Decide which onboarding steps do not need AI
Text exercise: Provide a process diagram, field rules and fictional data. Draft ownership, duplicate handling, retries and human exception handling.
Recently reviewed cases · 全民AI / AI for Everyone
Draft a project update from two ticket snapshots
Text exercise: Provide current and prior snapshots, a reporting period and status definitions. Draft an evidence-linked progress report.
Scattered voice requests? Draft a ticket for confirmation
Give a text AI a redacted transcript and ticket fields to draft a feature-ticket draft retaining open questions for human review. This is a drafting exercise; the source used Codex.
Moving a browser extension? Check compatibility before coding
Supply extension source and target browser versions to draft API, storage and permission differences with synthetic-data tests and recovery steps.
Small patch contains unrelated edits? Map each change to a requirement
Give a text AI a code diff and flag specification to draft a table of required fixes, extra edits and questions for human review. This is a drafting exercise; the source used Devin.
Navigation prototype keeps failing? Define checks and a stopping point
Supply prototype requirements and limits to design separate search, route-display and error checks using fictional places and explicit stopping rules.
Specify collage layers and delivery requirements
Provide your asset list, canvas size, audience and style. Draft layer, source, crop and label requirements with delivery checks.
Synthesized music clips? Plan note timing and levels first
Give a text AI an original musical idea and output requirements to draft a sound-event timeline and checking plan for human review. This is a drafting exercise; the source used Claude Code.
Clarify robot instructions before anything moves
Supply simulated scenes, API rules and coordinate units to draft action preconditions, boundaries and completion evidence without dispatchable commands.
Make a mathematical idea easier to check
Using Claude where available: Supply a small problem, known baseline and research limits to separate conjectures, candidate solutions and checking methods.
Changed link leads to the wrong page? Define acceptance checks
Give a text AI a test site and approved link destinations to draft a link, keyboard and layout checklist for human review. This is a drafting exercise; the source used GitHub Copilot.
Upgrading document search? Plan quality checks and rollback
Supply a test document set, baseline results and target design to plan a phased retrieval-system migration with citation checks and rollback criteria.
Does an agent save time? Track review and rework as well
Give a text AI low-risk tasks and an existing work baseline to draft a waiting, review, rework and acceptance log for human review. This is a drafting exercise; the source used Devin.
Dashboard totals look wrong? Check rows before charting
Supply synthetic CSV data, field definitions and duplicate rules to draft checks and totals that distinguish rows from unique items.
New vendors trigger false alerts? Test with synthetic invoices
Give a text AI synthetic invoices and fictional history to draft a test table separating new data from anomalies for human review. This is a drafting exercise; the source used Replit Agent.
Reusing a slide template? Catch stale dates and figures
Supply your template and verified current-period inputs to draft a slide-content map highlighting dates, footers and figures that need replacement.
Define dashboard status rules before building
Text exercise: Provide synthetic tickets, status definitions, roles and update frequency. Draft interface and event specifications covering duplicates, stale data and cancellations.
Keep practice teacher comments faithful to the notes
Use entirely fictional observations to draft a practice comment with a note reference for each claim, without inventing traits or family context.
Plan tests around a deletion deadline
Using Claude where available: Supply example rules, synthetic records and a time zone to cover before, at and after the deadline.
Background maintenance conflicts? Plan branches and dependencies
Give a text AI a maintenance backlog and baseline versions to draft a work queue identifying shared-file conflicts for human review. This is a drafting exercise; the source used Cursor.
Draft an order from the actual menu and your budget
Supply menu text, party size and budget to draft an order using listed items, with unclear prices and special requests flagged for staff.
Duplicate smart-home names? Check functions before deleting
Supply a de-identified device list and state snapshot to distinguish duplicates, separate functions and outages before reviewing dependencies.
Define evidence that a booking is complete
Use fictional booking requirements, mock screens and notifications to draft completion checks. Match dates, time zones, party size, location and provider confirmation.
Keep corrections when turning speech into a list
Supply a synthetic speech transcript to draft a read-back list that tracks additions, cancellations, replacements and uncertain quantities.