Jobs / Legal
Can agents do legal work?
Legal research, drafting, review, and support.
- US jobs
- 1,305,620 (0.8% of employment)
- Typical wage
- $121K a year (employment-weighted mean of occupation medians)
- Occupations
- 8 in O*NET 31.0
- Evidence
- 3 direct benchmarks, 1 also relevant, 3 context only
The answer today
Some tasksFirst measured 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
- Legal Research Bench by Vals AI: 55.3% of tasks completed · Muse Spark 1.3 Max, 22 Sep 2026
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
41% of the work in this family is measured by direct benchmarks. Some of the work is measured (20% to 50%). 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 |
|---|---|---|
| Prepare legal or regulatory documents | 16.4% | Measured |
| Consult legal materials or public records | 13.8% | Measured |
| Present information in legal proceedings | 12.4% | Not measured |
| Research laws, precedents, or other legal data | 11.3% | Measured |
| Advise others on legal or regulatory matters | 7.8% | Not measured |
| Discuss legal matters with clients, disputants, or legal professionals or staff | 7.6% | Not measured |
| Interview people to obtain information | 3.9% | Not measured |
| Mediate disputes | 3.1% | Not measured |
| Direct legal activities | 3.0% | Not measured |
| Supervise personnel activities | 3.0% | Not measured |
Our reading
Agents complete some legal research tasks. On Vals AI's Legal Research benchmark, the best models answer 55.29% of 208 private questions with every item of a lawyer-written rubric met; three models tie, with a standard error of about 3.5 points. Work products are harder: on Harvey LAB, run by Vals AI, the best model's memo or redline passes every criterion 25.42% of the time. RedlineBench gives 50.5% of rubric credit for one turn of a contract negotiation, with no completion rate. All three use LLM judges. No direct benchmark measures presenting in legal proceedings or advising clients. No legal benchmark is at 60% yet; the next step needs two from independent institutions.
Evidence: Legal Research Bench, Harvey Legal Agent Benchmark, RedlineBench by Crosby Legal, Harvey, micro1, and Vals AI. 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: Financial analysisConsulting workLegal workProfessional deliverablesLegal researchWeb researchContract redliningTax research
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.
| Occupation | SOC | US jobs | Work measured |
|---|---|---|---|
| Lawyers | 23-1011.00 | 755K | 38% |
| Paralegals and Legal Assistants | 23-2011.00 | 393K | 65% |
| Title Examiners, Abstractors, and Searchers | 23-2093.00 | 49K | 0% |
| Legal Support Workers, All Other | 23-2099.00 | 47K | — |
| Judges, Magistrate Judges, and Magistrates | 23-1023.00 | 24K | 0% |
| Administrative Law Judges, Adjudicators, and Hearing Officers | 23-1021.00 | 16K | 0% |
| Judicial Law Clerks | 23-1012.00 | 13K | 0% |
| Arbitrators, Mediators, and Conciliators | 23-1022.00 | 9K | 0% |
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 legal work?." https://canagentswork.com/jobs/legal/ (accessed 2026-10-11). CC BY 4.0.@misc{caw-jobs-legal,
title = {{Can agents do legal work?}},
author = {{Can Agents Work}},
year = {2026},
howpublished = {\url{https://canagentswork.com/jobs/legal/}},
note = {Accessed 2026-10-11. CC BY 4.0}
}