"Will AI replace this job?" is the wrong question. AI changes tasks. A role is a bundle of tasks; when some of them move to AI, the role changes shape — sometimes a little, sometimes a lot.
Break roles into tasks
For each role, list 5–10 key tasks and the hours per week each takes. Ask the people who do the work; managers often underestimate admin and overestimate the "core" work.
Example: an HR service advisor might spend
- 14 h answering routine policy questions,
- 8 h on data entry and record updates,
- 10 h on complex employee cases,
- 6 h drafting letters and case notes.
Ask three questions per task
- Can AI do this end to end with little human input? (automate)
- Can AI speed it up or improve it while a person stays in charge? (augment)
- Does it need human judgment, care, trust or accountability? (keep human)
What the evidence says
Anthropic's Economic Index, based on millions of anonymized conversations with its Claude AI, found that about 36% of occupations saw AI used for at least a quarter of their associated tasks, but only about 4% saw it used across three-quarters of their tasks. In other words, AI touches parts of many jobs and most of very few.
From tasks to capacity
Add up the hours per task, apply an assumption for how much time AI frees for automated and augmented tasks, and multiply by headcount. The result — hours or FTE freed — is the starting point for workforce planning: what new work, redeployment or reduced hiring it enables.
The workforce impact mapper does this math for your roles.