Discover useful things AI bots can do, with real examples, source links and practical notes for Canadians.
503 jobs
463 source-backed + 45 original plansA photo library is difficult to navigate when images lack descriptions that explain their contents to someone who cannot easily inspect them.
Source action: Adds descriptive photo metadata; initial run partly completed before connection failure. Expected output: Proposed review artifact: Image-ID caption manifest with uncertainty and completion status.
Modules with little test coverage leave engineers uncertain about edge cases and failure behavior when making code changes.
Agent job: Read assigned low-coverage modules and create runnable tests covering ordinary, edge and failure conditions. Expected output: A test pull request for engineer review.
Shipping notifications arrive across separate emails, making it awkward to gather tracking numbers in one place.
Agent job: Extract tracking identifiers from authorized shipping notifications and prepare matched entries for the Parcel app. Expected output: A reviewable set of shipment-tracking entries with ambiguous or missing identifiers flagged.
Aging stock and replenished stock need different pricing attention, but inventory changes can be difficult to follow product by product.
Agent job: Use Amazon inventory data to adjust clearance prices and restore prices after replenishment within approved rules. Expected output: Product-price changes tied to inventory age and restocking.
A familiar commute becomes uncertain when platform, arrival or cancellation changes appear after the traveller has planned their journey.
Agent job: Check live train information for an approved route and observation window and notify the commuter of changes. Expected output: Notifications about platform, arrival or cancellation changes.
Spatial features without a reliable geographic reference cannot be safely compared with mapped layers until alignment choices and residual errors are checked.
Agent job: Fit a transformation using supplied control points and a target CRS, inspect residuals and warp vector features. Expected output: Georeferenced vector output and residual evidence for GIS review.
Recurring session errors and user corrections can expose weak skills, but the patterns are difficult to turn into specific repair proposals.
Agent job: Analyze failed sessions and correction signals, propose skill changes and record comparisons while respecting protected bundled skills. Expected output: Proposed patches or new skills and a comparison report.
A sales conversation can stall when a technical product question requires evidence hidden in the repository or internal knowledge base.
Agent job: Investigate the codebase or internal knowledge source and prepare an answer to a live technical question. Expected output: A customer-facing technical answer for human checking.
Community members need timely answers while moderators also need to understand recurring questions and recognize when approved guidance is insufficient.
Source action: Author deployed overnight community help and morning theme summaries; tone required two revisions. Expected output: Proposed review artifact: Reply candidates, escalation reasons and a deidentified theme summary.
Understanding current service health can require checking streams, storage and several operational interfaces before answering a simple status question.
Agent job: Query authorized SSH and service interfaces for current streams, storage and service status. Expected output: A concise answer grounded in current operational state.
Colleagues without technical data skills can struggle to retrieve workplace information from the tools that hold it.
Agent job: Connect a shared agent to approved company data tools and answer colleagues' questions through Slack. Expected output: Answers to workplace data questions in the shared channel.
Confirmed travel details remain scattered across booking emails unless someone keeps the shared itinerary up to date.
Agent job: Watch authorized booking emails and append confirmed trip details to an itinerary Sheet. Expected output: Updated itinerary rows containing confirmed booking details.
Document feedback creates extra coordination when the relevant comment, proposed wording and source paragraph must be reconciled by hand.
Agent job: Locate a specified Google Doc comment, reply to it and insert revised prose at the relevant location. Expected output: A comment reply and revised text in the source document.
Finding an item locally can mean repeating the same stock question to several stores and keeping track of their answers.
Agent job: Call approved stores about a specified product's availability and return the resulting answers and transcripts. Expected output: Store-availability responses and call transcripts.
Road-trip photos lose their narrative when they remain a loose collection rather than following the days and places visited.
Agent job: Organize supplied local photos by day and location and assemble a themed digital scrapbook. Expected output: A chronological digital travel scrapbook for review.
Application deadlines are difficult to manage when supporting documents are scattered, missing or in formats the application cannot use.
Agent job: Gather licensing or grant materials, identify missing documents and convert files for the team's review. Expected output: An assembled application packet with identified missing materials.
Preparing a location-specific weather brief requires bringing current conditions, severe outlooks and relevant model imagery together without losing geographic context.
Agent job: Check weather conditions and severe outlooks and retrieve explicitly requested model fields for a bounded location. Expected output: A geographically focused image set for a weather brief.
Product photos taken on a phone are not ready to share as a coherent seasonal catalogue without layout and assembly work.
Agent job: Assemble supplied product photographs into a product catalogue PDF for merchant review. Expected output: A proposed product-photo catalogue PDF for the merchant to review before sharing.
Quarterly kickoff preparation requires connecting strategy, previous results and backlog choices so the team can discuss priorities with shared context.
Agent job: Combine strategy documents, prior results and the backlog into an OKR kickoff deck. Expected output: A quarterly kickoff presentation for review.
A music release needs coordinated imagery, transitions and lyric timing, which can be cumbersome to assemble from separate creative assets.
Agent job: Use supplied audio, timing analysis, images and a reference video to build a synchronized video; upload only with authorization. Expected output: A music video with transitions and lyric effects, with any upload separately approved.
Valuation analysts must repeatedly reconcile cap-table figures with deal inputs while preserving a consistent memo structure for review.
Agent job: Extract cap-table figures and combine them with structured deal inputs in a valuation-memo draft. Expected output: A consistently formatted valuation memo for analyst verification.
Offer preparation can become inconsistent when approved compensation, templates, candidate correspondence and presentation material are assembled separately.
Agent job: Read approved offer inputs and templates and draft the offer, candidate email and HTML call deck. Expected output: An offer draft, candidate-email draft and call deck awaiting human review.
A new on-call rotation needs unresolved issues and watch-items without reconstructing paging events, escalation discussions and recent deployments from scratch.
Agent job: Read paging history, escalation channels, deployments and the previous handoff, clarify unresolved status and draft the next handoff. Expected output: A rotation handoff with statuses and watch-items, plus a staged channel post.
Open municipal network data still needs assembly and diagnostic review before an engineer can understand what a stormwater simulation can support.
Agent job: Assemble and run a SWMMCanada model from municipal data, audit diagnostics and flag limits of the requested dry-week analysis. Expected output: A stormwater model run and diagnostic findings that identify unsupported capacity conclusions.
Counsel can lose time reconstructing the conversations, deal details and prior guidance needed to assess a new legal request.
Agent job: Gather relevant conversation, deal and prior-guidance context and maintain an intake tracker for counsel. Expected output: An enriched legal intake record and tracker for attorney review.
Tax preparation becomes harder when records span documents and entity history, leaving transactions and summary figures needing reconciliation.
Agent job: Analyze source documents, reconstruct entity history and prepare summary and profit-and-loss workpapers with human-guided cross-checks. Expected output: A reconciled tax review packet and profit-and-loss workpaper.
New service customers must fit existing geography, technician capacity and travel time rather than simply being added to the next available route.
Agent job: Check route geography, technician load and drive-time estimates while onboarding service accounts and updating routing tasks within approval boundaries. Expected output: Capacity-aware route assignments and updated routing tasks for dispatcher review.
Passport renewal can involve a difficult online process while sensitive identity details and final attestations still require the owner's direct involvement.
Agent job: Assist with the online renewal session and hand browser control back for sensitive identifier entry and photo upload. Expected output: Browser-assisted renewal work with owner-handled sensitive entry.
A growing skill library can accumulate duplicate responsibilities and conflicting instructions that make shared maintenance harder.
Agent job: Inspect skills for overlap, conflicting scope and inefficiency and propose consolidation or rewrites for their owners. Expected output: A skill-library audit with proposed consolidations and revisions.
Brand and legal-copy inconsistencies can remain unnoticed when asset files and their live versions are reviewed separately.
Agent job: Check marketing images against brand and legal-copy rules, verify live versions and create tasks for high-confidence findings. Expected output: Documented asset findings and correction tasks.
Teams evaluating AI visibility need to distinguish actual observed citations from technical website issues and speculative claims about what might improve visibility.
Agent job: Record brand mentions and citations for specified questions, inspect relevant pages and rank observations and proposals by evidence and effort. Expected output: An evidence-tiered citation and technical-gap audit with proposed changes.
Vulnerabilities, outdated dependencies and licence concerns can be dispersed across services, making a consolidated review difficult.
Agent job: Inspect service packages for vulnerabilities, obsolete dependencies and licence issues and consolidate the findings. Expected output: A consolidated dependency security and licence-risk report.
A streaming playlist may contain tracks absent from an owned music library, making local listening gaps hard to identify.
Agent job: Mirror Spotify playlists in Navidrome and record tracks missing from the local library in Obsidian. Expected output: Mirrored playlists and a note listing missing local tracks.
A professional profile may obscure relevant experience when its wording and emphasis do not match the responsibilities of a target role.
Agent job: Compare a target job description, the user's profile and relevant public profiles, then draft positioning changes without editing the live profile. Expected output: Draft headline, summary and experience text with identified positioning gaps.
A local inference optimization needs comparable benchmarks and a reproducible setup before its tradeoffs can be assessed or reused.
Agent job: Apply a runtime cache optimization, benchmark alternatives and document and package the local inference setup. Expected output: Benchmark results, documentation and a reproducible inference package.
Unused flight credits can be awkward to combine when arranging a new trip and checking what remains payable.
Source action: Reports combining unused flight-credit codes to book a Southwest flight. Expected output: Proposed review artifact: Proposed credit allocation and booking-review checklist showing any unverified remainder.
Market-size discussions are difficult to review when assumptions, calculations and supporting sources are disconnected from the presentation.
Agent job: Research a defined market, calculate market-size scenarios and assemble linked presentation, workbook and source materials. Expected output: A market-sizing presentation, calculation workbook and cited source document.
Public competitor information is only useful for strategy when it is compared with the firm's own positioning and context.
Agent job: Research competitors' registry details, positioning and public evidence, then compare them with supplied company context. Expected output: A company-specific competitor comparison and recommendations for human judgment.
Learners can revisit familiar material without discovering which concepts they cannot recall or which misconceptions still need attention.
Agent job: Use concept-based recall assessment, misconception tracking and scheduled review with a separate assessor and approved persistent records. Expected output: Concept-level learning records, recall assessments and a review schedule.
Family-tree records become difficult to follow across generations, especially when conflicting parent-family links are hidden in the underlying data.
Agent job: Traverse supplied GEDCOM ancestry from a root person, surface parent-family conflicts and write a numbered report with optional alternate trees. Expected output: A generation-by-generation ancestry report based on supplied records.
Recruiters need to compare evidence against job requirements without losing important details across differently formatted resumes.
Agent job: Read local resume PDFs against a supplied job description and structure the relevant evidence for human review. Expected output: A job-requirement evidence matrix comparing applicant resumes.
An AI-answer feature needs implementation, usage tracking and billing visibility before a team can evaluate its behavior in production.
Agent job: Implement an answer API through a pull request and user-controlled deployment, then monitor usage, latency and billing. Expected output: An answer-feature pull request and usage, latency and billing observations.
Designers need a practical way to inspect and adjust motion details without repeatedly modifying the production asset itself.
Agent job: Build a local interactive playground around an existing animation specification and iterate through controls and designer feedback. Expected output: An interactive motion playground with adjustable animation behavior.
An educator's learning logic can be difficult to communicate through static material when the lesson calls for interaction, illustration and narration.
Agent job: Translate teacher-directed requirements into a gamified web experience with coded interactions, illustrations and narration. Expected output: A narrated interactive learning simulation for educator review.
Restaurant operations become fragmented when ordering, customer history, loyalty and point-of-sale reporting live in separate workflows.
Agent job: Develop a restaurant application integrating ordering, customer history, loyalty and POS-linked reporting through directed iteration. Expected output: An integrated restaurant application for operational review before deployment.
Security reviewers need to distinguish real repository boundaries and existing controls from assumptions before relying on a threat model.
Agent job: Review an authorized repository offline, trace entry points and enforcement boundaries, verify anchors and produce a scoped threat model. Expected output: A source-grounded threat model separating scenarios, controls and unresolved questions.
Repeated illustration requests can drift from an established visual style when reference choices and generation instructions are recreated each time.
Agent job: Analyze selected reference illustrations, save a reusable style skill and generate requested illustrations through an image-model call. Expected output: A reusable illustration-style skill and matching image candidates.
Turning circuit requirements into fabrication files involves component, pin-mapping and geometry checks that can leave costly ambiguities for an engineer to resolve.
Agent job: Clarify requirements, verify components and pins, edit KiCad files and review electrical, design-rule and geometry checks. Expected output: A PCB fabrication candidate, validation findings and sourced cost estimates for engineer approval.
Internal questions can span wiki, code and warehouse sources, while unanswered topics disappear unless someone records and routes them.
Agent job: Build an agent that searches approved internal sources, cites answers, routes uncertainty to owners and records unanswered topics. Expected output: A reusable internal knowledge agent with cited answers and gap tracking.
Running history stored in a connected service can be difficult to interpret as a coherent training view.
Agent job: Retrieve running history through Intervals.icu and use it to build a training dashboard after the documented Garmin-access failure. Expected output: A dashboard based on connected running history.
A valuation model can become disconnected from current market conditions when its discounting inputs do not reflect available yield-curve data.
Source action: Use entitled market data to construct a discounted-cash-flow model with current yield curves. Expected output: Proposed review artifact: Reviewable model draft with curve mapping, assumptions, sensitivities and unresolved methodology questions.
Style inspiration does not readily show which wardrobe gaps matter most or when suitable items appear on sale.
Agent job: Analyze an approved Pinterest board, research styling methods, prioritize wardrobe needs and send scheduled matching retailer links. Expected output: A prioritized wardrobe plan and matching sale-item recommendations.
Creating a brand-consistent email campaign requires assembling copy, product photography, artwork and an editable draft across several tools.
Agent job: Write brand copy, retrieve product photos, create Figma hero artwork and assemble an editable Klaviyo draft. Expected output: An editable email campaign draft for owner review and manual sending.
Event outreach becomes harder when regional contacts, opportunities and salesperson ownership must be gathered from separate CRM views.
Source action: Query CRM contacts and opportunities by region, create a target list, and distribute salesperson lists into event coordination. Expected output: Proposed review artifact: Proposed audience list with inclusion reason, account/opportunity link, responsible owner and unresolved exceptions.
Product teams need a consistent basis for discussing usability differences instead of relying on scattered impressions of competitor screens.
Agent job: Assess competitor screens and documentation against a supplied rubric, compare internal designs and save the scored review. Expected output: An evidence-based usability teardown in files or a workshop board.
Messaging decisions are weakly grounded when a buyer role's duties, pressures and buying participants are reduced to unsupported assumptions.
Agent job: Research a specified role and company type using public evidence and distinguish supported findings from gaps. Expected output: A sourced buyer-role persona, messaging angles and explicit evidence gaps.
After an incident, the sequence of events and unresolved responsibilities can remain scattered across messages, pages and local evidence.
Agent job: Reconstruct timestamps from incident sources and draft a postmortem, action owners and a channel note, clarifying uncertain judgments. Expected output: An incident timeline, postmortem and channel-note draft for review.
Facilities decisions across locations are difficult when office-use patterns cannot be compared in a reusable, accessible view.
Agent job: Transform office-pattern data into a dashboard for reviewing usage across company locations. Expected output: An office-usage analysis dashboard.
Website owners need to understand changes in traffic channels and engagement without repeatedly rebuilding the same analytics comparison.
Agent job: Compare weekly traffic-channel and engagement metrics and generate an HTML performance report. Expected output: A website performance report based on the connected analytics data.
A meeting-caption application requires native implementation, debugging and packaging before its translation and subtitle interface can be tried.
Agent job: Create, debug and package a Swift meeting translation and subtitle app on a Mac using approved terminal commands. Expected output: A packaged native meeting-caption application for review.
A browser-game idea needs a playable implementation and exercised interactions before its creator can judge how the gameplay works.
Agent job: Create a multiplayer browser game, launch it and exercise its gameplay. Expected output: A playable browser game and its live location.
Musical ideas need repeated arrangement changes before MIDI parts, vocals, fills and dynamics work together in an editable session.
Agent job: Create Ableton Live sets through an MCP bridge and revise musical elements in response to the creator's feedback. Expected output: An editable Ableton session with requested arrangement revisions.
Custom phone automations can be difficult to assemble when API behavior, Shortcut formatting and validation errors must all be resolved.
Agent job: Research and test APIs, generate Shortcut XML, correct validation errors and sign the native shortcut file. Expected output: A native Apple Shortcut file for inspection before import or use.
Spatial layers need coherent projection, symbology and layout choices before they become a readable map someone can assess.
Agent job: Validate spatial layers, choose a CRS and symbology, add labels and cartographic elements, export the map and check the output. Expected output: An exported cartographic map for human visual review.
Broad tool descriptions can make command discovery cumbersome when an agent only needs concise references for a particular command-line interface.
Agent job: Enumerate CLI help recursively and save available actions as lightweight Markdown references. Expected output: Compact CLI action references or plugins for later agent use.
A large book corpus can hide connections between themes that are hard to discover through isolated searches and excerpts.
Agent job: Explore an indexed book corpus, select and order thematic excerpts, add links and revise trails to match author preferences. Expected output: Linked reading trails through selected excerpts.
Restaurant menu details are difficult to compare or reuse when they are unstructured and the currency of listed prices is unclear.
Agent job: Find the restaurant's own menu, assess whether its prices are current and extract structured menu data. Expected output: Structured restaurant menu data for review.
Transaction analysis and pitches require comparable evidence to be extracted from deal documents and presented in a coherent form.
Source action: Review transaction documents, build comparable-company analysis, and prepare pitch materials. Expected output: Proposed review artifact: Comparable-company worksheet, extraction register and pitch outline with unresolved normalization questions.
App-store images need consistent layouts and localized text, but translations and screen captures can introduce fit and claim errors.
Agent job: Drive localized simulator builds, capture screens, translate captions, render store images and review text and visual fit at human gates. Expected output: Localized app-store screenshot sets for final approval.
Lessons from a difficult trip can be forgotten when the next itinerary is researched without those practical constraints.
Agent job: Record a trip debrief and apply the approved layover preference during research for the next travel plan. Expected output: A revised travel proposal reflecting the recorded trip lesson.
Visual consistency can erode after code merges when styling violations are not connected back to the design system.
Agent job: Audit recently merged pull requests for styling violations, create issue tickets and optionally propose corrective pull requests. Expected output: Design-drift tickets and optional fix pull requests for review.
An article's framing can be hard to assess when its language is separated from missing baselines and competing evidence.
Agent job: Extract article text, scan language and sourcing, research omitted context and prepare a report for human judgment. Expected output: A contextualized article analysis with relevant counterevidence and baselines.
Underwriting reviewers need to locate policy exceptions and unsupported assumptions without losing the spreadsheet cells and financial evidence behind them.
Agent job: Retrieve borrower financials and peer spreads, inspect the underwriting workbook and prepare a cell-referenced exception brief. Expected output: A source-linked brief for reviewed underwriting-model and memo updates.
Committing to a software idea is difficult when similar products and relevant community discussions have not been checked.
Agent job: Search software and discussion sources through the idea-check tool and present comparable projects and its heuristic competition signal before coding. Expected output: A comparison of existing projects and evidence for a human proceed-or-pivot discussion.
Design work can enter the queue without essential context, leaving the team to chase missing references before it can begin.
Source action: Gatekeeper agents inspect tickets for missing information such as Figma links before Asana intake. Expected output: Proposed review artifact: Per-ticket completeness findings with missing item, reason, evidence and a proposed clarification question.
Executable decision trees can conceal missing cases, contradictions or unreachable branches that are hard to detect through ordinary reading.
Agent job: Research verification methods and implement checks for omissions, contradictions and unreachable paths in a compiled guideline library. Expected output: Decision-logic checks and findings for qualified review.
A field-service report can appear plausible in isolation while conflicting with service history or the operator's quality rules.
Agent job: Compare incoming pool-service reports with historical records and QC rules, flag suspect readings or omissions and notify approved audiences. Expected output: Quality-control flags and scoped notifications for manager or technician review.
Airline promotion eligibility is difficult to track when qualifying flight activity and corporate-ID receipt evidence must be matched manually.
Source action: Matches flight activity with corporate-ID receipts to track promotional reward eligibility. Expected output: Proposed review artifact: Activity-to-receipt ledger with supported, uncertain and excluded progress entries.
A storage product can meet a capacity claim yet fail to fit the intended space once its actual dimensions are checked.
Agent job: Inspect product specifications for both required storage capacity and physical fit before selecting candidates. Expected output: Product candidates evaluated against capacity and space constraints.
A shared video link may be unusable for its intended viewer because playback or access conditions are unclear.
Agent job: Open the supplied video and inspect playback, access restrictions and identifying metadata without bypassing restrictions. Expected output: A playback and access-status check for the shared video.
Choosing suitable music and directing it to the right room can require repeated app interactions and preference corrections.
Agent job: Select music for the requested room and use the owner’s preference feedback to refine later choices. Expected output: A proposed room-specific music selection, with actual playback and device targeting checked separately.
A work trip's hotel choice must balance airport access and worksite travel, while checkout still requires the traveller's decisions and sensitive details.
Agent job: Research hotels against airport and worksite constraints and assist with the selected reservation while the owner handles sensitive checkout steps. Expected output: Hotel options and an approved reservation attempt with owner checkout involvement.
Choosing dinner becomes repetitive when available ingredients and recently cooked meals are not considered together.
Agent job: Rank cookable recipes from recorded ingredients and meal history, then log the confirmed choice and update ingredient quantities. Expected output: Ranked meal suggestions and approved meal-history and ingredient updates.
Shopping rewards can remain uncredited unless purchase receipts are matched to the programme's missing-credit claim process.
Source action: Uses purchase receipts to submit missing-mileage claims and save supporting PDFs. Expected output: Proposed review artifact: Claim candidates with supporting-document manifest, exclusions and unresolved eligibility checks.
Duplicate contacts, missing fields and inconsistent CRM values make it harder to understand accounts and identify stale opportunities.
Agent job: Identify and merge duplicates under approved matching rules, enrich missing fields, standardize values and flag stale opportunities. Expected output: Reviewed CRM record updates and stale-opportunity flags.
A pull-request queue can stall on scattered review comments, rebase needs and small fixes that nobody has coordinated.
Agent job: Coordinate pull-request comments, rebases and bounded fixes. Expected output: Updated pull requests for review.
Single-cell expression data needs a reproducible processing sequence before researchers can inspect cell clusters and the markers supporting them.
Agent job: Validate raw counts, filter cells, build a neighbourhood graph, cluster cells and identify markers using approved analysis choices. Expected output: Cluster and marker tables, figures and reproducibility metadata for expert review.
Incoming bills need consistent classification and payee preparation before a person can review the payment run.
Source action: Classify incoming bills and stage payees. Expected output: Proposed review artifact: Proposed coding and payee staging list with source references and unresolved exceptions.
Loan-file checks require evidence from several exclusion-list portals, while possible matches and exceptions still need a qualified person's assessment.
Agent job: Search approved OFAC, SAM.gov and HUD portals with authorized borrower data and collect structured findings and page PDFs. Expected output: Exclusion-list findings and supporting PDFs for a loan-file reviewer.
Missing receipts leave accounting transactions unresolved when nobody has a clear, source-specific request to send to the responsible cardholder.
Agent job: Identify missing receipts, group requests by cardholder and prepare export instructions linked to unresolved transactions. Expected output: Tracked evidence requests and source-specific instructions for accountant review.
Designers need relevant production references, but useful public imagery and approved studio information can be scattered across sources.
Agent job: Collect and organize public visual references and retrieve approved set information from the studio library. Expected output: An organized reference collection for designer review.
Upcoming trial readouts are difficult to track when sponsor, drug, phase and timing details must be gathered across calendar entries.
Agent job: Traverse a trial-results calendar for a specified date range and extract the listed study details. Expected output: A table of sponsors, drugs, indications, phases and expected readout dates.
Daily nutrition tracking becomes cumbersome when recovery data, activity and logged meals must be reconciled against the user's own targets.
Agent job: Combine WHOOP recovery, Apple Health activity and recorded meals with user-set targets, calculate remaining macros and suggest meals. Expected output: A daily nutrition check and meal suggestions for personal review.
A job applicant needs to connect existing experience to a specific posting without introducing unsupported claims into their application.
Agent job: Read the supplied CV and job posting, explain fit against a rubric and return a tailored cover-letter draft. Expected output: A role-fit explanation and cover-letter draft for applicant review.
Choosing between cash and reward flights requires comparing availability, balances, point values and transfer paths that may change the apparent value.
Agent job: Search cash and award options, compare transfer paths and valuations and inspect authorized loyalty balances without transferring or booking. Expected output: A cash-versus-points comparison for the traveller's decision.
A circuit simulation is hard to trust when its predictions have not been aligned with fresh, correctly normalized instrument measurements.
Agent job: Use connected oscilloscope and SPICE tools to compare circuit-model behavior with fresh measurements and align time axes. Expected output: A simulation-to-measurement comparison for engineer interpretation.
Company revenue trends are difficult to interpret without aligning them with relevant economic series over comparable periods.
Agent job: Extract revenue from board decks, obtain matching FRED series and combine them into a chart and summary. Expected output: A company-growth comparison chart and economic-context summary.
A multi-city trip can be awkward to compare when each flight leg has different baggage rules and provider details.
Agent job: Select nonstop options across the specified three legs, inspect baggage and provider details and stop before continuing the booking flow. Expected output: A connected itinerary comparison with flight and baggage details.
The earliest driving-test appointment may be at another office, but dates and travel distances must be checked together.
Agent job: Inspect specified offices in the supported appointment portal and rank earliest openings by date, time, address and distance before booking. Expected output: A distance-aware shortlist of appointment openings for user selection.
Researchers need a reproducible comparison between experimental groups without losing track of count quality, sample metadata and the chosen statistical contrast.
Agent job: Validate counts and metadata, apply the selected formula and contrast and export differential-expression results and replay files. Expected output: Differential-expression tables, PCA and result plots with reproducibility files.
Legal practitioners comparing jurisdictions need procedural differences organized with citations so they can return to the governing rules before relying on them.
Agent job: Organize jurisdiction-specific civil-subpoena rules and notice categories in a searchable, cited comparison dashboard. Expected output: A jurisdiction-comparison dashboard requiring primary-rule verification by counsel.
Finding a suitable nearby class requires checking actual age bands, schedules and prices rather than trusting brief directory listings.
Agent job: Visit provider websites and compare class ages, schedules and prices for the requested area. Expected output: A comparison of nearby classes against age and scheduling needs.
Unexpected network exposure can be overlooked when current hosts and services are not compared with previous observations.
Agent job: Scan the approved owned network, compare hosts and services with prior observations, save evidence and alert on differences. Expected output: Network-change evidence and exposure alerts for operator review.
The best order for several city stops can change with departure time, making an intuitive route a poor basis for planning.
Agent job: Compare alternative orderings of three stops using travel estimates for the selected departure time. Expected output: A stop sequence with the lowest estimated travel time among compared options.
Different bottle sizes and promotions make supermarket prices difficult to compare, especially when one retailer's site cannot initially be accessed.
Agent job: Compare retailer pack sizes and promotions, calculate unit prices and revise the ranking after any owner-handled access challenge. Expected output: A unit-price comparison identifying inaccessible or newly checked retailers.
Creators need to understand which thumbnail patterns accompany stronger channel results without assuming that visual correlation proves the cause.
Agent job: Retrieve channel results and thumbnails, compare visual patterns with channel averages and suggest creative alternatives. Expected output: A thumbnail-pattern analysis and suggested alternatives for creator review.
Veterinary reference preparation requires comparable toxin-treatment information in one place, with every entry checked before clinical use.
Agent job: Assemble a pet-toxin treatment comparison spreadsheet in Obsidian for independent veterinarian verification. Expected output: A toxin reference spreadsheet requiring qualified verification of every entry.
A LaTeX paper can be held up by syntax, bibliography and cross-reference problems that only become clear after compilation.
Agent job: Select a template, edit and compile LaTeX, inspect errors, repair syntax and preview the rendered PDF. Expected output: Compiled paper files and a PDF preview for author review.
Adding one missing pickup item should not undo an existing cart or switch the store the shopper has already selected.
Agent job: Find the missing painting tool, preserve the selected store, verify pickup stock and update the cart without checking out. Expected output: An updated pickup cart retaining the existing contents and store.
Turning a portrait into calendar art requires translating shapes and colours into a time grid without disturbing real appointments.
Agent job: Map the reference portrait to a weekly grid and calendar palette, create approved coloured events and inspect week view. Expected output: A calendar-block portrait and a week-view screenshot.
Multi-projector installations can look wrong even when software settings appear correct, because routing, geometry and physical alignment must agree.
Agent job: Route output windows, adjust surface warping and blending and audit display assignments, resolution and performance with physical operator checks. Expected output: A configured projector-output mapping and audit findings for on-site review.
A guardian's intended app restrictions can be tedious to apply consistently across children's devices and availability settings.
Agent job: Use authorized phone mirroring to configure the guardian's selected Screen Time app-availability rules. Expected output: Configured app-availability settings for guardian verification on the affected devices.
A photo collection remains vulnerable when local and off-site backup routines are not configured and their execution is hard to inspect.
Agent job: Configure requested daily drive backups and weekly S3 backups with monitoring and dashboard logs under approved access and retention settings. Expected output: Scheduled local and off-site photo-backup configuration with monitoring and logs.
Event planning stalls when attendance responses and unanswered invitations are not collected into a clear register.
Agent job: Call an approved guest list within authorized hours and purpose, record responses and unanswered calls and compile the register. Expected output: An attendance register showing responses and unanswered calls.
Mobile growth findings can remain disconnected from the ad concepts and product changes needed to test a team's next ideas.
Agent job: Examine acquisition and product analytics and pass findings into ad creative and engineering specifications. Expected output: Creative concepts and engineering specifications grounded in growth analysis.
School deadlines and schedule changes become hard for parents to follow when notices are spread across multiple portals.
Agent job: Bring authorized school notices into a proposed family overview and calendar-update list for guardian review. Expected output: A parent-review overview of homework, closures and meetings, with proposed calendar and notice changes.
Creating a themed crossword requires fitting answers into a workable grid while producing clues and a usable way to play.
Agent job: Select theme answers, construct a grid from a scored wordlist, write clues and assemble puzzle JSON and an offline player. Expected output: Structured crossword data and an offline HTML player for review.
A PDF book can be awkward to read as an ebook when chapter navigation, illustrations and endnote links do not transfer cleanly.
Agent job: Convert a text-layer PDF into a structured ebook, preserve illustrations, build navigation and linked endnotes and check the output. Expected output: An ebook with chapter navigation, cover, illustrations and linked endnotes.
Rough handoff notes can leave clinical context scattered, making it harder for a physician to prepare consistent documentation and spot missing information.
Agent job: Transform rough clinical notes into handoff, referral, patient-explanation and missing-information drafts for physician verification. Expected output: Clinician-reviewable documentation drafts and missing-information prompts, with release approval retained.
Engineering drawings supplied as PDFs need editable geometry and clear uncertainty flags before anyone can consider using them for technical work.
Agent job: Extract PDF geometry and text with flagged OCR fallback, interpret quality findings and package CAD outputs with previews and a report. Expected output: DXF, optional DWG, previews and a quality report requiring engineering review.
Listening history and track lists are less useful when they cannot be carried into the service where someone wants to listen.
Agent job: Match ListenBrainz history or a supplied track list to Tidal, supporting a dry run and playlist-transfer export. Expected output: Playlist matches, a transfer export or a destination playlist.
Notes and transcripts can contain relationships that are easier to explain visually, but the author still needs editable control over the diagram.
Agent job: Convert explanatory notes or transcripts into an editable Excalidraw diagram for the author's final adjustments. Expected output: An editable explanatory-diagram draft that the author can inspect and adjust in Excalidraw.
Researched ideas need different wording and layout for a visual carousel, creating extra work after the article itself is written.
Agent job: Research the topic, write an article and transform its key points into an editable Canva carousel. Expected output: An article and editable carousel design link for review.
An evening plan can fall apart when dinner and entertainment are booked without checking whether the timing and travel distance work together.
Agent job: Propose dinner and comedy options, revise the plan after feedback and make only the explicitly approved reservation and ticket purchase. Expected output: A coordinated evening plan and approved booking or ticket-purchase attempts.
Returning a large retail purchase can involve customer-service coordination and pickup scheduling beyond simply requesting a refund.
Agent job: Coordinate the requested retailer return and pickup arrangement within approved order details, fees and collection times. Expected output: A return arrangement and scheduled pickup for owner review.
Understanding internal AI adoption requires connecting employee cohorts, interview logistics and transcript analysis rather than relying on isolated anecdotes.
Agent job: Identify adoption cohorts, coordinate interviews and analyze transcripts to extract employee practices. Expected output: An organized internal research process and findings on AI-use practices.
Compiler research demands coordinated implementation work and repeated comparison with executable tests before anyone can assess what the compiler handles.
Agent job: Coordinate parallel agents implementing a Rust C compiler through Git task ownership and iterative compiler and GCC-based tests. Expected output: Compiler code and test-comparison results for research review.
Reproducing a research paper requires tracking claims, failed experiments and proposed fixes across several interdependent lines of work.
Agent job: Coordinate paper reading, claim checks, reproduction attempts, fixes and rerun experiments within approved scope and compute limits. Expected output: Experiment results and a sourced research brief.
Site work can be delayed or disrupted when trades attend in an order that does not match the work's dependencies.
Agent job: Use the authorized client text line and approved partner contacts to coordinate trades in the required site sequence. Expected output: Coordinated trade attendance arrangements for site-lead review.
Property owners can miss important notices when an official mailing record still points to an old address.
Agent job: Trace misdirected property mail and prepare address-correction paperwork. Expected output: A correction-request review packet with property, authority and address evidence.
Slow-moving inventory needs attention, but turning product data into relevant promotional materials requires additional creative and offer setup work.
Agent job: Identify slow-selling products and generate related advertising creatives and promotional codes. Expected output: Product-focused promotional creative and codes for review before use.
Several recurring responsibilities can generate fragmented updates unless specialist routines are organized into a shared daily overview.
Agent job: Use supplied personal context to create five specialist roles, weekday routines and a consolidated brief while retaining existing bots. Expected output: A scheduled specialist-agent setup and consolidated chief-of-staff brief.
A menu can be difficult to navigate with a screen reader when categories and dishes lack clear heading structure.
Source action: Builds shareable menus with navigable headings for categories and dishes. Expected output: Proposed review artifact: Menu-page candidate outline with heading hierarchy, item/price associations and proposed navigation checks.
Selling an unused item requires turning supplied details into a marketplace listing before any buyer discussion can begin.
Agent job: Create a listing for the supplied item through authorized marketplace access, keeping negotiation and commitments within separately granted instructions. Expected output: A marketplace item listing with approved content.
An untagged or poorly structured PDF may be difficult to navigate, while automated tagging alone cannot establish meaningful accessibility.
Agent job: Preserve the original, classify and tag a working copy, inspect marked content, use approved OCR or fallback steps and stage a candidate. Expected output: A tagged PDF candidate and status entry for human accessibility checking.
Brand direction can be interpreted inconsistently in implementation when design tokens lack clear types, valid references or checked contrast.
Agent job: Translate brand direction into a typed design specification, check structure and contrast, repair references and export token files. Expected output: Validated design-token files for implementation review.
Event details in images and PDFs need to be copied into a calendar, creating opportunities to miss a date, time or location.
Agent job: Extract event details from supplied images or PDFs and create calendar entries within the approved scope. Expected output: Calendar events based on extracted document details.
Preparing lesson variants requires preserving shared learning objectives while adapting materials consistently across a unit's slides and documents.
Agent job: Read source-page images and objectives, produce matched slide and document variants and reuse project rules across the unit. Expected output: Differentiated lesson files aligned with supplied learning objectives.
A dividend watchlist can become stale when spreadsheet figures are not checked against current yield sources and their dates.
Agent job: Read the ETF tracking sheet, cross-check latest yields and post an approved daily summary. Expected output: A yield cross-check and daily watchlist summary for investor verification.
A product team needs to understand how game mechanics relate to market-specific requirements without losing the supporting rule sources.
Agent job: Query the approved knowledge source, compare game mechanics with market requirements and prepare a sourced briefing. Expected output: A sourced product-behavior and market-rule comparison for qualified review.
Engineers need help diagnosing difficult customer PDFs without exposing protected files outside the secure environment where they are held.
Agent job: Generate diagnostic tools for engineer execution in a secure sandbox, inspect returned non-sensitive structure and iterate on proposed fixes. Expected output: Diagnostic tools and code-fix candidates based on redacted structural evidence.
A brick-building idea needs a coherent part inventory and assembly order before it can be turned into understandable instructions.
Agent job: Create build.json, check collisions, support, order and inventory, then render and visually review an instruction booklet. Expected output: A checked digital build specification and HTML or PDF instruction booklet.
Documentation can fall behind merged API, CLI and configuration changes because the missing updates are not visible in the code review alone.
Agent job: Compare recently merged interface and configuration changes with accompanying documentation and report gaps. Expected output: A report of documentation updates missing from merged changes.
A recurring company metric can quietly disappear between disclosures, requiring exact-source comparison to distinguish omission from a mistaken reading.
Agent job: Compare successive earnings releases for recurring metrics and verify omissions against source lines and filing identifiers. Expected output: An evidence-linked account of missing metrics for human review.
A mathematical argument needs multiple forms of scrutiny when finite examples, adversarial objections and formal verification must be reconciled.
Agent job: Coordinate mathematical investigation, finite search, adversarial review and Lean formalization under the researcher's direction. Expected output: A proposed argument with proof sources and mechanical verification records.
Video scenes can lose visual continuity when generated across different models without shared story context and deliberate selection.
Agent job: Maintain story context, generate scene images, compare low-resolution clips across models and hand selected material to the creator's editor. Expected output: Scene images and selected video clips for final human editing.
Devotional writing needs a coherent study structure while quotations, interpretation and denomination-specific framing still demand the author's judgment.
Agent job: Develop devotional drafts and sermon outlines with the requested framing and export structured study notes. Expected output: Devotional drafts, sermon outlines and structured notes for author review.
Missing product snippets can leave an ecommerce team unsure whether its product schema needs correction or where the problem originates.
Agent job: Diagnose product-schema issues, propose a fix, back up the client site and apply only approved schema changes. Expected output: A schema diagnosis and approved site patch with a backup.
Storage warnings are hard to resolve safely when caches, logs, open deleted files and container data obscure what is consuming space.
Agent job: Survey storage and inode use, classify cleanup targets, perform explicitly approved cleanup and compare mount usage afterward. Expected output: A cleanup-target assessment and before-and-after storage comparison.
Handwritten production-schedule changes are easy to overlook when the clean digital version no longer reflects the annotated paper copy.
Agent job: Convert photos of marked-up production schedules into digital files and inspect the finished output. Expected output: Updated digital production schedules for coordinator verification.
A useful reading list needs working feeds, but relevant websites do not always expose obvious or valid RSS endpoints.
Agent job: Find relevant sites, verify that they expose real RSS feeds and select candidates matching the reading profile. Expected output: A verified list of candidate RSS sources.
A project breakdown is difficult to act on when approved responsibilities have not been placed where each role owner tracks their work.
Agent job: Break down the project, obtain the owner's review and distribute approved tasks into the appropriate role owners' pages. Expected output: Approved project tasks organized on their owners' pages.
Control reviews are difficult when walkthrough notes do not clearly connect process steps, responsibilities and potential segregation gaps.
Agent job: Read walkthrough notes and reference files and draft a process narrative, control matrix and flowchart with flagged gaps. Expected output: A process narrative, control matrix, flowchart and control-gap findings.
An approved code change still leaves engineers needing to decide which production behavior and failure signals deserve attention after release.
Agent job: Read an approved pull request and produce a production-monitoring plan for engineers to consult before deployment. Expected output: A change-specific monitoring plan for release review.
Writing a personal biography requires gathering public facts into a coherent narrative without letting unsupported claims enter the draft.
Agent job: Research the requesting person's public information and prepare a biographical draft for factual review. Expected output: A personal biography draft based on public research.
Interview documentation can be inconsistent when transcript details and the interviewer's actual judgment are not mapped into the required role form.
Agent job: Map the transcript to the role form, ask the interviewer for their assessment, revise the record and submit only after approval. Expected output: An interview record reflecting human assessment, staged for approved ATS submission.
Unsigned proposals can sit unnoticed when follow-up depends on remembering which opportunities have stalled and what was last promised.
Agent job: Check proposal status each business morning and draft follow-up messages for proposals that remain unsigned. Expected output: Proposal follow-up drafts for owner editing and sending.
Completed work is spread across task and time records, leaving billing staff to reconstruct what belongs on each client invoice.
Agent job: Review ClickUp and Harvest status, history and completed work, then prepare invoice drafts and a client work summary. Expected output: Invoice drafts and client work summaries for human review and manual sending.
Requesting an appliance repair is harder when the relevant lease context must be found and expressed clearly in correspondence.
Agent job: Prepare appliance-repair correspondence using the owner’s request and the relevant supplied lease context. Expected output: Repair-email drafts with factual questions and any lease interpretation reserved for owner or qualified review.
Upcoming equipment maintenance creates quoting work when prior service dates, customer contacts and estimate details are spread across business records.
Agent job: Query upcoming maintenance and service history, save an estimate in QuickBooks and attach it to a templated reminder-email draft. Expected output: A maintenance estimate and attached reminder-email draft for manual review and sending.
Regulated requirements are difficult to review when acceptance criteria are disconnected from the code and evidence used to test them.
Agent job: Read source code and Jira records and draft User Requirement Specifications with testable criteria and testing evidence. Expected output: Structured requirements drafts with code and test traceability for senior-engineer review.
Release owners need to turn scattered merged changes into an accurate account of what users should know.
Agent job: Group merged pull requests from a selected period, add contributor credits and prepare source-linked release notes for review. Expected output: A categorized release-note draft with change links and contributor credits.
Retail operators need to anticipate replenishment needs without manually translating changing sales demand into supplier orders.
Source action: Monitor stock demand and draft supplier reorders. Expected output: Proposed review artifact: Replenishment draft with SKU, quantity, unit, supplier, assumptions and exception reasons.
A visual idea can be difficult to express when creating each shape and brushstroke requires unfamiliar desktop drawing controls.
Agent job: Interpret a requested scene, draw it with Paint controls, inspect the canvas and save the artwork and a screenshot. Expected output: A PNG artwork file and a screenshot for the requester to review.
Creators with a narrated recording still need supporting visuals, captions and repeated edit passes before they can review a finished video.
Agent job: Find supporting footage, assemble B-roll and captions in Descript, revise treatments with creator feedback and export the edit. Expected output: A creator-reviewed video export incorporating the requested production revisions.
Processing several recordings separately makes it hard to track unfinished files, failed effects and requests to stop work.
Agent job: Submit approved recordings for requested enhancements, monitor their shared batch and report each file's outcome while respecting pauses and cancellations. Expected output: Completed media output paths alongside disclosed incomplete or cancelled items.
Materials researchers need comparable temperature-dependent transport estimates without losing sight of simulation convergence and uncertainty.
Agent job: Coordinate multi-temperature molecular dynamics, assess convergence and fit an uncertainty-aware Arrhenius relationship. Expected output: Diffusivity and activation-energy estimates with convergence checks and uncertainty information.
A promising crystal structure may still decompose into competing phases, making an isolated energy calculation insufficient for assessing stability.
Agent job: Retrieve competing structures, relax candidates consistently and construct a convex-hull stability and decomposition analysis. Expected output: A candidate stability and decomposition analysis, with electrochemical limits where requested and supported.
An agent release can contain many changes, leaving its operator unsure whether an upgrade is relevant to their installed version.
Agent job: Compare merged release changes with the installed agent version and recommend whether an upgrade merits consideration. Expected output: A version-comparison summary and upgrade recommendation, without performing the upgrade.
Authors with useful instructional material need help expanding and arranging it without losing their examples, guidance or branding.
Agent job: Organize and expand supplied instructional materials into an ebook draft, splitting work where context limits require it. Expected output: A formatted instructional ebook draft awaiting the author's claim checks, branding review and final cleanup.
Price charts alone leave readers searching across company announcements and news to understand possible explanations for market moves.
Agent job: Combine price data, company events and news into a source-traceable interactive chart annotated with potential catalysts. Expected output: An annotated market chart with traceable supporting data for review.
Moving an assistant between platforms risks losing useful identity settings and schedules or accidentally carrying secrets into the export.
Agent job: Package agent identities and schedules into a secrets-free archive for an owner-managed platform move. Expected output: A reviewable configuration archive for a separately authorized import, without assuming cross-platform compatibility.
Product researchers need a usable set of matching certifications without copying records individually or losing their source references.
Agent job: Search the selected certification portal, collect fifty matching records with links and check the exported file. Expected output: A checked CSV of matching certification records and source links.
Creators need consistent post-level performance records when visible engagement figures are otherwise scattered across their own profile.
Agent job: Collect thirty posts from an authorized profile and export visible engagement counts and post links. Expected output: A CSV of owned-post metrics and links for further analysis.
Readers of municipal finances need key balance-sheet figures without manually searching through a lengthy public annual report.
Agent job: Locate the municipality's public annual report and extract assets, liabilities and fund balance with the PDF citation. Expected output: A source-linked extraction of the requested municipal balance-sheet figures.
A useful passage in a long podcast can be difficult to isolate into a clip for review or reuse.
Agent job: Locate the user-named segment in a long recording and extract a clip for boundary and content review. Expected output: An extracted clip for the user to check before any separate publication decision.
A slide deck trapped in a web flipbook is difficult to search, quote accurately or examine outside its viewer.
Agent job: Navigate the presentation page by page and recover text and figures with slide numbers. Expected output: A slide-numbered extraction of the deck's text and figures.
Accounting teams need contract details in a consistent intake format while ambiguous terms and exceptions remain visible for review.
Agent job: Read signed contracts against accounting policy, extract revenue-recognition fields and link clauses while flagging non-standard terms. Expected output: An intake workbook with clause references and a review queue for ambiguous or exceptional terms.
People trying to avoid a crowded gym need to reconcile popularity patterns with the hours it is actually open.
Agent job: Read weekday popularity charts and opening hours to identify potentially quieter visit windows. Expected output: Suggested gym visit windows grounded in the displayed popularity and opening-hour information.
Finding an appointment before a deadline requires checking provider availability against existing commitments and personal time preferences.
Agent job: Search the requested practice, check the deadline, calendar and time preferences, and select and book a suitable slot. Expected output: A suitable appointment booking with details for the user to verify; a separate slot-approval step is not established by the source.
A signup problem that appears only at certain screen sizes can escape routine checks and leave developers without a concrete fix candidate.
Source action: Click-testing found a screen-size-specific signup defect and produced a fix PR. Expected output: Proposed review artifact: Reproduction checklist and bounded fix-review proposal with regression scenarios.
Potentially unclaimed funds are easy to overlook when records are scattered across official databases and identity matches remain uncertain.
Agent job: Search the requested unclaimed-property databases and identify candidate credits for the named person. Expected output: A list of candidate credits requiring official identity and eligibility checks before any claim.
Marketers need evidence of competitor search coverage gaps before deciding which missing topics deserve content investment.
Agent job: Use AIsa to compare the named sites' keyword coverage and seek approval when the quoted data cost exceeds the estimate. Expected output: An SEO gap report with competitor keyword evidence for editorial review.
Support teams see recurring customer confusion but may struggle to turn individual tickets into a prioritized list of missing help content.
Agent job: Scan incoming support tickets against company knowledge and suggest documentation gaps for human validation. Expected output: A reviewable set of help-document suggestions linked to the support context, without assumed issue creation or publication.
Small businesses need to identify plausible funding or bid opportunities before spending effort on applications that may not fit.
Source action: Find suitable grants or bids and draft applications. Expected output: Proposed review artifact: Eligibility screen, requirement-to-evidence matrix and application draft with unresolved fields visibly marked.
A reimbursement can appear unresolved when confirmation is buried in old correspondence, leading someone to pursue an unnecessary claim.
Agent job: Search the authorized airline correspondence for dated evidence showing whether the disputed fees were already reimbursed. Expected output: A source-linked account of the reimbursement evidence for the user to verify before taking further action.
Researchers need lawful access to cited papers without wasting time on broken links or mistaking abstracts for full text.
Agent job: Follow scholarly citations through publisher sites and university repositories to locate accessible full-text copies. Expected output: A list of lawful full-text locations, with unavailable papers clearly distinguished.
Responding thoughtfully to newsletters requires connecting an article with relevant experience that may be buried in past work records.
Agent job: Read incoming articles, search the author's stored work context and save possible personal-experience connections for review. Expected output: A private report of relevant response angles for the author to turn into their own comments.
Intermittent CI failures create repeated investigation work when teams cannot easily distinguish flaky tests from consistent defects.
Agent job: Review failed CI history, cluster signatures, check intermittent outcomes and rank a source-linked remediation queue. Expected output: A ranked flaky-test investigation queue with supporting run and test-path evidence.
People seeking public records can lose time approaching the wrong department or using an outdated request channel.
Agent job: Locate the official municipal records-request route and verify department contacts and legal-guidance links. Expected output: A verified request-channel summary with relevant official contacts and guidance links.
Researchers exploring a biomarker need trial details that distinguish relevant protocols from superficially similar study descriptions.
Agent job: Inspect trial protocols for the specified blood biomarker and record sponsor, phase and recruitment status. Expected output: A source-grounded shortlist of matching trials with their identifying study details.
Creators need to separate unusually promising topics from videos whose large view counts mainly reflect an already large audience.
Agent job: Research recent niche videos, compare views relative to channel size and summarize topic opportunities for creator selection. Expected output: A topic-opportunity summary with video metrics and comparison context, optionally refreshed on the requested schedule.
Finding a stock image that fits a layout requires checking dimensions, orientation and attribution rather than choosing on appearance alone.
Agent job: Search Pexels with the requested filters, present source metadata and download the selected image variant. Expected output: A selected image asset accompanied by source and attribution information for usage-rights review.
Travel planning can overlook tentative commitments discussed in messages when those commitments have not yet reached the calendar.
Agent job: Compare a proposed departure with authorized message context and flag a conflicting tentative commitment. Expected output: A travel-conflict warning with the relevant context for the user to resolve before changing plans.
Audio collectors need to catch questionable encoding before spending time tagging files or adding them to a trusted library.
Agent job: Run the existing packet and spectrogram analysis workflow and report audio readiness before manual tagging or import. Expected output: An audio-quality readiness report with flags for the owner to review.
Repeated payment failures or dwindling service credits can be buried in email until they disrupt tools someone relies on.
Agent job: Review connected billing notices and alert the owner to repeated payment failures or low remaining credits. Expected output: A billing-risk alert pointing the owner toward official account verification and user-controlled payment action.
Prospects can be left waiting when requested marketing material is hard to locate or its follow-up falls through.
Agent job: Locate marketing material requested by a prospect, email it and follow up through the connected business tools. Expected output: A prospect-facing collateral email and follow-up correspondence for the requested material.
Employees preparing a self-review need concrete evidence of their work rather than relying on memory or a blank competency template.
Agent job: Turn the competency spreadsheet into a role-aware interview, use completed Jira work as prompts and export the employee's evidence. Expected output: A reviewable evidence collection and development report for the employee's self-review.
Engineering teams repeatedly translate transformation specifications into functions, where mapping errors can spread into downstream data processing.
Agent job: Generate AWS Lambda data-transformation code from specifications for automated and human engineering review. Expected output: Candidate transformation functions awaiting mapping tests, access review and separate deployment approval.
Event organizers want invitation choices that reflect the same details without repeatedly explaining the brief for each design.
Agent job: Collect event information through the reusable intake skill and generate three invitation designs with editable links. Expected output: Three PNG invitation previews and corresponding editable designs for the user to select and revise.
Repeating a master document across many recipients invites mapping mistakes and makes missing information difficult to spot.
Agent job: Propose variable mappings, obtain confirmation and generate a draft file for each spreadsheet row while flagging missing values. Expected output: A confirmed mapping and a batch of draft documents with visible missing-value reports.
Vendor evaluations stall when reviewers must gather capability evidence and complete a long questionnaire from scratch.
Agent job: Research vendor capabilities, prefill the structured questionnaire and walk the reviewer through each answer for approval. Expected output: A researched questionnaire with answers ready for individual reviewer confirmation.
People who send voice messages need those recordings converted into usable agent input without manually transcribing each request.
Agent job: Use the supplied interface documentation to add a Telegram-audio transcription hook and return the resulting text to the agent's task loop. Expected output: Version-controlled integration changes for review and testing with the approved transcription service.
Trade-event attendees need to identify exhibitors that actually match their buying requirements before spending time on irrelevant booths.
Agent job: Inspect exhibitor pages for product-category and industry fit and gather relevant contact links. Expected output: A shortlist of matching exhibitors with contact links and fit information.
Recruiters need current evidence of relevant technical work when conventional profiles reveal little about a candidate's recent contributions.
Agent job: Watch selected repositories for recent contributors and compare public work with the supplied role brief. Expected output: A sourcing digest of potential candidates for recruiter review, with outreach kept separate.
Launching paid access requires billing configuration and application permissions to agree, with a way to check the purchase flow safely.
Agent job: Configure billing and gated access, implement the changes through a PR, deploy after user review and merge, and verify a test charge. Expected output: A billing implementation PR and test-charge verification results for the requested subscription flow.
Evolutionary analysis needs a reproducible path from raw sequences to supported relationships rather than an unexplained tree image.
Agent job: Validate, align and trim sequences, select a model, infer a tree and export support information and replay materials. Expected output: A Newick tree, support tables and a replay bundle for expert inspection.
Researchers examining a remote museum specimen need consistent viewpoints and annotation context that a single catalogue image cannot provide.
Agent job: Open the specified interactive specimen viewer, inspect requested angles and annotation pins, and assemble images with curator metadata. Expected output: An inspection brief with selected specimen views and curator information for specialist verification.
Server owners need to understand security weaknesses and configuration faults before changing settings that could also lock them out.
Agent job: Audit the test server, investigate firewall, SSH and proxy issues, apply configuration repairs and document the work. Expected output: A server audit, configuration repairs and documentation for owner inspection; per-change approvals are not documented.
Advertisers cannot interpret campaign performance confidently when conversion events may be misconfigured or missing from their destination.
Agent job: Configure the authorized Tag Manager events and test whether conversions reach the intended reporting destination. Expected output: A conversion-tracking configuration and destination-check findings for review.
Project owners may not know which outside people retain access to shared files as partnerships and staffing change.
Agent job: Inspect the authorized partner folder and inventory external addresses with access. Expected output: A spreadsheet of external file-access entries for owner review, without implied permission changes.
A declined travel payment can leave someone unsure whether a card problem or an incomplete booking needs attention.
Agent job: Connect payment and card notices, give activation instructions, observe later payment confirmation and add flight details to the calendar. Expected output: A payment-status explanation and calendar flight details, with actual ticketing still requiring verification.
Account owners need to notice changing usage and escalation patterns without repeatedly rebuilding charts from scattered customer signals.
Agent job: Watch account-usage and escalation signals, regenerate account charts and save them in the shared tracker. Expected output: Updated account-health charts and findings for human customer-strategy decisions.
Decision-makers face conflicting figures across trackers, messages and dashboards and need to understand which evidence supports a board recommendation.
Agent job: Trace project evidence across authorized sources, reconcile apparent discrepancies and prepare decision materials for authoritative-system verification. Expected output: A presentation, data appendix and executive brief explaining the reconciled evidence and remaining checks.
Infrastructure alerts demand context from logs and deployments before an operator can choose a safe response.
Agent job: Monitor service signals, investigate logs and prepare scoped infrastructure changes through the established human-review process. Expected output: An incident investigation and reviewable remediation changes with material actions awaiting operator approval.
Unexpected shifts in ad spending or conversions can be obscured by routine reporting, delaying attention to meaningful exceptions.
Agent job: Compare channel spend and conversions, investigate anomalies and report exceptions without changing budgets unless explicitly approved. Expected output: An anomaly alert with supporting spend and conversion context for the account owner.
Designing printable mechanical parts requires repeated geometry adjustments and visual checks before a physical assembly can be attempted.
Agent job: Write and revise FreeCAD Python geometry, inspect rendered iterations and export parts for human-controlled printing and assembly. Expected output: CAD geometry and printable exports for tolerance, fit and mechanical-safety review.
Legal teams can lose track of approved template versions when documents evolve across multiple repositories.
Agent job: Track legal templates, identify version changes and check whether the inventory stays aligned with the repositories. Expected output: An updated template inventory and change findings for counsel and document-owner review.
A local business needs a coherent path from search traffic to captured enquiries, including tracking and duplicate handling.
Agent job: Build a landing page, duplicate-protected lead capture, a geographic search-ad campaign and a consent-mode tag. Expected output: A landing page, lead-capture flow and search-campaign configuration with consent-mode tracking.
Owners trying to recover an item need readable location evidence and may need to compare its positions over time.
Agent job: Read the selected owner-registered item's Find My location and optionally compile timestamped UI observations into a position history. Expected output: A current location observation or approved position-history summary for the owner's property.
Keeping a nutrition journal becomes tedious when meal descriptions and photos must be converted manually into quantities and estimated nutrients.
Agent job: Interpret supplied meal entries, look up nutrition values, apply quantities and record reviewable estimates in the approved journal. Expected output: A nutrition journal with estimated entries for the user to correct, using approved health-data storage.
People recording meals and symptoms need consistent timestamps and a cautious view of patterns without treating coincidence as a diagnosis.
Agent job: Turn supplied meal and symptom messages into timestamped local records, run requested check-ins and prepare a weekly pattern review. Expected output: A timestamped wellbeing log and a weekly review of cautious correlation observations.
Trading preparation can become inconsistent when market inputs, candidate plans and past lessons are scattered across tools.
Agent job: Refresh market inputs, maintain the preparation board, record and challenge candidate predictions, and stage draft plans for human review. Expected output: A current preparation dashboard, reviewed candidate plans and lesson log, with trading decisions and execution kept with the user.
Launch teams need one view of blockers and ownership when readiness information is spread across specifications, issues and discussions.
Agent job: Cross-check launch documents, issue status, discussions and retrospectives, then record owner-tagged blockers and readiness changes. Expected output: An updated launch-readiness checklist with assigned blockers and changes for team review.
Project requirements drift as meetings add detail, leaving teams unsure whether the main document reflects the latest agreed inputs.
Agent job: Draft and revise the business-requirements document using meeting transcripts, project inputs and the supplied formatting conventions. Expected output: An evolving requirements draft for stakeholders to check against project decisions.
Teams producing branded work need reusable guidance that reflects current documents, design tokens and component-library conventions.
Agent job: Read approved brand sources, create a reusable brand-rule skill and rebuild it periodically with a change summary. Expected output: A reviewable brand-rule skill and update summary for brand-owner validation.
A recurring publishing schedule requires coordination between topic research, drafting, factual review and distribution that can otherwise become fragmented.
Agent job: Run scheduled topic research, drafting, SEO and fact review, Ghost publication and social distribution. Expected output: Content drafts, review findings and publication/distribution outputs; the source's precise human publishing gate is unclear.
Saved research becomes hard to reuse when bookmarks lack useful tags or their original relevance has been forgotten.
Agent job: Read suggested pages, save contextual tags, retrieve prior research and classify authorized untagged bookmarks. Expected output: A more searchable research library with contextualized bookmarks and retrieved references.
Useful work knowledge disappears into separate conversations, code changes and documents, making prior context difficult to recover.
Agent job: Ingest scoped work sources on schedule into a personal wiki, entity graph and retrieval index. Expected output: A searchable personal work wiki and index assembled from scoped sources.
Writing assistance can feel inconsistent when it lacks examples of how someone actually communicates and forgets their deliberate style rules.
Agent job: Read approved writing samples, create a personal voice skill and merge later patterns while preserving manual rules. Expected output: A reviewable writing-style skill grounded in supplied or authorized sent-message examples.
Finance teams repeatedly explain period changes while important variances are hidden across actuals, budgets and ledger detail.
Agent job: Compare close-period figures with prior results and budget, investigate material movements and prepare recurring commentary. Expected output: Variance commentary or a connected reporting artifact with figures and explanations for finance review.
Travel reminders become unreliable when itinerary changes do not reach every traveller or the people expecting updates.
Agent job: Record trip details, regenerate affected reminders and route bounded itinerary changes to the selected travellers and watchers. Expected output: Updated trip reminders and authorized itinerary notices reflecting the revised travel context.
Understanding a device market requires connecting product, company and regulatory claims that are scattered across unrelated sources.
Agent job: Gather company, product, regulatory and partnership information and connect it in a claim-cited market map. Expected output: A device-market map with source references for its substantive claims.
Researchers translating a question into biobank variables need to locate relevant fields without confusing schema descriptions with access to participant data.
Agent job: Search indexed public UK Biobank schema descriptions and rank field identifiers against the research question. Expected output: A ranked field list with relevance scores for researcher verification and separate data-access planning.
Readers following several research newsletters need to compare arguments and open questions without losing the evidence behind each claim.
Agent job: Read authorized subscribed articles and build a dashboard connecting their claims, evidence and unresolved questions. Expected output: A source-grounded research dashboard for comparing arguments across the permitted reading set.
Policy owners need to understand which new regulations may affect existing policies and products before deadlines are overlooked.
Agent job: Compare newly supplied regulations with indexed internal policies and product descriptions, then prepare applicability and action materials. Expected output: An applicability brief and deadline-bearing action table for qualified policy or legal review.
Design and engineering teams need trustworthy links between published components and implementation code without silently mapping the wrong properties.
Agent job: Resolve published Figma components, inspect code interfaces, confirm matches with the user and create checked parserless Code Connect templates. Expected output: Reviewable component templates with property, variant and nested-instance mappings; publication remains a separate decision.
Reaching a target company is easier to assess when decision-makers and possible mutual introductions are visible in one place.
Agent job: Identify decision-makers at the named company, inspect account connection information and rank potential introduction paths. Expected output: A table of prospective introduction routes and mutual contacts for review before any outreach.
Sellers need credible customer references that fit a prospect's situation rather than repeatedly asking colleagues for familiar names.
Agent job: Retrieve the prospect's CRM context, compare approved reference customers and rank suitable matches with explanations. Expected output: A ranked reference-customer shortlist for seller review and separate consent checks before contact or disclosure.
Support teams can miss an opportunity to close a customer loop when merged fixes are not connected to the original issue.
Agent job: Match staging-merged fixes with relevant customer issues to help staff identify possible follow-up needs. Expected output: Candidate fix-to-customer matches requiring release-availability checks and message approval before notification.
Replacing a legacy batch workflow is difficult when its essential behavior must be preserved across a new language and runtime.
Agent job: Reimplement the specified batch process in Python for AWS Lambda and iterate against supplied expected outputs. Expected output: A candidate migrated batch implementation with output comparisons for review before production use.
Moving an existing website archive into local content can expose broken markup, encoding problems and lost metadata.
Agent job: Retrieve the WordPress archive, convert HTML to MDX with metadata, repair conversion errors and run build checks. Expected output: A converted local-content archive and build results for completeness, link and preview review before deployment.
CRM exports often contain inconsistent records that can damage a new database or its downstream synchronization if imported blindly.
Agent job: Inspect and clean the export, propose schema changes, obtain approval, then import mapped records and verify synchronization. Expected output: A reviewed migration mapping, imported records and verification findings within the approved database scope.
Moving a live application risks inconsistent data, duplicate scheduled jobs and downtime if server preparation and cutover are poorly coordinated.
Agent job: Inspect the existing server, prepare matching services, transfer data, test staging and coordinate approved synchronization while the owner controls DNS. Expected output: A staged replacement application, migration checks and cutover support subject to backup, access and rollback approval.
Useful content ideas from client calls and daily work are easily forgotten before someone decides what is appropriate to publish.
Agent job: Review authorized transcripts and work logs, identify reusable ideas and add them to a content review queue. Expected output: A queue of work-derived content ideas awaiting owner selection and any client-attribution approval.
An older Windows application can be difficult to maintain when framework changes require both code migration and native verification.
Agent job: Migrate the specified .NET Framework application toward .NET Core, run Windows builds and tests, and iteratively address failures. Expected output: A candidate modernized application with build and test findings for engineering review.
Large collections of class components make architectural modernization repetitive and difficult to review as a single sweeping change.
Agent job: Divide class-component conversion into batches and prepare functional-component pull requests. Expected output: Batched modernization PRs for maintainers to review before merging and releasing.
Daily fleet reports lose value when they repeat unchanged findings and forget the context of issues already under investigation.
Agent job: Prepare fleet-health briefings using prior-run state and per-job notes to distinguish changes from ongoing findings. Expected output: A change-aware daily fleet briefing that preserves relevant follow-up context.
Keeping up with several agent products requires repeated visits to official sources and makes meaningful announcements easy to miss.
Agent job: Check the five selected official product sources on the user's weekday schedule and report linked changes. Expected output: A scheduled phone notification summarizing official product updates with source links.
Regulatory changes can be difficult to spot promptly when relevant notices are spread across numerous official websites.
Agent job: Use parallel agents to inspect the selected regulatory sources and report detected changes. Expected output: A regulatory-change report for qualified human interpretation and applicability review.
Hardcoded thresholds make business rules difficult to inspect and change without risking unnoticed differences in application behavior.
Agent job: Scope a shared configuration system, create per-rule tasks and implement tested replacements for hardcoded thresholds. Expected output: Reviewable configuration changes, rule tasks and test evidence for maintainer and domain-owner approval.
Households may keep paying an existing internet rate because comparing alternatives and waiting through provider chat takes effort.
Agent job: Handle provider support chat after the user logs in and negotiate the internet rate. Expected output: A provider rate offer or negotiation result for the user to inspect.
Vehicle buyers need comparable dealer offers but collecting inventory matches and negotiating by email can become an extended coordination burden.
Agent job: Find matching inventory, request authorized quotes and exchange approved competing offers within the buyer's negotiation limits. Expected output: A set of dealer quotes and negotiation correspondence for the owner to use when completing the purchase.
Simple photo reuse can be slowed by orientation problems or mismatched dimensions that require several manual editing steps.
Agent job: Rotate the selected photo or compose the requested pair side by side with matching heights. Expected output: Corrected or composed images for visual review before overwriting originals or sharing.
A home media collection becomes difficult to browse when file names, artwork, descriptions and categories are inconsistent.
Agent job: Standardize library folders and filenames, find or create posters, add descriptions, and organize categories and collections. Expected output: An organized media catalogue with revised naming, artwork and metadata.
Build delays across connected repositories are difficult to diagnose when dependencies and logs must be understood together.
Agent job: Analyze build logs and cross-repository dependency graphs to identify sources of build inefficiency. Expected output: A set of evidence-backed build-optimization findings for the engineering team's existing review process.
Audit preparation becomes uncertain when evidence is poorly organized and teams cannot tell which controls lack supporting material.
Agent job: Map authorized documents to audit controls, identify coverage gaps and propose file organization before any approved reorganization. Expected output: A control-evidence matrix, gap list and proposed naming or organization changes.
A crowded desktop is hard to navigate when filenames do not reliably describe the contents or suggest sensible folders.
Agent job: Read files in the authorized scope, categorize them by content, move them into appropriate folders and summarize changes. Expected output: An organized folder structure and a change summary, with consequential deletions kept for review.
Engineers checking implementation against specifications need precise evidence from both sources rather than an unsupported assertion that they match.
Agent job: Index specification documents and source symbols, retrieve bounded passages and return separately cited specification-to-code evidence pairs. Expected output: A reviewable collection of specification passages matched with code evidence.
Paying a bill from a photograph still requires checking the biller and amount and keeping the receipt where it can be found.
Agent job: Read the supplied bill, request transaction approval through the wallet and file the resulting receipt if payment completes. Expected output: An approval-gated bill payment and filed receipt, contingent on confirmed biller, amount, currency and payment method.
A voice ordering request needs a clear check of the chosen drink, total and pickup details before it becomes a purchase.
Agent job: Prepare the voice-requested coffee order through the merchant account and place it after confirmation of the drink and total. Expected output: An order confirmation for the confirmed drink and total, if the purchase completes.
Approved contracts can stall when staff cannot identify the current status or the correct destination in the contract system.
Agent job: Locate the contract and its status, preview the workflow move and perform the exact write only after confirmation. Expected output: A reviewed workflow-move preview and, if approved, the contract in the selected process.
An errand can disrupt the day when travel time, pickup reminders and the other person's arrival expectations are handled separately.
Agent job: Coordinate errand travel time, reminder and arrival message. Expected output: Calendar entries, reminder and ETA message.
An electric-vehicle camping trip needs route, charging and overnight availability checks that ordinary point-to-point directions do not resolve.
Agent job: Combine route planning with charger availability and repair warnings, then identify an available campsite without booking. Expected output: A proposed EV itinerary with charging considerations and a campsite option ready for user review.
Progressing through a game can require repeated short actions and careful checkpoints so a risky encounter does not erase useful progress.
Agent job: Observe a user-supplied game's emulator state, take bounded gameplay actions, verify progress and preserve save checkpoints. Expected output: Gameplay progress notes and checked saves toward the requested objective.
Designers need repeated content cards to stay visually consistent without losing the exact properties of the supplied design-system component.
Agent job: Inspect the provided Figma component and populate cards and asset placements from the structured content list. Expected output: A set of populated design-system instances for designer review.
Tax organizers ask for information scattered across local records, making manual field matching slow and missing data easy to overlook.
Agent job: Extract organizer fields, match them to local supporting records and populate the web form. Expected output: A populated tax-organizer draft with missing information flagged, ready for review.
Agency planners need past performance in the right planning format before they can compare channels or shape next year's media allocation.
Agent job: Pull historical cross-channel results into the agency's existing annual media-planning templates. Expected output: Populated media-plan templates with historical performance data for strategist review.
A fix merged into one SDK can leave a second repository inconsistent until someone translates the change into its local conventions.
Agent job: Read the merged diff, determine the equivalent change in the target SDK and open a linked pull request. Expected output: A target-repository PR linked to the original change for language-specific review and testing.
Scientific computing teams need to know whether a GPU port preserves comparable behavior before committing large cluster resources.
Agent job: Adapt the reference pipeline for GPUs, benchmark comparable stages, tune parameters and run only approved cluster workloads. Expected output: A GPU-adapted pipeline, benchmark measurements and run records for independent scientific and resource-use review.
A disputed collection bill is hard to challenge when the relevant insurance explanation and billing evidence are buried in separate systems.
Agent job: Locate the insurance explanation of benefits and related billing evidence, prepare a dispute and fax it with the supporting document. Expected output: A billing-dispute fax with supporting insurance evidence; the source does not establish final fax approval.
A B2B customer's email order must be translated into store items and quantities before the merchant can review pricing and quote accurately.
Agent job: Read the authorized customer email and create a Shopify draft order for the merchant to check. Expected output: A draft order awaiting the owner's quantity, price, tax and recipient review before a quote is sent.
Pull-request reviewers need an initial pass over security patterns, conventions and missing tests without giving up control over approval decisions.
Agent job: Inspect the diff and description and leave authorized review comments, keeping the default behavior comment-only. Expected output: First-pass PR review comments identifying potential issues and missing evidence for human assessment.
Household shoppers repeat the same product matching each week and still need to catch unfamiliar items before checkout.
Agent job: Map the Alexa shopping list to approved product URLs, ask about unfamiliar products and add matched items to the Walmart cart. Expected output: A reviewable grocery cart and any newly approved product mappings, with checkout completed by the owner.
Finding a nearby food pickup offer requires coordinating the service location, available store and a pickup time that works.
Agent job: Set the requested service location, select a suitable store and time, and prepare the local offer without checking out. Expected output: A selected food-pickup deal ready for the user to review before purchase.
Preparing for a business call is harder when the email discussion and calendar invitation are not obviously connected.
Agent job: Find the relevant email thread, match it to the authorized calendar invitation and prepare questions for the call. Expected output: A meeting-preparation brief with context and questions drawn from the matching sources.
A parcel-mapping project needs consistent boundary and elevation context before a human operator can plan a useful drone survey.
Agent job: Assemble parcel and GIS context, prepare a drone mission, process returned imagery and import it into the CAD base. Expected output: A proposed mapping mission and imagery-based CAD materials requiring operator flight control and geometry verification.
Client onboarding feels repetitive when the same workspace must be copied, personalized and shared correctly for each new engagement.
Agent job: Duplicate the approved onboarding folder, personalize its Docs, Sheets and Slides, and apply specifically authorized sharing. Expected output: A client-specific onboarding workspace with reviewed content and requested permissions.
Systematic reviews require a traceable selection process so missing studies, screening disagreements and biased evidence do not disappear into a final narrative.
Agent job: Follow the agreed protocol through database search, deduplication, screening, disagreement handling, bias review and manuscript preparation. Expected output: An evidence package and manuscript draft with an auditable selection trail for researcher review.
Recruiters need a grounded view of comparable roles, pay and local talent availability before discussing hiring expectations with managers.
Agent job: Use the job description to research comparable roles, compensation and location-specific talent pools. Expected output: A role-specific talent-market briefing for recruiter and hiring-manager discussion.
Vendor renewals are difficult to negotiate when proposed terms, actual usage and alternative quotes have not been compared together.
Agent job: Compare signed and proposed agreements with usage, internal discussion and competing offers, then prepare negotiation materials. Expected output: A material-change table and renewal-negotiation brief for commercial and legal review.
SaaS teams need to notice unusual business movements without manually reconciling subscription, activation and acquisition dashboards every week.
Source action: Combines subscription billing, product activation and acquisition data, computes weekly metrics against a trailing baseline and sends a narrative anomaly digest. Expected output: A weekly narrative metrics digest with anomalies and baseline comparisons for human validation.
People who want a copy of their account data need to find the correct export options without accidentally initiating an unwanted request.
Agent job: Navigate LinkedIn's data-export settings and select the desired archive, stopping before the archive request is submitted. Expected output: A prepared archive-selection screen ready for the user's submission decision.
A rejected airline compensation claim can be difficult to challenge when the stated reason has not been checked against flight evidence.
Agent job: Retrieve flight-status information, draft an evidence-backed challenge and send the final email only after explicit user approval. Expected output: A reviewed compensation-appeal email and supporting evidence, with any airline decision or settlement remaining uncertain.
Turning an object idea into a physical print involves several technical stages where dimensions, materials and machine readiness can be overlooked.
Agent job: Develop the explicit object request, generate a mesh, slice it and progress through upload and printing only at approved stages. Expected output: A reviewable model and sliced print file, with an authorized print and status record only after operator approval.
Newsletter editors need to turn a collection of stories into a coherent issue while preserving a final review before subscribers receive it.
Agent job: Rank collected stories, draft introductions, prepare newsletter HTML and send through the email platform after editor approval. Expected output: A reviewable newsletter issue and, once authorized, subscriber distribution through the approved list and send time.
Onboarding several vendors repeats document extraction, agreement preparation and portal entry while tracker updates can fall behind.
Agent job: Read vendor documents and templates, draft agreements, fill procurement forms and update the authorized onboarding tracker. Expected output: Vendor agreement drafts, prepared portal entries and an updated tracker for review before consequential submissions.
Daily point-of-sale totals need accounting-specific fields and formatting before they can be reviewed for a reliable import.
Agent job: Download the daily POS sales and accounting data, calculate the required fields and format a QuickBooks Online import CSV. Expected output: An accounting-import CSV awaiting mapping and reconciliation checks before any separate import.
Managing configuration across self-hosted machines is repetitive, but applying an unchecked change can disrupt networking or expose services.
Agent job: Edit the shared NixOS configuration repository for review while leaving configuration activation to the operator. Expected output: Proposed fleet-configuration changes that the owner reviews and activates manually.
Businesses need a clear way to revisit inactive customer relationships without composing every re-engagement message from scratch.
Source action: Identify inactive buyers and draft personalized win-back messages. Expected output: Proposed review artifact: Customer-specific draft candidates with inactivity evidence, exclusions and owner-review notes.
Reaching audiences in other languages requires coordinating dubbing jobs, translations and partial outputs without losing control of voice rights or cost.
Agent job: Resolve media files, target languages and workspace, submit dubbing jobs, manage documented pauses and eligible resumptions, and report each part's delivery status. Expected output: Reviewable dubbed outputs, or disclosed server-only results where applicable, with incomplete work and additional approvals identified.
Preparing agreements for many signers invites mistakes in recipient details, role assignments and signature-field placement.
Agent job: Generate agreement instances from the approved template, map DocuSign roles and fields, and queue envelopes for human review before sending. Expected output: Reviewable signature envelopes and, after authorized sending, signature-status tracking without signing for recipients.
Code cleanup and performance maintenance can remain neglected when developers must prioritize visible feature work during the day.
Agent job: Run scoped overnight maintenance to improve code quality, remove dead logic and address loading or bundle issues through new PRs. Expected output: Maintenance pull requests with changes and supporting checks for human review before merge.
Payroll planning becomes stressful when available cash, expected settlements and overdue customer payments are not considered together.
Agent job: Combine cash balances, anticipated settlements and overdue invoices into a payroll forecast and stage collection reminders. Expected output: A payroll-liquidity forecast and prioritized reminder drafts awaiting review before sends, postings or payments.
Payroll preparation can be delayed by inconsistent timesheets or unresolved exceptions that are discovered only at submission time.
Source action: Check timesheets and stage payroll for submission. Expected output: Proposed review artifact: Timesheet exception queue and staged payroll preview with unresolved entries held.
Podcast production requires both informed interview preparation and consistent release materials drawn from what was actually recorded.
Agent job: Research the topic and guest, write an interview plan, then turn the transcript into timestamped notes and promotional drafts. Expected output: Interview-preparation files, episode notes, descriptions and promotional drafts awaiting factual and publication review.
Portfolio owners need to see allocation drift and possible tax exposure before considering any change to their holdings.
Source action: Analyze portfolios for allocation drift and tax exposure, producing rebalancing suggestions for review. Expected output: Proposed review artifact: Allocation-review worksheet with input lineage, data gaps and adviser questions.
A sales proposal can stall when scattered source material must be converted into a clear scope and checked pricing.
Source action: Convert source materials into priced proposals for approval. Expected output: Proposed review artifact: Priced proposal draft with source-to-scope mapping, assumptions and unresolved commercial questions.
Community engagement takes time when useful conversations must be found and replies need to fit the actual context.
Agent job: Find relevant group posts, draft contextual replies and wait for the operator's approval before each post. Expected output: A queue of relevant reply drafts with their target conversations for individual approval.
Customer teams need a considered response to retention signals while keeping proposed outreach and concessions visible to human reviewers.
Agent job: Combine internal risk indicators with CRM context, draft re-engagement sequences and create internal discount-review requests. Expected output: Draft retention sequences and internal deal-desk tickets requesting discount review.
Preparing a CNC job requires coordinating geometry, stock, tooling and workholding while keeping potentially dangerous assumptions visible.
Agent job: Inspect the Fusion part and stock, confirm the machining plan, build setups and toolpaths, and report screenshots and rule-based warnings. Expected output: Reviewable CNC setups and toolpaths with warnings, requiring a machinist's collision simulation before any physical operation.
Rental owners need to understand each property's income and expenses rather than relying on a combined accounting view.
Agent job: Read the authorized property-management accounting records and prepare a profit-and-loss statement for each selected rental property. Expected output: Property-level profit-and-loss reports for accounting verification.
Employees completing travel claims must transfer transportation charges accurately from statements into the organization's required expense forms.
Agent job: Extract transportation charges from approved statements, map them into travel forms and verify the transfer. Expected output: Prepared travel-expense forms with transferred charges ready for employee or finance review.
Weather-model training requires consistent channels and derived variables, making raw forecast files difficult to use directly as learning inputs.
Agent job: Plan model-specific channels, fetch selected GRIB fields, calculate derived variables and export arrays with previews and manifests. Expected output: Training-data arrays, preview images and manifests for researcher validation of fields, transformations and scope.
Teams forget why earlier decisions were made and may miss the point when assumptions should be reconsidered.
Agent job: Create numbered decision records, connect superseding decisions and check for due reviews while keeping unapproved records proposed. Expected output: Searchable decision records and scheduled review notices only when a review is due.
Attachments shared in conversations become difficult to find and trust when originals, context and duplicate copies are scattered.
Agent job: Archive attachments from authorized channels, preserve provenance, calculate checksums and create searchable text companions and an indexed inventory. Expected output: A verified attachment archive with originals, provenance, checksums and searchable companions in the approved destination.
Documentation can appear complete while still failing to give a developer enough information to build a working integration.
Agent job: Generate integration code using only the selected documentation, leaving the team to test the result as evidence of documentation adequacy. Expected output: Candidate integration code whose test outcomes can reveal documentation gaps for human review.
Producing a short explainer requires connecting research, narration and timed visuals into a coherent video rather than separate unfinished assets.
Agent job: Research the chosen topic, prepare script and scenes, build Remotion components, inspect frames and add narration and captions before rendering. Expected output: A rendered vertical explainer for factual, asset-rights, voice and publication review.
Site owners need concrete accessibility findings while understanding which checks automation and available interaction evidence cannot actually establish.
Agent job: Confirm scope and target level, inspect available accessibility evidence, perform authorized checks and mark unsupported criteria as not tested. Expected output: A criterion-level accessibility report with evidence and explicit testing limits, requiring specialist review before any conformance claim.
A long-form fiction project needs continuity across chapters and repeated editorial passes before manuscript and supporting formats can be reviewed together.
Agent job: Build story context, draft and critique chapters, revise and typeset the manuscript, and prepare cover and audiobook materials. Expected output: A novel package containing manuscript and supporting production materials for editorial, rights and publication review.
Metagenomic research needs an auditable route from sequencing data to organism and marker summaries without treating tool outputs as clinical conclusions.
Agent job: Coordinate taxonomy, abundance, resistance-marker and optional pathway analysis on approved sequencing input, then export research results. Expected output: Research tables and figures for expert parameter, interpretation and scientific-publication review.
Designers evaluating desktop interaction ideas need working alternatives they can try, rather than relying only on static sketches.
Agent job: Coordinate design and engineering work to build multiple interactive desktop concepts for comparison. Expected output: Hands-on prototypes for designer evaluation and revision before any production or public release.
Running a modded game server requires compatible software, dependable launch routines and connectivity checks that can overwhelm a casual host.
Agent job: Collect preferences, inspect the approved modpack, prepare runtime and configuration, set up backups and check logs and a player connection. Expected output: A configured modded server with launch, backup and connectivity-check results.
Adding a client document to a bilingual website requires following repository conventions and updating both language listings consistently.
Agent job: Inspect localization conventions, add the supplied document, update bilingual listings, run checks and open a reviewable PR. Expected output: A checked bilingual document-publication PR for owner review and merge.
A customer seeking a refund for a stolen parcel may spend substantial time waiting for the right support representative.
Agent job: Monitor the support queue, shorten polling when an associate appears and request a refund for the stated stolen-package issue. Expected output: A support conversation and recorded response to the refund request, without assurance that a refund will be granted.
Unused cloud services can keep generating charges while receipts and the evidence needed for a billing request remain scattered.
Agent job: Review authorized receipts, identify the selected idle-service charges and prepare a billing request and follow-up for approval before sending. Expected output: A documented refund or credit request and any approved follow-up, with the provider's decision remaining uncertain.
Inbound enquiries can lose momentum when fit assessment, contact details and CRM handoff are handled separately.
Agent job: Assess incoming lead fit, gather relevant contact details, create a HubSpot deal and post a Slack summary. Expected output: A CRM deal and internal lead summary reflecting the qualification findings.
Parents want to preserve meaningful conversation moments while avoiding inaccurate milestones or unsupported interpretations of a child's development.
Agent job: Queue conversation quotes and observed moments for parental review and prepare a parent brief. Expected output: Reviewable memory candidates and a parent brief, with retention dependent on parental approval.
Support managers need to spot overloaded queues and urgent cases before uneven assignments turn into missed response commitments.
Agent job: Monitor queue volume and response deadlines, reprioritize tickets, reassign urgent work and warn about approaching breaches. Expected output: Ticket priority and ownership updates plus alerts about approaching response deadlines.
Creators can lose track of which ideas have already been published when newsletters and local drafts are stored separately.
Agent job: Sync published posts, match newsletters to local copies and archive used ideas with their publication links. Expected output: A reconciled content archive linking published pieces to local materials and used ideas.
Reviewing a past tax return can miss important omissions unless each supporting record is checked against the reported figures.
Agent job: Compare the prior return with authorized supporting records document by document and flag possible discrepancies for an accountant. Expected output: An evidence-backed discrepancy list for professional verification before any amendment, disclosure or submission.
Commission claims can disagree on agent identities and deal splits, leaving payment records difficult to reconcile.
Agent job: Match deal submissions to manager claims, agent identities and splits, then record commission updates and overrides. Expected output: Reconciled commission records with an override log.
Brokerage statements and tax returns make it difficult to see what investments earned after fees and tax.
Agent job: Reconcile authorized historical records under a reviewed plan and calculate annual net-return attribution. Expected output: A reconciled historical return analysis for financial review.
Infrastructure notes become unreliable when hosts and services change without matching documentation updates.
Agent job: Compare existing infrastructure notes with live SSH-accessible state and update the service inventory. Expected output: An updated infrastructure wiki with recorded discrepancies.
Useful facts remain scattered across notes and disconnected links, making relationships difficult to trace.
Agent job: Extract note entities, reconcile links with graph records and enrich them from authorized personal exports. Expected output: A reconciled knowledge graph and reviewed entity records.
Finance cannot explain why the approved headcount plan, employee roster and payroll totals disagree.
Agent job: Compare employee and cost-centre records across the roster, budget and payroll, then categorize differences. Expected output: A repeatable staffing reconciliation with differences for review.
Stale work tickets make it hard to tell which agent tasks still need attention and which can be closed.
Agent job: Check Code Queue records, report work status in Slack and propose stale-ticket cleanup. Expected output: A status report and proposed ticket changes.
Hours spent studying do not clearly show which parts of the curriculum have actually been covered.
Agent job: Compare logged study time with chapter-completion records and report completed and pending topics. Expected output: A subject-level study-effort and coverage summary.
Supplier stock arrives in emailed PDFs while the inventory system needs matched quantities and visible exceptions.
Agent job: Extract supplier quantities, match inventory records, update stock and flag large differences. Expected output: Updated stock records with significant variances flagged.
Transaction records from multiple systems are difficult to match without disturbing the accountant's authoritative ledger.
Agent job: Import records into a separate ledger, propose transaction matches and compare with accountant-managed records. Expected output: A parallel reconciliation ledger for accountant review.
A suspected repository compromise leaves scattered evidence and competing explanations that need a defensible timeline.
Agent job: Collect scoped repository evidence, reconcile a timeline, test hypotheses and redact the resulting forensic report. Expected output: An evidence-linked repository incident report.
Photographed scores cannot be edited directly, and recognition errors can distort notes, rhythm and instrument alignment.
Agent job: Compare optical-music-recognition outputs, check structure, repair suitable errors and export an editable score with previews. Expected output: Reviewable MusicXML and rendered score previews.
Product staff need to understand application journeys without asking engineers to demonstrate every path manually.
Agent job: Exercise requested application paths and record flow videos for product review. Expected output: Videos showing the requested application flows.
A delayed flight can invalidate a connection while alternative routes and airline support options are scattered.
Agent job: Contact airline support, investigate alternative routes and coordinate a replacement connection. Expected output: A proposed or reported replacement itinerary and support outcome.
An old résumé and later work history sit in separate places, making an accurate career update cumbersome.
Agent job: Locate the prior résumé in email and revise it using available subsequent career records. Expected output: An updated résumé draft for review.
Critical recovery safeguards can be buried in legacy programs and missing from the team's current understanding.
Agent job: Use a system map to inspect legacy claims-processing code and explain duplicate-transaction recovery logic. Expected output: An explanation of existing business safeguards in the code.
An invitation can be hard to find, and its parking and arrival details may be scattered across messages.
Agent job: Locate the invitation, extract practical details and map an arrival route from an approved starting area. Expected output: An event-arrival plan with location and parking context.
Repeated agent wakeups and overlapping checks can consume resources without improving when useful information arrives.
Agent job: Audit scheduled jobs, consolidate overlapping runs and batch nonurgent messages while retaining delivery requirements. Expected output: A revised agent schedule with reduced redundant checks.
A large structural refactor spans too many files for one isolated editing pass to coordinate safely.
Agent job: Delegate refactoring across isolated work areas, reconcile changes and gather pull-request test evidence. Expected output: A coordinated refactor pull request with check results.
A new identifier format must fit legacy programs without breaking business rules spread across the codebase.
Agent job: Apply expert-authored constraints to identifier changes across the affected legacy programs. Expected output: Reviewable identifier-refactor changes under the supplied constraints.
A finished manuscript may need substantial structural and reference changes before it fits a different journal.
Agent job: Adapt manuscript structure and references to the selected journal while preserving citation fields. Expected output: A reformatted manuscript and PDF for submission review.
Forecasts lag behind closed actuals, and scenario assumptions become hard to trace as models are rolled forward.
Agent job: Refresh the existing model with actuals, extend formulas and document scenario assumptions. Expected output: An updated forecast, scenarios and explanatory assumptions memo.
Project boards fall behind decisions and commitments spread across messages, meetings and calendars.
Agent job: Gather weekly communications, propose project-table changes and write updates after approval. Expected output: An approved, refreshed project board.
Tool-pilot evidence is spread across notes, meetings and web sources instead of one current evaluation view.
Agent job: Gather relevant pilot information and update the data behind the evaluation dashboard. Expected output: A refreshed tool-evaluation dashboard dataset.
Remembering when a relationship was last discussed is difficult when calls and messages have separate histories.
Agent job: Summarize authorized one-to-one messages and import qualifying call records into personal CRM history. Expected output: Proposed contact-history updates with communication dates and identity uncertainty available for owner review.
Repeated workout music becomes stale, while playlists often do not fit the length or pace of the activity.
Agent job: Select fresh upbeat tracks and assemble playlists suited to different workout durations. Expected output: Activity-length playlist proposals with track durations and substitutions visible before live replacement.
Event signup questions and calendar entry require separate steps that are easy to leave unfinished.
Source action: Completes event registration, asks required questions, and adds the calendar entry. Expected output: Proposed review artifact: Registration checklist with unanswered fields, approval needs, confirmation evidence and draft calendar details.
Generic fitness-activity names make a long history difficult to browse or recognize later.
Source action: Renames past fitness activities using context rules and keeps a change log. Expected output: Proposed review artifact: Before/after title proposal with activity IDs, rule rationale and unchanged ambiguous records.
Spam across email, professional messages and calendar invitations obscures communications that deserve attention.
Agent job: Identify unwanted messages and invitations, then remove or block them within the authorized scope. Expected output: A cleaned message and invitation queue.
A product model needs a polished motion presentation, but view selection, lighting and frame delivery require several coordinated steps.
Agent job: Render candidate views, obtain a selection, configure a camera animation and encode the completed frames. Expected output: A reviewable product-showcase animation.
Smart-home routines stop working when devices or entities change, leaving users to trace the broken references.
Agent job: Inspect current device state, identify unavailable entities and replace the broken automation with a working equivalent. Expected output: A repaired automation and updated dashboard for review.
A production defect needs both a code fix and confirmation that the service still behaves after deployment.
Agent job: Investigate reported application bugs, apply fixes, deploy them and check service health. Expected output: A reported deployed fix with service-health checks.
Some flashcards repeatedly consume review time, but changing them carelessly can damage study history or card relationships.
Agent job: Review collection conventions and history, propose exact card edits and reread approved changes. Expected output: An approved batch of repaired flashcards.
Changes to verified software can break proofs even when the intended specification should remain unchanged.
Agent job: Inspect proof states, edit proof code, run builds and audit assumptions after repair. Expected output: A reviewable proof repair with checker results and premise review.
Missing product descriptions, accessible image text or prices leave a store catalogue incomplete and harder to use.
Agent job: Inspect product records for missing information and correct the selected fields. Expected output: More complete product records for review.
Replacing a frontend dependency creates repetitive edits across components and risks inconsistent migration.
Agent job: Partition the replacement work, coordinate parallel changes and run the relevant checks. Expected output: A dependency-migration change set with test results.
A newsletter contains useful ideas that never reach audiences on the creator's other channels.
Agent job: Adapt newsletter material into channel-specific social posts and schedule the recurring work. Expected output: Social-post drafts or scheduled posts based on the newsletter.
A developer away from the workstation needs repository changes without losing branch, test and review controls.
Agent job: Receive a scoped request through Telegram, edit a branch, run checks and return a merge request for review. Expected output: A reviewable merge request and check summary.
Content ideas are easy to capture but difficult to turn into a researched backlog that can guide production.
Agent job: Investigate a proposed topic across relevant sources and save the findings to a backlog card. Expected output: A researched content card ready for editorial review.
Legal research is spread across authorities and prior work, making it difficult to assemble a traceable answer to a defined question.
Agent job: Research the scoped question, consult prior matter material and draft a cited memo with unresolved issues. Expected output: A cited legal memorandum for qualified review.
Salespeople spend substantial effort separating genuine buying signals from loosely relevant leads before writing outreach.
Agent job: Research intent signals, check opportunities and rank relevant contacts for a human outreach list. Expected output: A prioritized prospect shortlist with supporting context.
A journalist search can miss relevant people or include outdated employers without follow-up checking.
Agent job: Research journalists on a defined beat and revise the list in response to coverage and accuracy feedback. Expected output: A reviewable journalist contact-research list.
Award research can overrepresent visible regions or profiles while leaving important candidates undocumented.
Agent job: Investigate candidate profiles and review the list for missing information and geographic coverage. Expected output: A candidate-research list with completeness limitations visible.
Insurance alternatives are hard to compare when the current coverage details live in a declarations document.
Agent job: Read the supplied declarations page, research replacement options and prepare comparison or application fields. Expected output: A replacement-coverage comparison and prepared forms for review.
Conflicting contact records and sequence assignments can lead to duplicate or inappropriate sales outreach.
Agent job: Compare contact and sequence records across systems, hold conflicts and stage permitted corrections unsent. Expected output: Reconciled sales records with unresolved conflicts held for review.
Support staff repeatedly need to apply the same refund rules while still respecting the customer's specific request.
Agent job: Apply supplied refund instructions to incoming requests and resolve cases within the permitted policy. Expected output: A policy-based refund decision and resolution record.
A self-hosted flashcard library is of limited use when collections cannot synchronize reliably to a phone.
Agent job: Configure the sync service, test collection transfer and repair the reported database-schema failure. Expected output: A restored flashcard synchronization path.
Always-on feature flags leave obsolete branches and test dependencies that complicate maintenance.
Agent job: Remove retired flags in parallel, update affected tests and consolidate the changes for review. Expected output: A feature-flag cleanup pull request with validation findings.
Loan files need an official property-tax bill, but finding the correct local portal and document takes manual searching.
Agent job: Use the property address to find the county portal and retrieve the tax document. Expected output: An official tax-bill document for the loan file.
A company-specific question requires the right official filing rather than an unsourced summary or similarly named entity.
Agent job: Resolve the company, locate its filing and return evidence linked to the relevant document. Expected output: A filing-grounded answer with document references.
Technical migration answers may depend on documentation available only through an authorized account.
Agent job: Retrieve authorized technical documentation and incorporate source citations into the migration answer. Expected output: A migration-research answer grounded in the retrieved documentation.
Treaty research requires the actual official documents and status, not just a list of search results.
Agent job: Filter the official registry, retrieve matching treaty documents and report their status. Expected output: A filtered set of official treaty files and status details.
During an incident, responders can lose time finding which existing playbook fits the current symptoms.
Agent job: Gather incident context and identify the relevant response playbook. Expected output: A context-matched incident playbook for the responder.
Recurring emailed forms require repeated completion and return, while signatures create an important authorization boundary.
Agent job: Detect recurring emailed PDFs, place the supplied signature image and return the completed forms. Expected output: Completed recurring forms; signature authority requires separate review.
Chess-engine numbers alone do not explain which moments deserve review or how mistakes connect across a game.
Agent job: Analyze supplied PGN evaluations and clocks, select critical moments and build a player-focused review with interactive boards. Expected output: A game-review report and linked coaching boards.
Proposals can contain weak statistics or risky terms that are difficult to notice while reviewing layout and wording.
Agent job: Check proposal claims, flag problematic terms and mark up the document. Expected output: A proposal with factual and contractual review annotations.
Old repositories accumulate without a clear decision about which projects should stay public, become private or be retired.
Agent job: Inventory owned repositories, group them by activity and purpose, and apply explicitly approved cleanup decisions. Expected output: A reviewed repository-retirement plan and authorized changes.
Sales managers cannot review every booked call individually while keeping track of the team's common issues.
Agent job: Find booked appointments, review their recordings and summarize sales-call findings in Slack. Expected output: A manager-facing sales-call digest.
Card benefits and posted spending are difficult to compare, so unused credits or unsuitable card choices may go unnoticed.
Agent job: Compare authorized posted transactions with researched current benefits and flag potential mismatches or unused credits. Expected output: A card-benefit and spending review for the owner.
Inventory bundles need price checks, but reviewing each combination and proposed correction creates manual work.
Agent job: Inspect inventory and bundle context, compare prices and prepare adjustments for approval. Expected output: A queue of proposed bundle-price corrections.
Support leaders need concrete coaching examples from many human and bot conversations rather than isolated anecdotes.
Agent job: Review support interactions and summarize recurring response-quality issues for leadership. Expected output: A support-quality coaching summary.
Creators can keep publishing on habit even when recent channel evidence suggests a different focus.
Agent job: Compare recent channel performance, update the publishing strategy and recommend where to focus. Expected output: A channel-performance-based strategy revision.
Traffic and conversion data may show an offer is not working without revealing a practical next experiment.
Agent job: Review early funnel evidence, revise the offer and build supporting acquisition material. Expected output: A revised offer and funnel experiment for review.
Starting a new reporting period requires repetitive workbook cleanup without damaging formulas or structure.
Agent job: Copy the period sheet, update its name and clear prior actuals and notes while retaining the template. Expected output: A clean next-period tracking sheet.
Parents need to notify the correct teacher about an absence without searching through school contacts at the last minute.
Agent job: Prepare an absence notice for the responsible teacher and verify recipient identity with the guardian. Expected output: A teacher-notice draft and contact reference, awaiting guardian authorization before sharing child information.
A household routine can require several device controls whose names and current states are not obvious.
Agent job: Resolve requested devices, check their states and perform the approved lighting, climate or media routine. Expected output: The requested device-state changes with a task result.
Hardware validation requires coordinated bench access, flashing and result triage that can delay an engineer's next decision.
Agent job: Trigger the scheduled bench program after code completion, collect test results and prepare engineering triage. Expected output: A hardware-test result and triage report.
Model-evaluation runs are hard to reproduce when commands, failures and decisions are scattered across a session.
Agent job: Review the execution plan, run the evaluation and record commands, outputs and dead ends in a notebook. Expected output: A reusable evaluation notebook with an execution record.
Second-hand equipment listings change quickly, and worthwhile offers may require more judgment than a simple keyword alert.
Agent job: Monitor selected marketplaces and consult a specialist when a listing needs further assessment. Expected output: A shortlist or alert for potentially relevant equipment listings.
A sector screen requires comparable company identifiers and market sizes instead of an unstructured list of names.
Agent job: Filter listed companies by sector and capitalization range and collect their basic market details. Expected output: A sector-screening table with tickers, countries and capitalization.
Site selection involves overlapping spatial requirements and exclusions that are difficult to evaluate consistently by eye.
Agent job: Check and align spatial layers, apply the supplied proximity and exclusion rules and export candidate areas. Expected output: Reviewable site-screening geometry based on the stated criteria.
Investment-review documents use inconsistent formats, making opportunities and downside scenarios difficult to compare.
Source action: Extract standardized financials from document sets, model scenarios, and assess opportunities against supplied investment criteria. Expected output: Proposed review artifact: Financial-extraction and scenario-plan candidate with source links, assumption register and reviewer exceptions.
Potential projects are scattered across permit and government records and need investigation before an investment discussion.
Agent job: Search matching public records and prepare an opportunity memo with financial-model drafts. Expected output: A public-record-grounded investment-review draft.
Microscopy images require consistent segmentation and measurements before a researcher can assess cell patterns.
Agent job: Run local cell segmentation, measure morphology and export masks, overlays and measurement reports. Expected output: Reviewable cell masks and morphology measurements.
Finding suitable library books and placing holds can become a separate recurring household admin task.
Agent job: Select books that match the supplied needs and request them through the authorized library account. Expected output: A selected reading list and requested library holds.
A traveller wants to know when a specific fare changes without repeating the same search every day.
Agent job: Retrieve the selected flight fare and configure a daily change-notification task. Expected output: A fare-watch task for the chosen itinerary.
A new developer must coordinate several repositories and local setup steps before contributing productively.
Agent job: Guide interactive setup of the project's frontend, backend and end-to-end repositories. Expected output: A configured development workspace with the setup steps completed.
Brands need creator partners whose audience evidence fits the campaign rather than relying on superficial follower counts.
Agent job: Compare opted-in creator profiles and authenticated performance data, then prepare partnership materials. Expected output: A creator-partner shortlist and reviewable outreach materials.
Jobseekers waste time sorting old or unsuitable listings when they need recent remote openings.
Agent job: Apply recency and remote-work filters to a job board and collect matching roles. Expected output: A shortlist of current matching job openings.
Relevant industry events are spread across sources, and isolated dates do not reveal a practical trip plan.
Agent job: Filter events by industry, location and budget and compare combinations suitable for one trip. Expected output: A shortlist of event-based travel options.
Slow tests delay development, but performance changes can silently reduce what the suite checks.
Agent job: Refactor slow tests and compare test counts and covered lines before proposing the change. Expected output: A test-performance pull request with coverage-preservation checks.
A finished social draft still needs to be entered into the right live composer without accidentally publishing it.
Agent job: Place the drafted post into the selected social composer and stop before publication. Expected output: A staged social post awaiting a publishing decision.
Campaign edits span several advertising interfaces, making it difficult to review changes before money is committed.
Agent job: Read normalized campaign data and stage proposed changes while keeping new campaigns paused. Expected output: Paused campaign drafts or staged edits for approval.
Support staff need the customer's context and an appropriate escalation path before they can work a complex issue.
Agent job: Gather customer context, start a scoped support session and escalate unresolved issues to the team. Expected output: A contextual support session with escalation when needed.
A veterinary bill must be transferred into an insurer's claim process and followed through subsequent updates.
Agent job: Enter the authorized bill into the insurer's claim flow and summarize later insurer messages. Expected output: A submitted claim and update summaries, without a payout guarantee.
Creators struggle to track relevant directories, submission requirements and whether an approved resource has been accepted.
Agent job: Research suitable publishers, submit an approved resource and track receipts and review status. Expected output: A directory-submission status register.
A long cooking process requires repeated temperature checks while the cook still needs safe control of equipment and handoff.
Agent job: Read cooking telemetry, estimate progress, make bounded adjustments and stop scheduled checks at handoff. Expected output: A monitored cooking session with bounded adjustments and handoff.
Important deadlines become buried in archived bills and correspondence instead of appearing in an actionable queue.
Agent job: Review archived correspondence signals, record deadlines in an attention ledger and notify the owner. Expected output: A deadline-attention queue and relevant alerts.
An errand is easy to forget when its relevance only becomes obvious after a destination is chosen.
Agent job: Resolve the selected destination and filter stored errands for relevant reminders. Expected output: Contextual errand reminders for the destination.
Small administrative commitments remain unfinished because they are scattered across conversations and lack timely follow-up.
Agent job: Track outstanding administrative threads and surface reminders for unfinished commitments. Expected output: A reminder queue for unresolved administrative work.
Repeated support complaints are difficult for engineers to recognize when each arrives as a separate ticket.
Agent job: Group ticket feedback into recurring engineering themes and report the findings in Slack. Expected output: A support-feedback theme report for engineering review.
Survey owners need to understand how agent-generated submissions could affect the data their own form collects.
Agent job: Run the owned survey through the agent and inspect completion timing and submitted responses. Expected output: A controlled survey-agent test record for researcher review.
Cancellation signals need to become a measurable product experiment rather than remaining in a passive dashboard.
Agent job: Connect cancellation and usage evidence, deploy a reviewed retention nudge and report observed outcomes. Expected output: A retention experiment and its reported outcome data.
A technical idea can absorb implementation effort before anyone establishes whether its risky assumptions hold.
Agent job: Agree observable questions, build disposable prototypes, test edge cases and compare the evidence. Expected output: A feasibility verdict with prototype findings and trade-offs.
A cloud bill can rise without a clear connection to the workload or recent change responsible.
Agent job: Compare spending with history and correlate spikes with monitoring and code changes. Expected output: Ranked cost-anomaly causes, owners and recommended actions.
Sales teams cannot explain lost opportunities when lead activity is fragmented across CRM, calls and email.
Agent job: Trace lead touchpoints across systems, identify drop-offs and summarize missed handoffs. Expected output: A lead-journey analysis and executive dashboard.
Published guidance becomes outdated when policies change, but affected pages are hard to identify and assign.
Source action: An analyst’s agent checks earlier domain content when policies change and opens assigned remediation tickets on a recurring basis. Expected output: Proposed review artifact: Candidate correction tickets with exact passage, policy reference, owner question and proposed review priority.
People lose track of chosen routines when completion is remembered informally rather than recorded consistently.
Agent job: Ask the selected habit questions, log confirmed completion and summarize streaks and weekly activity. Expected output: A habit log and periodic summary.
Rental-platform income and local remittance forms require repeated transfer and checking each reporting period.
Agent job: Retrieve authorized rental earnings and prepare the corresponding portal entries and remittance steps. Expected output: A prepared or reported monthly rental-tax submission.
Animation from one skeleton cannot be reused cleanly on another without inspecting and correcting bone mappings.
Agent job: Import the source animation, propose target-rig mappings, obtain corrections and bake the approved transfer. Expected output: A retargeted animation track on the chosen rig.
Moving agent configuration between platforms risks silently losing meanings that do not translate directly.
Agent job: Inspect the selected configuration, preview target files, repair incompatibilities and validate the translation. Expected output: Target-platform configuration and a report of unresolved differences.
Subtitle localization must preserve readable dialogue and timing relationships while handling structural and translation errors.
Agent job: Translate cue text in context, check subtitle structure and readability, then review the separate retiming pass. Expected output: Localized subtitle tracks with timing changes available for review.
Experimental guide-count results need transparent aggregation and annotation before a researcher can prioritize follow-up work.
Agent job: Validate the supplied count table, aggregate gene depletion and combine annotations into a ranked report. Expected output: A traceable experimental-hit triage report.
Repository issues arrive with inconsistent detail, duplicate reports and unclear priority, slowing meaningful investigation.
Agent job: Read issues, apply existing labels, check duplicates and request missing reproduction information. Expected output: A triaged issue queue with reproduction requests and duplicate decisions.
Researchers need to decide which parts of a paper deserve close reading before committing to the full text.
Agent job: Fetch abstracts and section structure, retrieve versioned source material and perform selective analysis. Expected output: Section-grounded paper notes with cached source context.
Large media files can consume disproportionate storage while the path to safe re-encoding is unclear.
Agent job: Audit collection sizes and configure a GPU-assisted transcoding service for selected media. Expected output: A media-size assessment and configured transcoding setup.
An unread service request can leave an appointment unarranged even when the provider has already contacted the owner.
Agent job: Find the service email, ask the owner to choose a date and draft the appointment reply. Expected output: An appointment-response draft awaiting approval.
Exploratory modelling notebooks are difficult to rerun reliably as shared production work.
Agent job: Translate notebook data retrieval, training and evaluation steps into a reusable pipeline. Expected output: A structured ML pipeline derived from the notebook.
A reference image needs several generation, import and material steps before it becomes a usable scene asset.
Agent job: Generate a model from the authorized image, track completion and import the result into Blender for inspection. Expected output: A textured 3D asset prepared in a reviewable scene.
A controls dashboard can reveal a code vulnerability without giving its owner a reviewable remediation change.
Agent job: Investigate the reported dashboard finding and prepare a proposed security-remediation change for code review. Expected output: A reviewable fix pull request or change proposal; deployment and verified remediation remain separate.
A positive sales reply can stall between qualification, scheduling and CRM updates if each step needs separate coordination.
Agent job: Research the reply context, coordinate available slots and record a confirmed meeting with exception handling. Expected output: A qualified meeting booking and updated sales records.
Infrastructure know-how remains in configuration files and people's heads instead of a repeatable operating guide.
Agent job: Inspect approved files and configuration, then document prerequisites, commands, checks and failure handling. Expected output: A draft Markdown runbook with prerequisites, commands, checks and failure-handling guidance for maintainer review.
Captured ideas and clips accumulate as unprocessed notes instead of a clear list of next actions.
Agent job: Review captured notes, update context and organize actionable items in the inbox index. Expected output: A reviewable task list linked to captured material.
Business ideas often rely on isolated anecdotes rather than recurring customer pain and evidence that could disprove them.
Agent job: Verify relevant communities, group pain evidence and rank commercial hypotheses with contrary evidence and tests. Expected output: A research-grounded set of testable business hypotheses.
Useful advice remains trapped in private email, while publication risks exposing personal details or unreviewed claims.
Agent job: Anonymize candidate Q&A, stage entries privately and publish only owner-reviewed selections. Expected output: An approved public knowledge base derived from advice correspondence.
Handwritten lettering is difficult to reuse consistently across designs without a working font and visual checks.
Agent job: Label photographed glyphs, build a font, inspect sample text and revise visible defects. Expected output: A checked font candidate and character previews.
Language-lesson notes pile up without turning into short listening practice that fits a daily routine.
Agent job: Review earlier lesson notes and generate scheduled short audio lessons. Expected output: Daily listening-practice audio based on prior lessons.
Developers can produce working code faster than they learn to understand the patterns inside their own project.
Agent job: Read project code and learning state, then create annotated lessons, questions and prerequisite-gap prompts. Expected output: Project-specific learning lessons and progress records.
Generic product demos fail to reflect the prospective customer's actual data and operational questions.
Agent job: Use authorized prospect recordings, messages and spreadsheets to tailor an interactive HTML demonstration. Expected output: A customer-specific interactive demo.
Expense receipts are scattered across images and PDFs, making weekly spending totals difficult to maintain.
Agent job: Extract new receipt fields into a sheet and refresh the expense dashboard. Expected output: An updated expense ledger and weekly spending view.
Expense claims need readable, consistently organized receipts, but phone photos arrive tilted, uneven and hard to check.
Source action: Straightens photographed receipts, performs OCR, presents extracted fields, binds approved fields to a digest and packs selected receipts into an A4 PDF. Expected output: A reviewed receipt packet with approved extracted fields.
Recurring customer requests are difficult to turn into a scoped engineering proposal without losing the underlying evidence.
Agent job: Group repeated feedback and, at the configured threshold, draft a specification, issue and pull request. Expected output: A feedback-grounded feature proposal and draft PR.
Saved technology ideas rarely become reviewable experiments because research, scoping and implementation happen in separate places.
Agent job: Review bookmarks, draft a prototype brief, obtain approval and delegate implementation to a coding agent. Expected output: A prototype pull request with review visuals.
Spoken product ideas are easy to lose before anyone examines competing solutions and a realistic initial scope.
Agent job: Turn voice notes into tagged concept records with competitive research and a proposed MVP scope. Expected output: A researched product-concept note.
Working technical notes are useful to their author but remain inaccessible or unstructured for the wider team.
Agent job: Combine authorized notes and rough documentation and publish organized team material through the documentation service. Expected output: Published team documentation derived from working notes.
Repeatedly unopened mailing lists fill the inbox, but choosing which subscriptions to remove still needs owner control.
Agent job: Identify likely unwanted lists, present a selection and execute approved unsubscribe requests. Expected output: A reviewed set of unsubscribe actions.
A customer deck can lag behind what the buyer is saying in the current discovery conversation.
Agent job: Use a branded slide master and recent call context to tailor the presentation. Expected output: A customer-specific presentation draft.
Earnings updates must be reflected consistently in both a financial model and the presentation used to explain it.
Source action: Retrieve earnings information, revise the financial model, and carry its context into a summary presentation. Expected output: Proposed review artifact: Update-plan candidate with source-to-cell mapping, assumption changes and corresponding slide revisions.
Household food lists become unreliable when shelf contents, recipes and existing shopping entries are checked separately.
Agent job: Review stock photos and ingredients, ask about ambiguous items and reconcile inventory with the grocery list. Expected output: An updated household inventory and consolidated grocery list.
Ground transport remains to be arranged after a flight is booked, especially when the traveller is busy or away.
Agent job: Retrieve the flight details and coordinate an authorized ground-transport reservation. Expected output: A ground-transport booking associated with the flight.
Reserve spreadsheets require careful review of formulas and factor changes against prior filings and guidance.
Agent job: Inspect reserve triangles, retrieve authorized historical material and prepare cell-linked review findings. Expected output: An actuarial-review brief linked to workbook cells.
A coupon's advertised value may not match the amount actually shown at checkout.
Agent job: Stage the product and coupon in checkout and inspect the resulting price without purchasing. Expected output: A checkout-price verification for the selected discount.
A claim may appear well supported while its citations repeat the same origin or ignore later corrections.
Agent job: Search supporting and contrary evidence, examine source independence and revisions and draft a cited assessment. Expected output: A source-aware factual-claim review.
Rental listings can describe a pool without making it clear whether guests have exclusive access.
Agent job: Inspect selected listings and confirm the evidence for private versus shared pool access. Expected output: A rental shortlist with the pool claim checked.
A sale label does not show whether a digital title is genuinely discounted relative to its price history.
Agent job: Compare current offers with title history and give an evidence-limited buy, wait or skip assessment. Expected output: A historical-price comparison for the selected media.
Order paperwork can disagree on dates, deposits or commissions, creating problems for fulfilment and customer communication.
Agent job: Cross-check order documents, resolve the identified mismatch and prepare the register, packing list and email draft. Expected output: A checked fulfilment package and customer-message draft.
Prospecting lists can incorrectly identify a store's commerce platform, making technology-specific targeting unreliable.
Agent job: Inspect store-level evidence and verify whether the candidate uses the required platform. Expected output: A verified platform-attribution list.
A parcel may stop progressing between routine status checks without the recipient noticing promptly.
Agent job: Check the supplied tracking identifier over time and flag apparent transit delays. Expected output: Parcel-progress summaries and delay alerts.
An important sent message can be forgotten while its author repeatedly checks for a response.
Agent job: Watch the specified conversation and notify the owner when a reply appears. Expected output: A reply notification for the watched message.
A preferred seat may become available after booking, but repeated manual checks are inconvenient.
Source action: Checks seat availability repeatedly and optionally changes an existing reservation’s seat. Expected output: Proposed review artifact: Seat-change review checklist covering flight, current and preferred seat, possible cost and required reservation-confirmation evidence.
Court availability is useful only when the time and outdoor conditions make the booking practical.
Agent job: Check the requested court times and condition outdoor notifications on the forecast. Expected output: Court-availability alerts matched to the requested conditions.
Restaurant openings are difficult to catch, while frequent automated checking can put the account at risk.
Agent job: Check selected reservation openings over time; the reported attempt resulted in account suspension. Expected output: Reservation-availability findings if checks succeed; access may be blocked.
Ticket buyers need to follow several markets for suitable seats without losing track of price changes.
Agent job: Check selected ticket sites, track the preferred seats and notify the owner before purchase. Expected output: A ticket-availability and price-change watch.
Revised attachments can disagree on quantities, discounts or delivery promises, leaving sales unsure which terms are actually approved.
Agent job: Compare the supplied quote chain and approval messages; isolate changed commercial fields and unresolved authority. Expected output: A version timeline, field-by-field conflict workbook and approval questions, with page and message references.
A buyer changes specifications or submission instructions after the response is nearly ready; affected sections are easy to miss.
Agent job: Trace each supplied amendment into the response sections, forms and acknowledgement evidence that need review. Expected output: An amendment impact matrix and missing-acknowledgement checklist; no submission or assurance of compliance.
Teams can expose the same audience to overlapping tests without documenting possible interactions.
Agent job: Compare experiment plans and allocation exports to find shared audiences, time windows and changed surfaces requiring analyst review. Expected output: A collision map, overlap counts where identifiers permit, and analyst questions distinguishing potential interference from established bias.
Capitalization, spelling and missing UTM fields make campaign attribution hard to reconcile.
Agent job: Audit supplied outbound links against the approved campaign taxonomy and propose a reversible mapping. Expected output: A link-level defect list, canonical naming map and corrected candidate URLs with unchanged destinations.
When an item is unavailable, procurement receives alternatives described with different terminology and incomplete technical details.
Agent job: Compare supplied datasheets with an engineer-approved requirement list and separate matches, contradictions and missing evidence. Expected output: A requirement-to-datasheet evidence matrix and clarification questions for a qualified reviewer.
Recurring charges and vendor reference variations create false alarms while real duplicates may have small formatting differences.
Agent job: Group candidate duplicates and assemble evidence for an accountant, preserving every source record. Expected output: A candidate-pair register showing matching and conflicting fields, linked documents and unresolved payment status.
Prepaid balances become hard to explain when service dates, credits and recognition conventions are scattered.
Agent job: Build a review-only prepaid roll-forward using accountant-supplied policies and service periods. Expected output: A formula-driven schedule of opening balances, additions, credits, period expense and closing balances, plus missing-policy questions.
Order forms, CRM notes and provisioning records may disagree on an account’s current service entitlement.
Agent job: Compare approved commercial records with provisioned features and flag contradictions for account-owner resolution. Expected output: An entitlement evidence table by product and effective date, plus a discrepancy packet; no access changes.
Completed tasks do not necessarily show that the customer’s agreed acceptance criteria were met.
Agent job: Match approved acceptance criteria to supplied test results, deliverables and customer sign-off records. Expected output: An acceptance evidence index with supported, missing and contradictory items and draft reviewer questions.
Stale question sheets and rating definitions can undermine a structured interview before any candidate is assessed.
Agent job: Compare blank distributed interview kits against the approved master, checking question order, wording and rubric version only. Expected output: A kit discrepancy report and a clean replacement candidate for human approval.
The careers page, agency brief and posting draft can list conflicting location, schedule, employment type or application deadline.
Agent job: Check supplied posting copies against the approved requisition and report factual mismatches without changing selection criteria. Expected output: A source-linked mismatch checklist and corrected draft passages for recruiter review.
Ambiguous wording or a mismatched answer key frustrates learners and creates avoidable disputes.
Agent job: Check an instructor-owned question bank against supplied teaching material and answer keys, documenting possible ambiguity for educator review. Expected output: An item-level QA report with cited lesson passages, suspected key errors and proposed clarified wording.
Course revisions can leave some promised learning outcomes unassessed while repeatedly testing others.
Agent job: Map approved learning objectives to existing assessment items and identify missing or weak evidence coverage. Expected output: An objective-by-item coverage matrix with confidence, source references and educator questions.
Freelancers can lose track of which changes were included, approved as extras or still undecided.
Agent job: Compare the approved scope with dated requests and approvals to build a factual change ledger. Expected output: A request-by-request scope ledger, evidence links and draft clarification questions; no invented charges or claims.
A finished design can combine licensed images, client-supplied assets and fonts with different permitted uses.
Agent job: Inventory embedded or linked assets and attach the supplied licence or permission evidence to each intended use. Expected output: An asset-permission manifest with missing evidence, explicit restrictions and questions for the owner or qualified adviser.
A product launch stalls when supplier spreadsheets contain invalid headers, duplicate handles or broken variant groupings.
Agent job: Write a schema-aware preflight checker; identify risky overwrites and variant collisions; propose a corrected COPY with a row-by-row change log. Never infer missing identifiers. Expected output: Corrected CSV, exception workbook, reversible import instructions and checker script.
Automatic transcripts can misstate names and specialist terminology, making lessons hard to search or follow.
Agent job: Find inconsistent terminology, speaker labels and timestamps; propose transcript edits with uncertainty markers and an issue list for human listening. Expected output: Reviewed transcript draft, caption correction list and glossary; no certification of accessibility.
Teams repeatedly search old decks and may reuse unsupported case results, expired offers or another client’s confidential details.
Agent job: Index reusable passages by service; attach source date, owner and permission status; quarantine conflicting numbers and confidential examples rather than picking a winner. Expected output: Searchable claim library, reusable evidence-backed snippets and a review queue.
Reporting staff need a clear evidence trail from programme totals to public statements, while activity counts can be mistaken for outcomes.
Agent job: Reconcile denominators and periods; map each draft claim to a source cell; distinguish activity, outcome and unsupported causal language; flag gaps without inventing results. Expected output: Claim-to-evidence workbook, caveat list and editable short impact-report draft.
Return counts alone do not show whether a particular size or product has a disproportionate issue or simply sells more.
Agent job: Join minimized item-level returns to their original sales lines using merchant-generated opaque tokens; form sale-date cohorts, separate missing reasons and compute returned-unit rates against eligible sold units. Surface patterns and competing explanations without causal claims. Expected output: Reconciled analysis workbook, three evidence-linked hypotheses and a data-quality report.
A sponsor deck may promise exposure without an auditable audience baseline, defined placements or a fulfilment checklist.
Agent job: Reconcile audience figures; enumerate each sellable placement; flag double-booked benefits; draft a clear evidence-backed sponsor inventory and post-event proof checklist. Expected output: Sponsor inventory workbook, draft short deck and fulfilment checklist; no sponsor list harvesting or outreach.
A workbook can open successfully while formulas, dates or unsupported features behave differently.
Agent job: Inventory formula and feature differences; build invariant checks for key totals; compare sample outputs; deliver an exception report and proposed repair patch on copies. Expected output: Migration QA workbook, formula-difference inventory and reproducible acceptance checks.
Customers may see conflicting regular or holiday hours across the website and public profiles.
Agent job: Compare each public listing against the approved schedule; distinguish regular hours from special dates; prepare exact correction drafts and an evidence matrix. Expected output: Dated hours-consistency report with screenshots and ready-to-review correction checklist.
A small business may turn an attractive idea into public claims before checking deliverability, differentiation, customer objections and evidence. Multiple independent lenses can expose assumptions before money is spent.
Agent job: Draft one bounded offer, obtain six independent role reviews, reconcile disagreements, and prepare a customer test and human decision packet. Expected output: Versioned offer draft, six structured reviews, evidence audit, qualitative competitor-response analysis, disagreement log and proposed customer test plan awaiting owner approval.
A returned receipt can mix accepted and rejected stock. A draft supplier invoice may omit retained units or bill returned units unless each movement is traced to its original purchase-order line and billing rule.
Agent job: Build a source-linked quantity bridge across receipt, return and invoice copies, keeping physical disposition separate from the owner’s billing-entitlement rule. Expected output: A draft quantity-bridge CSV and exception memo showing retained accepted/rejected units, eligible units under supplied rules, already billed units and unresolved draft-invoice differences; awaiting procurement/AP approval.
A variable-weight delivery can leave a small open quantity even after a supplier considers it final. Closing every paid order risks hiding a real backorder; leaving every residual open obscures genuine purchasing commitments.
Agent job: Compare residual quantities, settlement evidence, owner-defined tolerances and explicit delivery-completion statements to form a closure-review queue without closing orders. Expected output: A draft PO residual queue separating candidate final variances, confirmed pending deliveries and unresolved cases, with tolerance calculations, completion evidence and a purchasing-owner decision column.
A helpdesk view and a messaging provider log can disagree after retries or relationship errors. Retrying solely because a message is absent from the helpdesk risks duplicate contact, while a provider acceptance event does not prove delivery or reading.
Agent job: Join synthetic helpdesk and provider event copies into a state-by-state ledger, separating transport evidence, display presence and uncertain identity mappings. Expected output: A draft reconciliation ledger with provider-accepted, explicitly delivered, missing-view, retry-duplicate and unknown states; any possible retry is an unsent owner-review candidate.
A later thank-you or a different issue can reopen an old conversation. Conversation counts then mix acknowledgement traffic, continuation of the original issue and genuinely new incidents, distorting operational reporting.
Agent job: Apply an owner-defined episode policy to synthetic conversation histories and prepare a reviewable issue-episode table with an explicit ambiguity queue. Expected output: A draft episode table, post-resolution classification queue and count-reconciliation note; conversation IDs remain unchanged and uncertain episodes are not forced into a single metric.
A reported breach can arise from calendar, pause-state or qualifying-response differences rather than slow service. Without an interval-by-interval workpaper, the team cannot distinguish event-data gaps from a disputed SLA interpretation.
Agent job: Reconstruct a first-response clock from synthetic event copies and a supplied policy, exposing included business minutes, exclusions and unresolved contract terms. Expected output: A draft clock ledger with interval formulas, qualifying-response evidence and differences from reported breach flags; contractual interpretation and any service-credit decision await the owner.
A score based only on returned surveys can conceal cases that never received a survey because a template was unapproved, paused or outside an allowed messaging window. Comparing channels without those denominators can misrepresent survey coverage.
Agent job: Reconstruct case-level eligibility and survey transport stages from synthetic status exports and template history, keeping missing ratings missing. Expected output: A draft coverage table of resolved, policy-eligible, sendable, attempted, explicitly delivered and responded cases by channel/time; reason-coded coverage gaps and cautious interpretation notes await the owner.
A monthly retainer can cover a fixed time quota while only excess work is separately billable. Splitting the threshold-crossing entry and reconciling prior billing is difficult when entries span projects or contract rules leave rounding, rollover or eligible work unclear.
Agent job: Prepare a deterministic time-allocation workpaper using owner-supplied retainer rules, preserving threshold crossings and holding ambiguous entries for billing review. Expected output: A draft entry-level included/overage/unresolved allocation, entitlement reconciliation and billing questions; no invoice, charge or contractual entitlement is established.
A copied project can look ready while work items, dependencies or required fields are absent. A later schema change can also make it unclear whether an omission is an intentional exclusion, incomplete export or failed copy.
Agent job: Compare source and destination export copies against an owner-defined copy scope, producing a source-to-copy manifest and evidence-based launch blockers without repairing the project. Expected output: A draft task/relationship/field coverage manifest with missing, intentionally excluded, ambiguous and unobservable categories, plus launch-review questions for the project owner.
Copied teaching material can retain old due dates, lose section-specific exceptions or place unlock, due and close times in an impossible sequence. A uniform date shift cannot infer an educator’s intended holidays, local due times or availability windows.
Agent job: Compare source/destination course-configuration copies against an educator-supplied term mapping and calendar, preparing date exceptions and candidate questions without touching the LMS. Expected output: A draft assignment-date bridge and section-override matrix showing out-of-term dates, unlock/due/close contradictions, holiday conflicts and missing overrides; educator decisions remain pending.
A board move can disagree with a work order’s assignee/date or a calendar’s recipients. Automatically propagating the newest-looking value could send work to the wrong person because export freshness and the authoritative dispatch decision are separate questions.
Agent job: Compare three synthetic representations of the same existing dispatch decision, identify mismatched dates/assignees/recipients, and prepare a dispatcher review sheet without scheduling or sending anything. Expected output: A draft three-way dispatch-consistency sheet with missing assignments, schedule differences, recipient mismatches, stale-snapshot caveats and source-linked correction questions pending dispatcher approval.
Partial receipts, repeated export rows, changed promises and cancellations can produce different supplier delivery percentages. A buyer needs to see which original due quantities and actual receipt events support each numerator and denominator before relying on a score.
Agent job: Build a due-obligation ledger and receipt-event allocation, apply a buyer-supplied quantity-based on-time policy, and explain report differences without changing source records or judging the supplier. Expected output: A draft due-line ledger, receipt-allocation CSV and score-discrepancy memo, including excluded events, unresolved quantities and a buyer sign-off column.
A blanket agreement can have several releases, cancellations and receipts with different link quality. Treating goods received as agreement consumption, or restoring cancelled quantities without a policy, can hide outstanding releases and misstate the quantity still available before expiry.
Agent job: Reconstruct an agreement-line to release-line to receipt-line roll-forward using explicit references and owner-approved commercial rules; isolate orphan releases, invalid dates and terms requiring buyer review. Expected output: A draft agreement roll-forward CSV, release/receipt bridge and expiry-question memo showing gross releases, allowed restorations, net release consumption, open deliveries and unresolved linkage.
A supplier/manufacturer lot identifier can differ from internal receipt batches and finished-goods lots. Missing split, transfer or production links can make a trace appear complete even when some received quantity or downstream destination is unaccounted for.
Agent job: Construct an evidence-linked manufacturer/item/lot crosswalk and directed receipt-to-batch-to-production-to-delivery graph, with unit-aware balance checks and explicit unresolved edges for quality-owner review. Expected output: A draft lot crosswalk, genealogy edge list, reached document/destination-code table and gap register. It describes trace evidence and uncertainty; it does not clear stock, determine safety or initiate a recall.
A mixed-project timesheet can carry a whole-sheet billed label while only some time rows have invoice coverage. A useful review must distinguish genuinely unmatched minutes in a complete supplied ledger from absent links, cancelled documents and known reversals; it cannot infer chargeability from a header status.
Agent job: Build a row-to-invoice coverage ledger from minimized copies, reconcile signed coverage movements under supplied state rules, and explain discrepancies between row evidence and the sheet-level label. Expected output: A draft time_coverage.csv, document_link_exceptions.csv and billing_review_questions.txt showing source-linked coverage, uncovered minutes where provable within the supplied scope, unresolved rows and header conflicts. No invoice or charge recommendation.
A lag adjustment can be hard to review when a project tool requires a relation to be recreated. The owner needs evidence that only the intended lag changes, that dependency identities and directions remain intact in the proposed scenario, and that downstream date constraints are interpreted under the correct calendar.
Agent job: Prepare a hypothetical before/after dependency graph from supplied copies, enforce a lag-only change allowlist, and compute conditional downstream earliest-bound deltas under an explicit owner-approved scheduling model. Expected output: A draft dependency_delta.csv, schedule_implications.csv and schedule_review_questions.txt with unchanged-edge checks, unsupported constraints, downstream lower-bound shifts and deadline conflicts. No live rescheduling or claim that an edit was applied.
After course reuse, duplicate-looking question records can have different identities, versions, scoring settings or quiz dependencies. A cleanup list based on text similarity alone can obscure fixed-version links, latest-version pointers and random category selection, even when no student data is involved.
Agent job: Prepare a version-aware question inventory that separates exact observable content matches from near-match hypotheses, retains every original ID and maps direct and category-based quiz references before presenting educator/admin review questions. Expected output: A draft question_match_review.csv, question_usage_edges.csv, category_growth.csv and keep_separate_questions.txt with preserved identity/version records, evidence for candidate matches and a protected/unobservable register. No merge, deletion recommendation, answer revision or publishable bank file.
Try a broader search or clear one of your filters.
Quotes use each, case or carton units and different minimum orders, hiding the actual cost and surplus quantity.
Agent job: Normalize supplied offers to an approved unit of need and flag assumptions that prevent fair comparison. Expected output: A formula-based comparison of usable quantity, pack rounding, quoted freight and total cost, with unknowns left blank.
A layout or plugin update can leave the enquiry or booking path unusable without an obvious homepage failure.
Agent job: Create and run bounded browser tests for specified paths; capture failing steps and screenshots; produce a reproducible handoff without sending real enquiries. Expected output: Three journey tests, test run evidence, regression report and maintenance notes.
Listeners and producers struggle to locate a useful explanation spread across dozens of episodes.
Agent job: Create a topic taxonomy, index claims to episode/timecode and build a simple searchable archive prototype; return matching excerpts with source links and no unsupported answers. Expected output: Episode topic index and static searchable archive prototype using the creator’s own licensed material.
New staff may miss handover details when training is informal and each location’s expected finish differs.
Agent job: Convert approved observations into location-specific completion checks, clear exceptions and a trainer signoff sheet; flag ambiguous standards for the owner. Expected output: Original editable onboarding checklist pack for three non-hazardous service types, with version and review date.
Vendor opportunities are scattered and differ in eligibility, dates, fees and application deadlines.
Agent job: Review a bounded list of official opportunities; record deadline, eligibility, fee and source timestamp; flag unknowns; rank fit without claiming acceptance or sales potential. Expected output: Original short regional opportunity brief with source links, deadlines and a suitability explanation; no copied listings or photos.