Skip to main content
A/KAI Kills What?

Research brief · ranked 25

The 25 jobs safest from AI in 2026

Same pipeline, ranked from the other end: well-evidenced occupations whose core tasks current AI systems cover the least. These are jobs where physical presence, dexterity, regulated judgment or in-person trust still carries the work - and where adoption evidence is thin, not absent. Scores measure task automation potential, not job disappearances; a low score with strong evidence is the honest reading of "AI-resistant today". Pipeline as of 2026-11-16, refreshed hourly.

Jobs safest from AI: rank, occupation, AI exposure score, evidence count, workforce size and industry
#OccupationAI exposureEvidence
1Stockers and Order Fillers11%1
2Sound Engineering Technicians13%1
3Pharmacy Technicians14%1
4Film and Video Editors17%5
5Training and Development Specialists18%1
6Tutors18%1
7Photographers20%2
8Sports Medicine Physicians20%2
9Medical Transcriptionists21%1
10Accountants and Auditors21%4
11Lawyers22%4
12Computer Programmers22%3
13Search Marketing Strategists23%3
14Secretaries and Administrative Assistants, Except Legal, Medical, and Executive27%3
15Financial and Investment Analysts27%4
16Marketing Managers28%4
17Graphic Designers29%4
18Editors31%2
19Project Management Specialists31%2
20Human Resources Specialists32%3
21Bookkeeping, Accounting, and Auditing Clerks33%3
22Web Developers35%4
23Software Developers35%9
24Computer Network Support Specialists36%1
25Clinical Data Managers37%8

Why "safest" is still evidence, not a promise

"Low exposure" means today's best models and agents demonstrably cover few of the core task statements - not that the job is future-proof. Tasks get re-scored as new capability evidence lands (see the live radar), so treat this ranking as a stock-taking you can re-check any month. Method and limits: how we score AI exposure.

Cite or embed

Every job links to an interactive dossier; the full table is open data at /data/jobs.csv and the data page. Attribution: task statements © O*NET OnLine, U.S. Department of Labor (CC BY 4.0); scores by AIKillsWhat.com, methodology.