Law was one of the last professions to take AI seriously. It is now one of the fastest to adopt it.
In 2023, a partner at a major New York firm estimated that AI tools had reduced the time required for first-pass contract review by 70%. That's not a projection — it's what was already happening in production workflows at firms with access to enterprise-grade legal AI. By 2025, those tools had reached mid-market firms. By 2026, solo practitioners were using them.
What AI Does Well in Law
The legal profession, viewed from the outside, looks like a unified entity. From the inside, it is a bundle of tasks with wildly different automation profiles.
Contract review and due diligence: AI is now better than junior associates at first-pass review of standard commercial contracts. It is faster, more consistent, and available at marginal cost. The 2023 Goldman Sachs analysis cited legal work as one of the professional categories most exposed to AI disruption — and contract review was the specific function cited most often.
Legal research: Finding relevant case law, identifying precedent, summarizing statutory history. These are tasks that law school trains associates to do well over years of practice. They are tasks that large language models can perform in minutes, with citations. Not perfectly — AI legal research still requires verification by a trained practitioner — but well enough to change the economics of who does it.
Document generation: Standard legal documents — NDAs, employment agreements, standard commercial terms — can now be generated at high quality from templates that AI can adapt intelligently to context. The volume of work that required a lawyer now requires a lawyer to review AI output. That is a structurally different job.
E-discovery: The review of large document sets for relevance in litigation was already being automated before generative AI. The current generation of tools has accelerated this dramatically.
What AI Does Not Do Well in Law
The list is shorter, but it matters.
Judgment in novel situations. The cases that end up in court — the ones that actually require lawyers — are almost by definition the ones where the precedent doesn't cleanly apply. The value of a senior litigator isn't their ability to find case law. It's their ability to construct a theory of the case in circumstances where the law is genuinely uncertain.
Client relationships. The client who is facing a criminal charge, a divorce, a business crisis is not primarily buying legal knowledge. They are buying the judgment, presence, and accountability of a specific person they trust. That is not automatable in any near-term sense.
Courtroom performance. Oral argument, cross-examination, jury selection — these are embodied skills that require reading a room, adapting in real time, and performing under pressure. AI cannot do this, and the courtroom remains the most human part of legal practice.
Ethical and strategic judgment. When to settle and when to fight. When a client's stated goal is not their actual interest. How to navigate a situation where the legally correct path and the right path diverge. These require a human with experience and accountability.
The Actual Picture
What's happening in law is not replacement. It's restructuring.
The profession is bifurcating. At the top — senior partners, specialist litigators, deal lawyers with irreplaceable relationships — the work is becoming more valuable because the commodity layer beneath them has been automated away. At the bottom — junior associates doing document review and basic research — the economics are changing fast.
The law schools graduating 40,000 JDs a year into a market that needs fewer first-year associates for commodity work is a structural problem that hasn't been publicly named yet.
For lawyers: the move is the same as in every other profession. Identify the fraction of your work that requires genuine judgment, accountability, and relationship — and move aggressively toward it.