AI and Lawyers: What's Actually Happening and What to Do

The honest assessment

The legal profession has spent the last two years simultaneously freaking out about AI and pretending it doesn't exist. Both reactions are understandable. Neither is helpful.

Here's what AI can already do in legal work. It can review contracts and flag non-standard clauses faster than a paralegal. It can summarise case law, draft first-pass legal research memos, and generate standard-form documents like NDAs, employment contracts, and simple wills. Claude in particular handles long legal documents well... paste in a 50-page lease agreement and ask it to identify all the break clauses, and it'll do it in about eight seconds. It would take a trainee an afternoon. Tools like Harvey AI (built specifically for law) are already being used by Allen & Overy and other major firms.

What AI can't do is practise law. It can't stand up in court and adjust its argument based on the judge's body language. It can't sit with a divorcing couple and navigate the emotional minefield of asset division. It can't make the strategic call about whether to settle or fight based on twenty years of knowing how this particular opponent operates. It doesn't understand the difference between what the law says and what actually happens in practice. And it hallucinates case citations, which in law is not a minor inconvenience. It's professional negligence.

The parts changing fastest: document review, due diligence, legal research, and routine contract drafting. These are the tasks that have historically kept large numbers of junior lawyers and paralegals employed. That pipeline is narrowing. Some firms are already hiring fewer trainees because AI handles the work those trainees used to do. If you're a mid-career solicitor who still spends significant time on document review... that's worth thinking about seriously.

The profession has a structural advantage though. Regulation. You need to be qualified to give legal advice, appear in court, and handle client money. AI can assist with all of these things but it can't replace the regulated human. For now. The Solicitors Regulation Authority is watching this space carefully, and the rules will evolve, but the regulatory moat buys lawyers time that other professions don't have.

Your exposure level: Medium

Medium exposure for lawyers comes with a massive asterisk. It depends entirely on what kind of law you practise and where you sit in the hierarchy.

If you're a conveyancing solicitor doing volume residential transactions, your exposure is closer to high. AI can already handle most of the document preparation, land registry checks, and standard correspondence. If you're a complex commercial litigator or a specialist in something like maritime law or patent prosecution, your exposure is genuinely low. The complexity, judgement, and domain expertise required create a real barrier.

For most lawyers sitting somewhere in the middle... doing a mix of advisory work, client management, document drafting, and research... medium is right. AI will change how you do the work, but it won't eliminate the need for you to do it. The catch is that it might mean firms need fewer of you. If one lawyer with AI tools can do the research and drafting that used to require two, the maths isn't complicated. The firms that are honest about this are already restructuring. The ones that aren't are just doing it more slowly.

The 90-day action plan

1. This week: test AI on real legal research. Take a question you recently researched manually. Ask Claude and ChatGPT the same question. Compare their answers to yours. Note where they're right, where they're wrong, and where they hallucinate. Understanding the failure modes is essential if you're going to use these tools responsibly.

2. Week two: run a contract review experiment. Take a standard contract you know well. Paste it into Claude and ask it to identify unusual clauses, missing provisions, and potential risks. Compare its analysis to what you'd flag. i was genuinely surprised the first time i did this. It catches things. Not everything. But things.

3. By day 30: build a document drafting workflow. Pick three types of documents you draft regularly. Create detailed prompts that include your firm's standard positions, preferred language, and specific client requirements. Test and refine until the first draft quality is high enough that you're editing rather than rewriting. Time the before and after.

4. By day 45: learn to use AI for case preparation. Use Perplexity to research opposing counsel's recent cases and published positions. Use Claude to summarise lengthy witness statements or expert reports. Feed it the key documents and ask it to identify contradictions or gaps. This is preparation, not advice. The distinction matters.

5. By day 60: address the ethics head-on. Write a short internal note on how your firm should use AI tools. Cover client confidentiality (don't paste client data into free-tier tools), the duty to verify AI-generated research, professional insurance implications, and SRA guidance. If you're the one who wrote the policy, you're the one they consult.

6. By day 75: identify your defensible niche. What do you do that requires genuine expertise, judgement, and human interaction? Complex negotiations? Tribunal advocacy? Specialist advisory work? Whatever it is, that's where you should be spending more of your time. Use AI to free up time from the routine work so you can focus there.

7. By day 90: have the conversation with your firm. Present what you've learned. Show the time savings. Show the limitations. Make recommendations that are practical, not theoretical. "We should explore AI" is useless. "I saved 6 hours last month using Claude for contract review and here's my workflow" is valuable.

AI tools you should be using this week

• Claude for Work — The best general AI tool for legal work right now. Handles long documents well, follows complex instructions, and is less prone to hallucination than some alternatives. Use it for contract review, document summarisation, and legal research. But always verify the output. Always.

• ChatGPT for Work — Better for client-facing drafting and communications. Use it for demand letters, client update emails, and marketing copy for the firm. The GPT-4 model handles legal language well, though it's occasionally too American in its references for UK practice.

• Microsoft Copilot for Work — The integration with Word and Outlook makes it practical for daily legal work. Draft documents in Word with AI assistance, summarise email chains, and pull together meeting notes. If your firm uses Microsoft 365, this is the lowest-friction option.

• Perplexity for Research — Searches current sources with citations. Useful for checking recent case law, regulatory updates, and opposing counsel's public positions. Not a substitute for Westlaw or LexisNexis for authoritative research, but excellent for quick fact-finding.

What to say in meetings

When the managing partner asks about AI strategy: "I've been testing Claude on our standard contract review process. The first-pass quality is good enough to cut review time significantly, but everything still needs a qualified eye before it goes to clients. I can walk through my workflow." That's the right balance. Practical, not evangelical. Acknowledges both the opportunity and the risk.

If a client asks about AI and your fees: "We use AI tools to handle the routine research and drafting faster, which means more of your budget goes toward the strategic work that actually affects the outcome." Clients don't want to pay for a trainee to spend three days on research that a tool can do in an hour. This answer respects that.

In team meetings when people are anxious: "The tools are good at first drafts and research. They're not good at judgement. Our job is to get faster at the first part so we have more time for the second part." Simple, true, and slightly calming.

If the worst happens

If you're made redundant from a legal role, you have options that many professionals don't. A law degree and practical legal experience transfer into compliance, risk management, corporate governance, regulatory affairs, and consulting. Every regulated industry needs people who understand rules, obligations, and how to structure agreements. You just won't be called a lawyer.

The natural adjacent moves: compliance officer, contract manager, legal operations specialist, regulatory consultant, or in-house counsel. The in-house market is actually relatively strong because companies are bringing legal work internal to save costs, and one good in-house lawyer with AI tools can handle a remarkable amount. Legal tech companies are also hiring... they need people who understand both the law and the technology.

i'll say something that might sound harsh but is genuinely meant to help. If you've been made redundant from a legal role and your reaction is "but I'm a qualified solicitor," you're thinking about it wrong. Your qualification is a foundation, not a lifeboat. What matters now is what you can do, not what you're called. If you can demonstrate that you know how to use AI tools effectively in a legal context while understanding the ethical and practical limits, that's a combination very few people have. That's your edge. Use it.