Jobs / Healthcare
Can agents do healthcare work?
Clinical and technical healthcare work, and healthcare support.
- US jobs
- 17,725,580 (11.4% of employment)
- Typical wage
- $76K a year (employment-weighted mean of occupation medians)
- Occupations
- 116 in O*NET 31.0
- Evidence
- 3 direct benchmarks, 2 also relevant, 2 context only
The answer today
Some tasksNo change since 10 Oct 2025.
Rule: Agents complete 20% or more of the tasks (or projects) on a direct benchmark.
The evidence that sets this step
- MedAgentBench by Stanford University: 69.7% of tasks completed · Claude 3.5 Sonnet v2, 24 Jan 2025
Next step: Most tasks needs 60% or more of the tasks (or projects) on two direct benchmarks from independent institutions.
How much of the work is measured
1% of the work in this family is measured by direct benchmarks. Little of the work is measured (less than 20%). On 10 Oct 2025: less than 1%.
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 |
|---|---|---|
| Assist individuals with special needs | 8.1% | Not measured |
| Train others on health or medical topics | 6.5% | Not measured |
| Maintain health or medical records | 6.4% | Measured |
| Administer basic health care or medical treatments | 6.4% | Not measured |
| Confer with healthcare or other professionals about patient care | 3.4% | Not measured |
| Monitor health conditions of humans or animals | 3.4% | Not measured |
| Assist healthcare practitioners during medical procedures | 3.0% | Not measured |
| Analyze health or medical data | 2.8% | Measured |
| Evaluate patient or client condition or treatment options | 2.6% | Measured |
| Explain medical information to patients or family members | 2.0% | Not measured |
Our reading
Agents complete some healthcare tasks, all inside clinical software. On Stanford's MedAgentBench, Claude 3.5 Sonnet v2 completes 69.67% of 300 short electronic health record tasks (January 2025; no leaderboard, so newer models are not measured). On Microsoft's HealthAgentBench, Claude Code with Opus 5 passes 55% of 162 trials on 54 tasks such as X-ray report correction. On Stanford's PhysicianBench, GPT-5.6-sol completes 50.0% of 100 e-consult workflows on one try, and 31.0% in all three runs. No benchmark measures work with patients: assisting people with special needs and giving basic care are the largest activities. Step 3 needs a second benchmark, not from Stanford, at 60% or more.
Evidence: MedAgentBench, HealthAgentBench, PhysicianBench by Microsoft Research and Stanford University. Written 24 Sep 2026.
Benchmarks for this work
Direct evidence
Work benchmarks mapped primary to this family. They set the step when they report how many tasks agents complete.
Also relevant
These benchmarks cover part of this family's work, or have no result yet. They do not set its step.
Context only
These benchmarks span many occupations, or they test how agents behave. They give context but do not set the step.
Functions: Professional deliverablesMedical image readingClinical records workData engineering and SQLComputer useEnterprise workflowsProcedure execution
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 healthcare work?." https://canagentswork.com/jobs/healthcare/ (accessed 2026-10-11). CC BY 4.0.@misc{caw-jobs-healthcare,
title = {{Can agents do healthcare work?}},
author = {{Can Agents Work}},
year = {2026},
howpublished = {\url{https://canagentswork.com/jobs/healthcare/}},
note = {Accessed 2026-10-11. CC BY 4.0}
}