How Lawyers Actually Use AI in Practice
A practical view from daily legal work: where AI helps, where it fails, and how lawyers should use it responsibly.
There is a lot of noise around AI and the legal profession right now. Some of it is hype. Some of it is fear. As a practising lawyer working primarily in civil matters, I prefer a more practical conversation: how does AI actually fit into the work lawyers already do every day?
In my practice, AI does not replace legal judgment. It helps with the repetitive, time-consuming parts of legal work: organising drafts, reviewing contracts, translating documents, and finding a starting direction for research. Used properly, it saves time and improves consistency. Used blindly, it creates risk.
Key takeaway AI is useful in legal practice when it works under a lawyer's supervision. The lawyer still controls the facts, strategy, legal reasoning, verification, and final work product. |
Drafting: AI Helps After the Legal Thinking Is Done
A lot of people assume that using AI to draft means typing one short prompt and receiving a finished legal notice, plaint, contract, or application. That is not how it works in serious legal practice, and it should not be how any lawyer uses AI.
My process starts before AI enters the picture. I sit with the client, understand the facts, identify what the client needs, and then do the legal research required to decide what must go into the draft. Only after that groundwork is complete do I bring AI into the workflow.
My drafting workflow
1. Understand the client's facts and documents.
2. Identify the legal issues, reliefs, risks, and procedural requirements.
3. Research the relevant law, limitation issues, and practical strategy.
4. Prepare clear drafting instructions with the necessary facts and points.
5. Use AI to organise that substance into a clean, structured first draft.
6. Review, edit, and finalise the document myself before it goes anywhere.
The important point is this: I do not ask AI to create a reliable legal document from incomplete facts. I give it the substance. It helps me structure that substance more efficiently.
Where Lexi helps with drafting consistency
This is where a law-specific AI tool like Lexi is more useful than a generic chatbot. Lexi is designed around legal workflows, including drafting documents in a firm's style, reviewing and redlining contracts, and supporting legal research with verified citations. For drafting, that means I can work with the formats and templates my firm actually uses instead of accepting a generic output that needs to be rebuilt from scratch.
That consistency matters. In legal drafting, format, terminology, sequencing, and phrasing conventions are not cosmetic details. They affect readability, enforceability, client confidence, and the time required for senior review.
Contract Review: A Second Set of Eyes That Does Not Get Tired
The other major way I use AI is in contract review. This is not because AI understands the client better than I do. It is because AI can perform a patient second pass over dense language and flag issues that are easy to overlook during a manual read.
What I ask AI to check
· Clauses that could work against my client.
· Missing clauses that should be included for enforcement or risk control.
· Unusual obligations, hidden restrictions, or one-sided language.
· Inconsistent definitions, timelines, penalties, and termination rights.
· Clauses placed under misleading or generic headings.
An anonymised matter that changed my process
In one anonymised matter, a client came to me with an employment offer and asked me to review it before signing. It looked like routine work. I checked compensation, notice period, termination rights, and the usual employment terms. Nothing immediately stood out.
As a second pass, I ran the document through AI. It flagged a restriction I had missed. The clause was not placed under a heading like 'Non-Compete' or 'Restrictive Covenant'. It was buried under a generic 'General Obligations' section and phrased almost like a formality.
Read closely, however, the clause restricted the employee from joining any company in a related line of work for two years after leaving. It was not limited to direct competitors. It had no meaningful geographic limit. In practical terms, it could have blocked the client from working across a large part of his own industry.
I had skimmed past it because it was not where a lawyer would normally expect that kind of restriction to appear. That is exactly what made it risky. Many contracts place important obligations in predictable locations. This one did not.
We went back to the employer and negotiated the restriction into a narrower, clearer, and commercially more reasonable form. The client signed with a better understanding of what he was agreeing to. Without that second pass, he may have accepted a restriction that affected his career mobility for years.
Practical lesson AI should not be the only reviewer of a contract. But as a second reviewer, it can be extremely useful because it does not get tired, assume where clauses should appear, or skip over boring sections. |
Translation: The Quiet Time-Saver
One use case that does not get discussed enough is translation. A meaningful part of legal practice involves documents moving between English and regional languages: client documents, property papers, notices, correspondence, pleadings, and supporting records.
AI gives me a workable first translation quickly. I still review the legal terminology, context, and meaning myself, because a literal translation can be dangerous in legal work. But the time saved is significant. Work that used to consume a large part of the day can now be completed in a fraction of the time, leaving more room for strategy, drafting, and client communication.
Legal Research: Useful Starting Point, Not Final Authority
I do not treat AI as my primary source for legal research. Case law, statutes, procedure, and precedent are areas where mistakes can have serious consequences. A confident but incorrect answer is worse than no answer at all.
That said, AI can be useful when I am stuck and need a starting direction. It can help frame the issue, identify possible legal angles, or suggest what to look for next. But every proposition, case, section, and citation must be verified against actual sources before I rely on it.
For legal research, source traceability matters. If an AI tool gives me a proposition of law, I need to know exactly where it came from. This is why legal-specific tools with verified citations are more useful than generic answer generators. Even then, verification remains the lawyer's responsibility.
The Tools I Actually Use
For legal work, my primary tool is Lexi because it is built around legal workflows rather than general-purpose writing. I use it most for drafting, contract review, templates, and legal research support.
For non-legal side tasks, early brainstorming, or general writing assistance, I may use tools like Google Gemini or ChatGPT. But for actual legal work product, the distinction matters. Legal work needs structure, terminology, context, source traceability, and a workflow that respects professional responsibility.
Where AI Falls Short
AI is not magic, and it is not always reliable. Drafts can come out weak. Important context can be missed. Generic tools can produce confident language that sounds correct but is legally incomplete. AI can also hallucinate, especially when it does not have enough relevant information and tries to fill the gap.
Common risks I watch for
· Vague prompts producing vague drafts.
· Incomplete facts leading to incomplete legal analysis.
· Overconfident summaries that hide uncertainty.
· Fabricated or inaccurate legal citations.
· Generic drafting that ignores local practice, court format, or firm style.
· Lawyers treating AI output as final instead of as a draft for review.
The rule I follow is simple: verify every single output. Whether it is a draft, a flagged clause, a translation, or a research summary, nothing goes to a client, court, counterparty, or senior lawyer without my own review.
Where Legal AI Is Headed
AI is changing every profession, and law is no exception. I do not believe it replaces lawyers. What it changes is the way lawyers work.
Used carefully, AI can make legal work faster, more consistent, and easier to review. Used blindly, it can expose the lawyer, the client, and the practice to serious risk.
I am already seeing a concerning habit among some younger lawyers: feeding AI half the facts of a case and expecting a complete, courtroom-ready draft in return. That is not responsible use. AI can assist with legal work, but it cannot replace the lawyer's duty to understand the matter.
My Advice to Younger Lawyers
1. Do your legal homework first
Before using AI, understand the facts, documents, procedural posture, client objective, and legal issue. If you do not understand the matter, AI will not solve that problem for you.
2. Give AI complete instructions
Good output depends on good input. Give AI structured facts, the purpose of the document, the audience, the jurisdictional context, the required tone, and the points that must be included or avoided.
3. Verify every output
Treat AI output as a working draft, not a final answer. Check the law. Check the facts. Check the citations. Check the formatting. Check whether the result actually serves the client's objective.
Final Thought
AI is a tool. A very useful one, when used correctly. It can help lawyers draft faster, review more carefully, translate more efficiently, and begin research with more direction.
But the lawyer remains responsible for the final work product. The judgment, verification, ethics, strategy, and accountability still belong to the lawyer. That has not changed, and it should not change.