AI and the legal function: what changes (and what won't)

A general counsel who announces that “AI is going to transform legal work” is saying something both true and useless. True, because standard contract review, regulatory monitoring, and case-law research are already shifting massively toward AI tools. Useless, because that statement says nothing about what actually needs to change inside your own legal department, with your own contract volumes, your own litigation exposure, and your own risk appetite. Here’s the distinction that matters: what shifts fast, what holds, and why the line isn’t where most people assume.

The legal role isn’t a binary either

Like most support functions, the in-house counsel role is a bundle of highly heterogeneous tasks. A contracts lawyer drafts standard clauses, negotiates sensitive points, monitors regulatory change, advises the business on a specific risk, and — occasionally — handles litigation. Each of these activities has a radically different automation profile, and treating them as a single block blocks any useful decision about how to organize the legal function.

The question isn’t “will AI replace my lawyers” but “what share of their week can shift to AI without degrading the quality of advice given to the business.” That question has a precise, measurable, organization-specific answer.

It depends directly on the profile of the legal team in question. A contracts team handling hundreds of standardized files a year doesn’t have the same automation potential as a litigation team following a dozen complex, high-stakes cases. Applying the same AI roadmap to both under-exploits the first and over-promises on the second.

What shifts fast

Certain legal tasks already have high, well-documented automation potential:

• Standard contract review and drafting — reviewing boilerplate clauses, flagging deviations from a template, and generating first drafts of simple contracts are largely automatable today, particularly for high-volume vendor agreements and NDAs.
• Legal research and regulatory monitoring — aggregating case law, tracking sector-specific regulatory changes, and summarizing lengthy texts are tasks where AI delivers substantial time savings, freeing up hours previously spent on manual, repetitive monitoring.
• Documentary compliance tracking — checking file completeness and reconciling regulatory obligations against existing documentation lends itself well to automation, provided the underlying document repositories are properly structured to begin with.
• Translation and terminology harmonization — across multilingual contracts or international subsidiaries, AI sharply accelerates work that used to be highly time-consuming, without removing the need for legal review on terms carrying significant contractual weight.

What holds

Other activities remain, for now and for a long time, largely out of automation’s reach:

• Negotiating sensitive clauses — weighing commercial position against legal risk in a high-stakes negotiation stays a contextual judgment call no model has mastered, because it requires reading the real balance of power between parties, not just the text of the contract.
• Litigation strategy — assessing leverage, reputational risk, and whether to settle or go to trial depends on factors AI cannot evaluate on its own, including the leadership team’s own risk tolerance at that particular moment.
• Relationships with regulators and authorities — managing a data-protection audit, a sector inquiry, or a regulatory dispute involves institutional diplomacy and accountability that remain irreducibly human, where how you respond often matters as much as what you say.
• High-exposure commitment decisions — any decision that significantly binds the company legally stays, and should stay, validated by a senior lawyer, however well an upstream tool has prepared the ground.

AI in legal doesn’t make the lawyer redundant. For the first time, it makes visible how much of their time was never actually legal work to begin with.

The legal function also becomes the AI-compliance owner

This shift comes with a new role for the legal department: guardian of AI-usage compliance inside the company itself. The EU AI Act imposes documentation obligations on high-risk uses starting August 2026 — December 2027 for entities covered by the Omnibus mechanism. These obligations aren’t limited to AI software vendors: they apply to any company deploying AI systems in high-stakes processes, including in HR or production.

Legal departments therefore find themselves both a user of AI in their own processes and the party accountable for AI-usage compliance elsewhere in the company. A department that hasn’t itself clarified where and how it uses AI in its own files will struggle to police that question for every other department.

How to measure it in your own organization

A role-by-role diagnostic applied to the legal function starts from job descriptions and real case volumes — contracts reviewed per month, share of time spent on monitoring, negotiation, litigation. Each identified task is checked against a reference library of roles already pre-scored to estimate its current automation potential, and scores are then consolidated by role and by practice area (contracts, litigation, compliance, IP).

The output doesn’t say “legal will lose X positions.” It says how many hours of contract review can be freed up this year, and toward which higher-value work — strategic advisory, managing emerging regulatory risk, steering AI compliance — that time can be redeployed.

Three horizons for the legal department

1. 0–6 months — automate document review. Standard contracts, regulatory monitoring, completeness checks: these shift first, under a lawyer’s supervision on edge cases.
2. 6–18 months — refocus roles on advisory work. Job descriptions evolve toward more strategic advice to the business and less repetitive document production.
3. Beyond 18 months — own the company’s AI compliance. Legal structures the documentation the AI Act requires across all of the organization’s high-risk uses, not just its own.

Legal doesn’t become a lighter function with AI. It becomes a more strategic one — provided you know precisely which tasks to hand over, and which to keep.