Will AI replace computer research scientists?
Part of the work, not the whole job. AI could take on about 59% of the working time in a typical computer research scientist job today. The other 41% still needs a person.

41% of the work still needs a person.
The bracket under the bar is the evidence range for the share AI could take on: 35 to 59%. One end comes from how well AI did each responsibility in real use, or a second rating where that isn't measured. The other is our figure, which also follows research across the industry.
9 responsibilities in a typical computer research scientist job. Median pay $140,300 a year in the U.S. (BLS, May 2025).
Still needs a person
- Meet people to resolve problems
- Invent new computing technology AI helps
- Analyze problems and design computing solutions AI helps
- Design computer hardware and software AI helps
- Gather requirements from users and vendors
- Build math models of technical problems AI helps
AI can do or help with
- Analyze problems and design computing solutions AI helps
- Invent new computing technology AI helps
- Build math models of technical problems AI helps
- Design computer hardware and software AI helps
- Set work priorities for the team AI helps
- Judge whether project plans are feasible AI helps
The biggest responsibilities on each side, ranked by a person's time on the left. "AI helps" means AI takes on part of the work, or most of it with a person checking what it produces, so one can show on both sides. See all 9, with the reason for each and the cost of handing them to AI.
Is this your job?
Upload your résumé to see your Human Edge: how much of this job your experience holds in the 41% that still needs a person. You also get a free plan with what to learn next. We never keep your résumé.
Questions people ask
Will AI replace computer research scientists?
Part of the work, not the whole job. AI could take on about 59% of the working time in a typical computer research scientist job today. The other 41% still needs a person. That is a split of the work, not a forecast of jobs lost.
Which parts of the job can AI do?
The biggest ones AI can do or help with: analyze problems and design computing solutions, invent new computing technology, build math models of technical problems, design computer hardware and software, set work priorities for the team, judge whether project plans are feasible.
Which parts still need a person?
The biggest ones: meet people to resolve problems, invent new computing technology, analyze problems and design computing solutions, design computer hardware and software, gather requirements from users and vendors, build math models of technical problems.
How is this measured?
The U.S. Department of Labor lists 15 responsibilities for this job (O*NET 31.0), and the split covers the 9 people in it do today. Each is scored on 4 questions: can AI do it, does it get it right, must a person sign for it, and do people need to deal with a real person. The split is weighted by how much of the job each responsibility takes up. The evidence range for the share AI could take on is 35 to 59%. One end comes from how well AI did each responsibility in real use, or a second rating where that isn't measured. The other is our figure, which also follows research across the industry. Scores version 0.8.13, updated November 20, 2026.
Responsibilities from O*NET 31.0 (U.S. Department of Labor). Pay from the Bureau of Labor Statistics, May 2025. Scores are calibrated to published research and updated every month. How it works.
Related jobs
- Will AI replace document management specialists? 40% needs a person
- Will AI replace IT support specialists? 39% needs a person
- Will AI replace operations research analysts? 39% needs a person
- Will AI replace information security engineers? 38% needs a person
- Will AI replace digital forensics analysts? 38% needs a person
- Will AI replace cybersecurity analysts? 37% needs a person