AI in Legal: Contract Review in Minutes, Research at Scale
Law is language, and language is what AI does best. That is why the legal profession β famously slow to change β is adopting AI faster than almost anyone expected. The work of reading a thousand pages to find the three that matter is exactly the kind of work a model can do in minutes. The catch is that in law, a confident mistake is not an embarrassment. It is a liability.
Where AI helps most
Document review, at machine scale
Due diligence for a merger, discovery for a lawsuit, a lease portfolio for a property deal β all of it means reading enormous volumes of text for risk. AI review tools now surface the unusual clause, the missing warranty, the inconsistent date, and summarize what each document says. What took a team of junior lawyers three weeks now takes an afternoon, with the humans checking the machine's flags rather than reading every page.
Research that reads everything
Legal research used to mean knowing which keywords to search. AI assistants can now answer a question over a firm's whole library β statutes, case law, past memos β and point to the source for each claim. For small firms and solo practitioners, that is a leveller: research capacity that once required a big-firm library is now available to anyone.
Drafting the first version
Contracts, letters and filings all start from patterns, and AI is good at patterns. A first draft in seconds, tailored to the facts, lets lawyers spend their time on the negotiation and the judgment β the parts clients actually pay for. Standard documents for small businesses, like NDAs and service agreements, are becoming cheap enough to do properly instead of copying a template from the internet.
Access to justice
Most people cannot afford a lawyer, and most legal need goes unmet. AI-powered legal help β explaining a tenancy dispute, drafting a small-claims letter, translating legalese into plain language β is reaching people the profession never served. It is not a lawyer, and it should not pretend to be one. But a good answer to a simple question is far better than no answer at all.
Where to be careful
Law has already produced AI's most famous cautionary tales: lawyers who filed briefs citing cases the AI invented, complete with realistic-looking citations. Hallucination is not a bug lawyers can tolerate β every citation must be verified, every claim checked against a real source. Confidentiality is the other hard line: client information fed into a public chatbot may not stay confidential, and privilege rules were not written with AI in mind. Regulators are also asking who is practicing law when software gives legal advice. The firms getting this right treat AI as a very fast paralegal: useful, supervised, and never the one signing the filing.
What it means for people in law
Junior lawyers are spending less time on document review and more on client work β which is good for their careers and hard for the old training model, since review was how juniors learned. Partners are rethinking billing: when a task takes minutes instead of hours, charging by the hour stops making sense, and value-based pricing is finally arriving. For clients, legal help is getting faster and cheaper, especially for routine work. The lawyer's durable advantage is the same as ever: judgment, negotiation, and the trust of a client who needs a human to stand behind the advice.