How it works
How the result is built, and where it stops.
Still In Demand reads two things: the tasks you actually do, and public data about those tasks and the jobs around them. No AI model writes your result. Every figure on it comes from one of the sources below, and the page names which.
Your AI diagnosis
You tick the tasks you do from the job's core task list (O*NET 31.0) and say how much of your week each takes. For each task we read Anthropic's observed AI use: how often people use Claude on that task, and when they do, whether the AI does it end to end (automation) or works alongside a person (augmentation). Weighted by your week, that gives the three shares on your result.
- Your band. Low when under 5% of your ticked week shows AI use: none of your tasks has observed use. Moderate from 5% to 25%: roughly one to three of them. High at 25% and above.
- Confidence. The share of your ticked week that has task-level evidence, and how many tasks you ticked. It is shown next to the band, not hidden under it.
- The federal view. The Bureau of Labor Statistics' relative AI exposure category for your job (Low, Moderate, High, Very high), published in August 2026. BLS also counts what AI could do in theory; your band counts what it is observed doing on your tasks. When the two disagree, the result says so.
- What holds up. The work activities your job rates highest among those that rely on a person: deciding and persuading, caring for people, and hands-on work.
A task with no observed AI use is not a safe task. It means no one has measured it yet. Exposure is not a forecast of job loss: BLS says so about its own categories, and so do we.
Your moves (the Switch Plan)
Every move is a real US occupation that is stable or growing to 2035. Each is scored six ways, and ranked on this weighting: skills that carry over 30%, experience kept 20%, AI touch 20%, pay kept 15%, demand to 2035 10%, effort 5%. It is a decision aid, not a probability.
- Skills that carry over: the similarity of your job's skills, work activities and knowledge to the target's, the method behind the US Department of Labor's own Skills Matcher, adjusted by your answers.
- Experience kept: what you know and how you like to work. A move that keeps too little of it is not offered.
- AI touch: the same task-level reading, on the target job. A move is marked down when it carries more AI use than your job today.
- Pay kept: the target's US median against your pay band, or your job's median if you skipped it.
- Demand: BLS projected growth 2025 to 2035 and openings a year.
- Effort: extra preparation (O*NET job zone), extra entry education (BLS), and whether the job is built on a specialist degree.
Some moves are never offered: a promotion within your own line (chief executive), jobs that need a police or fire academy unless you already work in protective services, and degrees above a bachelor's unless you allow two years to retrain. If fewer than three moves pass every check, the plan shows fewer and says why. It never pads the list.
Sources
- O*NET 31.0 Database, U.S. Department of Labor, Employment and Training Administration: skill and task data for every occupation, with its job titles.
- BLS Employment Projections 2025 to 2035, Tables 1.2 and 1.12, and the AI exposure categories (August 2026).
- BLS Occupational Employment and Wage Statistics, national, May 2025.
- Anthropic Economic Index: observed AI use by task and by occupation, and robot exposure.
- Microsoft Research, Working with AI: AI applicability scores.
This site incorporates information from the O*NET 31.0 Database by the U.S. Department of Labor, Employment and Training Administration (USDOL/ETA). Used under the CC BY 4.0 license. O*NET® is a trademark of USDOL/ETA. Still In Demand has modified all or some of this information. USDOL/ETA has not approved, endorsed, or tested these modifications. Anthropic Economic Index and Microsoft data are used under CC BY 4.0 and have been summarised per occupation by Still In Demand; neither has endorsed this site.