What a Semester With Legal AI Taught Us
AI tools for lawyers are everywhere now. We ran them through a full semester in our Entrepreneurship Clinic and came away with three big lessons.
- Students and junior lawyers are already using AI. Employers and educators need to set guardrails and teach sound usage.
- Different tools produce very different results. Seeing that firsthand trains better judgment.
- AI's best value is sharpening legal judgment-spotting issues, stress-testing reasoning, and improving drafts.
Legal AI Prep: Start With Reality, Then Add Discipline
We kicked off with training on mainstream legal research platforms and general AI tools. Before that, student usage ranged from "never" to "I start all case research on ChatGPT."
Our goal was to move everyone to the middle: use AI to boost speed and quality, but never outsource judgment. That meant clear expectations, tight prompts, and verification habits baked into each assignment.
Tool Differences Are Real-and They Matter
We ran simulations where student teams tackled the same transactional tasks-drafting, research, market benchmarking, and client emails-using different AI tools. All teams found a standard construction contract, but only half reached the industry-accepted form on the first pass. The rest got there after changing their approach and queries.
That single exercise landed two points: tool coverage and constraints vary, and your search method still drives results. Some teams also missed the client's core question in their emails-proof that no tool can replace your job of framing issues and answering the brief.
General-purpose tools (like those built on popular LLMs) were the most responsive and forgiving of loose prompts-but more prone to unreliable outputs. Specialized legal tools hallucinated less, yet often refused to complete tasks beyond their guardrails. Pick based on the task, then verify.
Delegation, Not Dependence
AI excels at issue-spotting and double-checking. That helps new and experienced lawyers-if you treat AI like a fast junior you always review.
In a live exercise, students built a risk list for a new business opportunity. After consolidating a strong list as a group, we asked an AI tool to issue-spot the same scenario. It added a couple of items no one had raised-plus a few irrelevant ones. Useful, but only with independent judgment in the loop.
Tie this to duty of competence and client confidentiality. Tech competence now includes understanding AI limits and risks. See the ABA's guidance on competence for emphasis on staying current with technology.
A Simple AI Protocol for Clinics and Firms
- Policy: Define approved tools, use cases, and red lines (e.g., no client confidentials in public tools).
- Disclosure: Tell clients when and how AI is used, if your engagement terms call for it.
- Source control: Require citations, links, and document retrieval for any AI-surfaced authority.
- Prompting: Standardize task-specific prompts for research, drafting, and client comms. Iterate, don't improvise.
- Verification: Read every cited case. Cross-check key facts, market terms, and calculations.
- QA workflow: First do your own work. Then use AI to stress-test, expand, and edit. Compare results and revise.
- Data hygiene: Strip identifiers. Use enterprise or walled tools for sensitive matters.
- Tracking: Log where AI helped (time saved, issues found, errors avoided). Improve your prompts monthly.
How We Taught It
We used AI at the end of each exercise. Students first wrestled with the task, then used AI to challenge their thinking and polish the product.
This sequencing preserved learning, raised the quality of the final work, and reduced drafts. It also drew a bright line between AI as a tool versus a crutch.
What Lawyers Should Do Next
- Choose tools by task: general LLMs for brainstorming and drafting, specialized tools for research and retrieval.
- Interrogate, don't accept: ask for sources, definitions, and counterarguments.
- Benchmark your prompts: save the winners, retire the duds, and share across the team.
- Keep judgment central: if you can't verify it, don't delegate it.
Bottom Line
AI won't replace legal judgment-it can refine it. Treat AI like a tireless junior: give clear instructions, demand sources, and verify everything.
Do the first pass yourself. Then use AI to pressure-test, add angles you missed, and tighten your writing. That's how you get faster without lowering your standards.
Further Learning
- Structured AI training by job role for building practical skills you can apply to legal work.
Your membership also unlocks:
