The Remix Career AI Impact Index 2026
How exposed 100 common mid-career roles are to AI, task by task. Version 2, October 2026. Ratings for the original 25 roles were last reviewed in August 2026; the other 75 were rated in October 2026.
Most of the debate about AI and work happens one headline at a time. This index puts 100 common mid-career roles side by side and rates them with the same method, so they can be compared.
For each role, we sorted its core tasks into three groups:
- Automate: tasks AI can increasingly handle from start to finish.
- Augment: tasks AI speeds up while a person still owns the result.
- Stays human: work that depends on accountability, trust, judgment or physical presence.
Each role then gets an exposure band, from Low to Moderate up to Very High. The band is our judgment about how much of a role's typical work is open to change, built from published research on AI and occupations and from government task and employment data. Bands are not surveys, and they don't measure layoffs.
What the index shows
1. Most roles sit in the middle of the scale. Of the 100 roles, 52 rate Moderate or Moderate to High, 29 rate High or Very High, and 19 rate Low to Moderate. Seven rate Very High: accountants and bookkeepers, administrative assistants, credit analysts, customer service representatives, IT support specialists, payroll specialists, and translators and interpreters.
2. When a job splits, the routine side is the exposed side. Of the 100 roles, 30 carry a split rating, because one kind of work inside them is far more exposed than another. In all 30, the routine, production or administrative work rates higher than the judgment and relationship work. Tax preparation is a clear case: simple returns rate Very High, while complex returns and advice rate Moderate.
3. In the Moderate band, AI mostly changes how the work gets done. Of the 37 roles rated Moderate, 29 are classed as augment-heavy, meaning most of the change comes from tools that speed up work a person still directs. Many are management and relationship roles, such as IT managers, sales managers, fundraisers and real estate agents.
4. Hands-on and in-person work anchors the bottom of the scale. The 19 roles rated Low to Moderate include seven of the 13 healthcare roles and six of the seven roles in skilled trades, transportation and public safety. In these jobs AI mostly reaches the paperwork around the work, not the work itself.
5. For the most exposed roles, government projections point the same way. The U.S. Bureau of Labor Statistics projects 12 of the 100 roles to lose at least 1% of their jobs between 2025 and 2035, and 11 of those 12 are in this index's top two bands. BLS's projections name AI as a factor for 20 of the 100 roles, and for 17 of them BLS expects AI to hold down employment. Those projections reflect many forces besides AI, and the overlap describes these 100 roles only.
6. High exposure is not the same as decline. Of the 29 High and Very High roles, 13 are still projected to grow at least 1% through 2035. At the other end, three of the 10 fastest-growing roles in the index rate Low to Moderate: physician assistants, mental health counselors, and speech-language pathologists.
A highly exposed role can keep growing while the mix of tasks inside it changes. Each role page explains where the work is moving.
The 100 roles
Sorted from most to least exposed, then alphabetically. Select a role to see its tasks in each group and the directions experienced people in that role tend to move into.
How the ratings work
We start from tasks, not job titles. For each role we use the U.S. Department of Labor's O*NET task lists (version 31.0) and sort the core tasks into the three groups above.
The band is our judgment, and we show what it rests on. For every role we look at:
- Two research measures of AI exposure: the AI Occupational Exposure measure from Felten, Raj and Seamans (2021) and the task-level estimate of exposure to large language models from Eloundou, Manning, Mishkin and Rock (2024).
- The Bureau of Labor Statistics' AI exposure categories, released in August 2026, which combine those two measures with three others, including observed AI use, into four relative groups.
- BLS employment projections for 2025 to 2035, including the reasons BLS gives when it expects an occupation's share of jobs to change.
- The role's task mix: how much of the typical week is information work AI can complete, how much is work AI speeds up while a person stays in charge, and how much depends on presence, relationships or accountability.
Our band and BLS's category answer different questions. BLS ranks every occupation against all others on how much of its work AI could touch or has been observed touching, and BLS says exposure "does not imply job loss." Our band weighs what kind of change it is. That is why some roles BLS places in its highest group, such as management consultants and marketing managers, rate Moderate here: most of their exposed tasks are ones AI speeds up rather than takes over. Both appear in the data file so you can compare them.
When the evidence splits within a role, we split the band by type of work instead of forcing a single number. We don't publish percentages or scores, because occupational data doesn't support that level of precision.
The full method, with source tiers and a bibliography, is on How we read AI impact.
Limits of this index
- These are estimates of exposure. They are built from published research about tasks and occupations. They don't measure hiring, wages or layoffs.
- Ratings describe a typical version of each role. A specific job can differ with the employer, the specialty and the person's seniority. The free AI Impact Check rates your own job from a description of what you do.
- One hundred roles is still a selection. We chose them for relevance to people in the middle of their careers, for employment size and to cover every major type of work. The set leans toward office and professional jobs, so it is not a sample of the whole labor market, and the counts above describe these 100 roles only.
- The scale's lowest band is Low to Moderate. Roles with very little exposure, such as plumbers and electricians, share that band with nurses and teachers.
- Some roles combine occupations. 10 roles draw on more than one government occupation code, and Product Manager has no code of its own. For combined roles, the projected change is our sum of BLS's figures, not a number BLS publishes. The data file labels each mapping.
- Projections are BLS's, not ours. Employment figures come from BLS's 2025–35 projections, which reflect many factors besides AI.
- AI is changing quickly. We'll update the ratings and date each revision.
Use the data
Download the data (CSV, 100 roles). Each row lists the role, its exposure band, the tasks in each of the three groups, common next moves, the sources behind the rating, the role's occupation codes, BLS's AI exposure category and BLS's projected employment change for 2025 to 2035.
Remix Career's ratings, task groupings and text are published under a Creative Commons Attribution 4.0 license (CC BY 4.0). You can reuse them as long as you credit Remix Career and link to this page. The license covers our own work only. Figures from the Bureau of Labor Statistics and O*NET, and the research measures we cite, belong to their publishers and remain under their own terms, so cite those sources directly when you use them.
Remix Career. AI Impact Index 2026: AI exposure across 100 mid-career roles. Version 2, October 2026. https://www.remixcareer.com/ai-impact-index
Writers and researchers with questions about the method can reach us through the contact page.
Check your own job
The free AI Impact Check sorts the tasks you describe into the same three groups. It takes a role and a few sentences about your week, and it doesn't need an account.
Thinking about a bigger change? Start with the free Career Fit Analysis. It shows your career archetype and how you work best.