Lawyer · Advises clients, drafts the documents that bind them, and argues their case when it reaches court — carrying personal legal liability if the advice is wrong.
Software can already draft a first-pass non-disclosure agreement, summarize a thousand-page discovery production, and flag the risky clause in a merger contract — tasks that used to occupy junior associates for entire billable weeks now take minutes.
This page separates what is already being automated from what has proven far more resistant, sets out how the risk compares with the rest of the profession, and looks at the new roles opening up around AI-assisted legal practice.
Roughly two-fifths of the profession's work — document review, first-draft contracts, legal research summaries and due-diligence flagging — is already well suited to automation, and a widely cited 2023 Goldman Sachs analysis ranked legal services among the most exposed white-collar fields to generative AI. The remainder depends on courtroom persuasion, client judgment and a personal legal liability no software can hold, which keeps the profession's overall exposure moderate rather than severe.
Scored from the tasks, not the job title. Lower is safer.
Jobs AI cannot take →A licensed lawyer can be sued for malpractice, disbarred or held in contempt; no AI system can be sued, disbarred or held personally accountable, so a human must remain the legally responsible party for any advice given.
Reading a judge's or jury's reaction in real time and adapting an argument on the spot is a live, improvisational skill current AI cannot replicate inside an actual courtroom.
Weighing a client's risk tolerance, finances and non-legal priorities against the law's technical answer to give practical advice requires judgment that goes beyond correctly stating what the law says.
Constructing a genuinely new argument to extend or challenge existing precedent, rather than pattern-matching against past cases, is exactly the kind of reasoning current AI tools struggle to originate reliably.
Reading motive, building trust and knowing when to bluff or concede against an opposing lawyer remain fundamentally interpersonal skills that resist full automation.
Machine-learning-assisted 'technology-assisted review,' court-approved in the US since 2012, already handles the first pass of sorting huge document productions by likely relevance, work that once occupied teams of associates for weeks.
AI drafting tools now produce workable first drafts of standard agreements — NDAs, simple leases, routine employment contracts — from a short set of client instructions, leaving a lawyer to review and adjust rather than draft from a blank page.
AI research tools can summarize case law and statutes quickly, though a well-publicized 2023 US sanctions case, in which lawyers submitted a brief citing entirely fabricated AI-generated cases, showed the real risk of trusting the output unverified.
In mergers and acquisitions, AI tools now scan huge data rooms of contracts to flag clauses that might trigger risk — change-of-control provisions, unusual liabilities — a first-pass task that once consumed junior associate weeks.
As AI produces more first drafts, junior lawyers' time shifts toward reviewing and correcting machine output rather than drafting from scratch — raising a real question about how the next generation learns judgment it used to build through drafting practice.
Clients increasingly push for flat fees or subscription pricing instead of hourly billing, especially as AI compresses the time a task actually takes, making pure hourly billing harder to justify to a client who knows the software exists.
Alternative legal service providers and legal-process outsourcers now handle commoditized tasks like contract review and compliance monitoring more cheaply than a traditional law firm, competing directly for work firms once did in-house by default.
Following well-publicized cases of lawyers filing briefs with AI-fabricated citations, courts in multiple countries have begun issuing standing orders that require lawyers to disclose or certify their use of AI in filings, with sanctions for uncorrected errors.
Builds, configures and audits the AI drafting and workflow tools a firm actually uses, translating between legal practice and software development — a hybrid role barely staffed a decade ago.
Manages the technology-assisted review workflow for large litigation, setting the search parameters and quality checks that keep an AI-assisted document review defensible in court.
Advises companies on complying with emerging AI-specific regulation, such as the EU's AI Act, a specialism that barely existed before the early 2020s and is growing fast alongside AI adoption itself.
Runs a corporate legal department's budget, outside-counsel relationships and technology stack — a role built around exactly the efficiency and cost-control pressures AI and alternative fee arrangements have intensified.
Demand for lawyers tracks economic activity, regulation and litigation volume more than it tracks automation risk alone — new fields like data privacy, AI governance and climate-related disclosure are creating legal work faster than routine drafting work is disappearing.
The deeper tension is between AI's efficiency gains and a business model built on billing time: a firm that lets AI cut a task from ten hours to one has to either bill for one hour or find a different way to charge for the judgment that still took years to build — a structural problem the profession has not yet solved.
From ancient royal envoys and Venice's resident ambassadors to the Congress of Vienna and the UN, still negotiating peace one careful sentence at a time.
AI-resistant 78 🚓The sworn officer who patrols, investigates and keeps the peace, from Vidocq's undercover detectives and Peel's 1829 constables to today's contested public trust.
AI-resistant 66 📋Government's permanent, exam-selected staff who administer law and deliver public services — the machinery that keeps running as elected officials come and go.
AI-resistant 42