Jobs / Education and social services
Can agents do education and social services work?
Teaching, library work, counseling, and social services.
- US jobs
- 11,769,270 (7.6% of employment)
- Typical wage
- $60K a year (employment-weighted mean of occupation medians)
- Occupations
- 86 in O*NET 31.0
- Evidence
- 0 direct benchmarks, 0 also relevant, 2 context only
The answer today
UnmeasuredNo change since 10 Oct 2025.
Rule: No direct benchmark gives a success rate.
Next step: A direct benchmark that reports how many tasks agents complete.
How much of the work is measured
0% of the work in this family is measured by direct benchmarks. Little of the work is measured (less than 20%). On 10 Oct 2025: 0%.
Work = the O*NET 31.0 work activities of the family's occupations, weighted by how important each task is and by US employment. A benchmark measures an activity when its tasks make the agent do it and its grading checks it. Method.
| Work activity (O*NET) | Share of the work | Direct benchmark |
|---|---|---|
| Develop educational programs, plans, or procedures | 7.2% | Not measured |
| Advise others on educational or vocational matters | 5.7% | Not measured |
| Teach academic or vocational subjects | 5.5% | Not measured |
| Assess student capabilities, needs, or performance | 4.8% | Not measured |
| Monitor individual behavior or performance | 4.2% | Not measured |
| Prepare informational or instructional materials | 4.1% | Not measured |
| Maintain current knowledge in area of expertise | 2.9% | Not measured |
| Set up classrooms, facilities, educational materials, or equipment | 2.6% | Not measured |
| Collaborate in the development of educational programs | 2.4% | Not measured |
| Maintain operational records | 2.3% | Not measured |
Our reading
No direct benchmark gives a success rate for teaching or social services work, so the family is unmeasured. Ai2's TutorMoments-Preview gives context only, because it scores single tutoring moves, not completed teaching: GPT-6 Astra makes the move that teachers called for at 0.93 of 520 key moments from real math tutoring transcripts, in short replays with a simulated student that an LLM judge scores. Nothing measures whether a student learns. GDPval is broad and sets no step. No benchmark measures lesson planning, grading, advising, or case work. A direct benchmark that reports how many of these tasks agents complete would give the family its first step.
Evidence: TutorMoments-Preview, GDPval by Ai2 and OpenAI. Written 10 Oct 2026.
Benchmarks for this work
Context only
These benchmarks span many occupations, or they test how agents behave. They give context but do not set the step.
Functions: Professional deliverablesTutoring
What we do not know
- Benchmarks measure tasks, not whole jobs. A job is a bundle of tasks, judgment, and relationships.
- Capability is not adoption. A task that agents can do in a benchmark may still be done by people at work.
- Most results come from single runs. Reliability over many runs is often unknown.
Occupations in this family
Sorted by US employment (BLS OEWS May 2025). O*NET sub-occupations share the employment figure of their SOC group. Work measured = the share of the occupation's O*NET work that direct benchmarks test.
How to cite
Credit the original work first: the benchmarks listed on this page, and their institutions.
Then, if you used this page:
Can Agents Work. "Can agents do education and social services work?." https://canagentswork.com/jobs/education-social-services/ (accessed 2026-10-11). CC BY 4.0.@misc{caw-jobs-education-social-services,
title = {{Can agents do education and social services work?}},
author = {{Can Agents Work}},
year = {2026},
howpublished = {\url{https://canagentswork.com/jobs/education-social-services/}},
note = {Accessed 2026-10-11. CC BY 4.0}
}