Prompt Testing for Law Firms
Software testing methods applied to legal AI prompts, so that your firm knows how reliable a prompt is before a junior lawyer uses it.
The Problem
The Approach
Software quality assurance applied to legal AI prompts: the testing principles used in software development, adapted for legal work.
Systematic Testing
Run each prompt 20 times on real and synthetic data. Measure consistency (the same input gives the same output) and accuracy (the output against a lawyer's review).
Five Validation Layers
The AI checks its own answer, scores its confidence, and is checked by rules written in code. The firm's own context is added, and each task is graded by complexity, so a junior lawyer knows when to escalate.
Automation
Automated test runs, quality dashboards, version control for prompts and regression testing, so that less of the testing is done by hand.
Knowledge Transfer
A methodology handbook, team training and documentation, so that your team can maintain the prompts after the engagement ends.
Service Packages
The price depends on the size of your firm and the work involved. Email hello@counsel.directory for a quote.
What the Work Takes
Legal Work
How law firms work, how reliable legal work has to be, and the duties of professional conduct.
AI and Prompt Engineering
How large language models work and fail, and how to write prompts for them.
Software Testing
Quality assurance, version control and regression testing, applied to prompts.
Testing a legal prompt takes all three: knowledge of the legal work, of the AI model, and of how software is tested.
Get in Touch
Interested in prompt testing for your firm? Describe what you need.
Pilot projects are open to firms in England and Wales. No prompt in this library has published test results yet. When one has, its page will show them.
Email: hello@counsel.directory
Free Resources
The prompts and the methodology on this site are free to read and use.