AI for lawyers: practical, confidential, and by the rules

What AI means for legal practice

In 2023, a New York lawyer stood before a judge because his brief cited six court cases that did not exist. He had asked ChatGPT to find case law and filed the answer without reading it. That is the story every lawyer hears when AI comes up, and rightly so.

But it is not the whole story. A contract review that costs a junior 4 hours takes AI 10 minutes. Due diligence across 200 documents goes from a week to a day. A targeted case law question yields a concise, annotated list within minutes. That is not marketing, that is what happens once you set it up properly.

The problem is that "setting it up properly" carries a much higher bar in legal work than in marketing or finance. Confidentiality, accuracy, and liability are not side issues, they are the foundations. Bar rules, the duty of confidentiality, professional liability: a hallucination in a pleading can trigger disciplinary consequences. A wrong assumption in advice can cost millions. This article shows where AI already works in a law firm today, where it is outright dangerous, and how to start without repeating the New York mistake.

Where AI is already being used

Contract review and due diligence

This is the clearest win. Kira Systems scans hundreds of contracts for specific clauses, risks, and deviations from standard terms. For an M&A transaction with 50 contracts or more, that is the difference between a week of junior hours and a few hours of review. The human reads the signal, not the whole stack. Alternatives: Luminance, eBrevia, and LeXial for smaller firms.

Legal research

ROSS Intelligence searches case law and statutes using natural language. No more puzzling over search terms in legal databases, just ask the question the way you would ask a colleague: what does the Court of Appeal say about missed deadlines for insurance claims after a loss event. For the US and UK, Lexis+ AI, vLex, and since 2024 Westlaw as well come with built-in AI. Important: pick a tool that cites from a controlled database, not a chatbot that guesses from memory.

Drafting documents

LawGeex drafts and reviews legal documents against pre-set criteria. Standard contracts, NDAs, terms and conditions, summonses, deeds: AI produces the first version, the lawyer refines and signs. For a small or mid-size firm, Claude or ChatGPT with a good prompt and your own template library often does the same work, at a fraction of the cost of the heavy enterprise packages. Start there.

Predictive analysis

Based on prior rulings, AI predicts the likely outcome of a case. Not as a replacement for legal judgment, but as one data point alongside your own read. Especially useful for settle-or-litigate decisions or estimating timelines. Tools: Lex Machina (mainly US), Premonition. In smaller jurisdictions this is less developed, simply because the body of case law is smaller.

Standard correspondence and client management

The dullest category and also the place where you can start tomorrow. Letters to opposing counsel, invoices with explanatory notes, intake questions for new clients, matter summaries for periodic updates. Low risk, immediate time savings, no tooling investment. If you do not know where to start, start here.

The challenges

Confidentiality

Attorney-client privilege is sacred and legally protected. Every AI tool has to meet data protection rules, and data must never be used to further train the model. Cloud AI with a data processing agreement works fine for most work. For the most sensitive matters (M&A, criminal defense, international arbitration), running on your own servers or in a private EU environment is worth serious consideration. Case by case, not the default.

Hallucinations and factual errors

The biggest risk in legal work and the reason the New York lawyer walked into sanctions. AI cites incorrect case law in a confident tone, invents statutory provisions, and misreads clauses, all in sentences that look authoritative. Generic ChatGPT as a research aid for briefs is simply a bad idea. Full stop.

The fix is threefold. One: work with systems that only cite from controlled source documents (official court databases, EUR-Lex, your own contract library), not from training data. Two: every AI claim points back to its source so you can always verify. Three: for pleadings and client advice, thorough human review is the minimum, not an option. If you cannot commit to that, do not use AI for the task.

Transparency

If AI makes a recommendation, the lawyer has to be able to explain why. Opaque, untraceable algorithms are unacceptable in court and in advice to clients. Ask every vendor explicitly how the output is built and which sources were consulted. Under the AI Act, this is also a legal requirement for high-risk applications, and legal work almost always falls in that category.

Liability

If AI misses a clause in a contract review, who is liable? The vendor? The firm? The lawyer? Under professional rules, the lawyer remains ultimately responsible, so arrangements for human oversight are not optional. Align your professional liability insurance with your use of AI too. Most policies only cover AI errors when documented human review took place, and "I trusted the tool" is not a defense.

Bias in historical data

Legal data carries historical bias: decisions that used to be normal but read as discriminatory today. AI systems trained on that data reproduce those biases, especially in predictive systems for case outcomes or sentencing. Work with systems that are transparent about their training data and methods, and treat outputs as a signal, not a verdict.

Generic ChatGPT as a research aid for briefs is simply a bad idea. Every AI claim has to point to a verifiable source.

Getting started safely with AI: a roadmap for law firms

Month 1: analysis and goals. Which processes cost the most time right now: contract review, standard correspondence, case law research? Which of those carry high liability risk and which are low risk? Train your team in AI basics, then specifically in legal AI, not the generic online course.

Month 2-3: pilot on a low-risk process. Do not start with pleadings. Start with standard correspondence or contract templates. Tools: Claude or ChatGPT (business version with a data processing agreement) plus a fleshed-out template library. Approach: AI produces the draft, the lawyer reviews. Measure: time per document, error rate in the first version, client satisfaction.

Month 4-6: scale and expand. Only once the first pilots are running do you move to contract review or case law research. Here you set up a system that draws on your own contract library or a controlled case law database. Document oversight arrangements and everything you do.

Month 7+: comply with the AI Act and tighten confidentiality. Document every AI application for the AI Act: which AI you use, for which processes, what data it processes, how oversight is arranged, how clients can object. For the most confidential work, explore on-premise options.

How a project runs

A project runs in phases. First a short analysis to find the spots with the highest impact in your firm. Then a targeted pilot on one tool (a contract scanner or due diligence assistant, for instance) to test whether it fits your way of working. If it does not work, we stop and pick something else. If it does, we scale. Between phases you evaluate whether the project continues, so the firm adjusts step by step and can shape the direction. For matters with strict confidentiality requirements, we look case by case at your own servers or a private EU environment.

The future

AI makes legal services more accessible, faster, and more affordable. Not by replacing lawyers (that cannot and must not happen), but by freeing them from routine work so they can focus on what actually adds value: thinking, negotiating, advising, arguing.

The firms investing in AI now are building a lead that is hard to catch up with. Not just in efficiency, but also in client satisfaction (faster answers, better predictability on timelines) and in talent (juniors do not want to spend their days on document review). The story of the New York lawyer is not a reason to avoid AI, it is a reason to organize it seriously.

Want to know which processes in your firm are best suited for AI? Take the free AI Scan and get a personal analysis in your inbox within 1 minute.