Start with the week, not the headline
List 5–10 things you actually do each week — using verbs: schedule, reconcile, interview, inspect, explain, draft, decide. The reading below scores each against repeatable-pattern vs. human-context markers so you get a task-by-task view, not a single verdict on your title.
Open the companion tool →Why one number for a whole job misleads
A job title hides variation: a clinic coordinator who schedules and also explains options to anxious patients does two very different things. A tool-by-task view makes the trade visible — which tasks compound when paired with AI and which get replaced if you don't add judgment.
This article does not predict whether you will keep a job. It gives you an inspectable map and a 30-day experiment with evidence, so a manager, mentor or career service can discuss real work, not a headline.
The protocol
- Write 5–10 recurring weekly tasks with concrete verbs. Skip title, write action.
- Run the reading. Each task is checked for human-context signals (decide, coordinate, care, explain, trust, negotiate) and repeatable-production signals (draft, format, transcribe, reconcile, sort).
- Choose one repeatable task to redesign with AI assistance and one context-heavy skill to deepen. Define what good review looks like.
- Re-check in 30 days with one work sample or feedback. Roles and tools change; treat this as a recheck cadence, not a one-time forecast.
What the tool does
list your role's actual weekly tasks, get a task-by-task exposure reading and a ranked list of which skills compound with AI versus get replaced by it. It runs in this browser; nothing is sent or stored. This is not career advice — it is a structured check that leaves you with a next conversation.
Limit first: This is not career advice, a layoff forecast, or a guarantee about employment. It cannot know your employer's plans. Bring the map to a career counselor, manager, union or worker-rights adviser for context.
Run the honest exposure check →One human next step
Bring one task map and one evidence-backed 30-day experiment to a manager, mentor or career service. The prompt is: “Here is the task I would redesign with assistance and the review step I would keep — does this match how you see the work?”
Research log and safety checks
AI-assist path: not used for a verdict. This artifact is intentionally local and inspectable; a remote model would add an opaque prediction where the reader needs to see which task is the issue. §6.10: re-check cadence is the honest frontier-pace answer — not a new prediction.
§0.5 protection result: category is job-skill-disruption; crisis-adjacent: no; fear/urgency/scarcity toolkit used: no. The protective function is free and the paid feature is convenience only. No manufactured urgency on automation risk.