What AI changes in 120 jobs, task by task

How much of a typical week stays with the person?

At a middle pace, 43% of the average week stays yours. The share tools could take depends a lot on how fast employers adopt AI. At a steady pace it's about 10% of the week. At a middle pace it's 31%, and at a fast pace 34%.

The jump from steady to middle is the big one. Once tools are good enough and in use, the tasks that move are mostly the same ones. A faster pace adds a few points more.

Which fields keep the most of the week?

Hands-on and people-heavy work holds up best. Electricians, truck drivers, nurses and store managers keep the largest share of their week. Legal and admin roles keep the least, because so much of their week is reading, drafting and checking documents. Marketing roles have the largest share tools could take.

What makes a task hold up?

We tagged every task with the main thing that keeps it with a person. Tasks with nothing holding them in place are the most exposed. Tasks that need you there in person, or that run on a relationship, are the least exposed. Work someone has to answer for sits in between.

Does the main part of a job change first?

For each role we noted which task is the biggest reason employers hire for it. In 62 of 120 roles, that task lands in "tools could take it." First-pass document review for paralegals, sourcing for recruiters and ad-hoc data pulls for analysts are three examples.

In 56 roles, the task that holds up best is one of the smaller reasons the job exists. For a paralegal, that's exhibit prep and hearing logistics. For a data analyst, it's presenting the number in the room and saying when the data can't support a claim. Those are the tasks worth getting better at.

What matters besides AI?

Every report names the main risk to the job outside AI. For 65 of 120 roles, that risk matters more than AI does. Interest rates drive work for mortgage processors and title examiners. Branch consolidation drives it for bank roles. School enrollment, research grants and federal budgets drive it for teachers, graduate students and contract staff.

For managers it's flattening. Organizations are removing layers, and 30 of our 32 manager and director roles carry that as their main risk.

How do managers compare?

Managers and directors keep 54% of their week at a middle pace, compared with 38% for the other working roles and 41% for students. Their work runs on judgment and relationships. Their risk is that fewer manager seats exist.

Pathcast also scores pressure on entry-level hiring, from 0 to 100. Students average 31 and other working roles 33, against 20 for managers. A higher score means fewer openings for new people.

Which common advice do the reports turn down?

Each report looks at the advice people in that job usually get and says when it fits. Across 120 reports we looked at 394 pieces of common advice. The one turned down most often is to buy a credential or certificate first. It's turned down, for that role and moment, in 77 reports.

Going back to school is turned down in 21 reports. Avoiding AI tools altogether is turned down in 15.

Where can people go next?

Each report links the role to new jobs that AI is creating, matched on skills the person already uses. Our catalog has 47 of these jobs, each with the training to get there. The most common starting point is AI workflow builder, suggested first for 19 roles. Expert AI trainer comes next, for 9. AI deployment strategist, clinic AI agent specialist and AI evaluation specialist follow, at 8 each.

All 120 roles

Shares of the week at a middle pace. A star marks roles where the risk outside AI matters more. Each role links to its full report.

The full data, including the steady and fast paces, is in the CSV download, free to reuse.

Using this research

You're welcome to quote, chart and republish these findings. Please credit "Pathcast by Agori Labs" and link to pathcast.ai. Journalists, researchers and educators can ask for a role's full data at [email protected].

Suggested citation: Agori Labs (2026). What AI changes in 120 jobs, task by task. Pathcast. pathcast.ai/research/ai-and-120-jobs