Skip to content

AI vs. human capability

AI Autonomy Scores by Task

How autonomously AI can handle each legal task, on a 1-5 scale with confidence intervals.

MeasuredUpdated 2026-07-08

Across 15 legal tasks in HAQQ's usage data, autonomy scores mostly cluster between 2.78 and 3.10 on a 1-5 scale, with no task reaching consistent independent execution (4+). One task, advising on business transactions and lawsuits, stands higher at 3.38, though its wide confidence interval (2.87-4.09) signals inconsistent rather than reliably strong performance. The lowest, explaining tax law, sits at 2.78. The pattern reads as collaborative, not autonomous: AI functions as a capable assistant across the legal workflow, not a substitute for judgment on any single task.

The data

AI Autonomy Scores by Task - How autonomously AI can handle each legal task, on a 1-5 scale with confidence intervals.
TaskAutonomy (1-5)95% CI low95% CI high
Read & rewrite legal documents2.952.873.03
Advise on business transactions & lawsuits3.382.874.09
Advise on financial & legal matters2.812.613
Advise on market conditions & legal requirements2.862.732.97
Obtain patents & meet legal requirements2.972.643.28
Modify compensation policies for compliance2.952.53.4
Analyze building codes and by-laws2.92.433.38
Analyze probable outcomes using legal precedents3.12.573.62
Determine liability per laws and precedents2.892.763.02
Examine legal documents for court adherence2.882.583.17
Examine records for compliance2.872.673.07
Explain tax laws2.782.53.05
Gather & analyze legal research data2.822.543.1
Interpret & explain policies and laws2.842.623.05
Interpret laws, rulings, and regulations2.892.633.14

Measured data - pre-processed from HAQQ's real Legal AI Index usage dataset (134,000+ data points across 30 countries) or HAQQ's own /50 cross-model legal benchmark.

Back to the Legal AI Index report

Related research

All research graphs