What AI will never replace. Rethinking the lawyer’s profession from what endures

Everyone is worried about whether AI will replace lawyers. What if we asked the opposite question instead: what, in this profession, has stood the test of time — and why? A shift in perspective to refocus practice on what matters.

Everyone is stirred up about the same question: is AI going to replace lawyers? It is obviously the wrong question.

The real question is: what has withstood the centuries, and why? If we want to understand how artificial intelligence will transform the practice of law, we need to change focus. Rather than looking at the latest developments in GPT or a legaltech company’s product announcements, we need to look in the rear-view mirror — over the long term. AI, like any technology, will merely reveal what holds firm, and what was incidental.

The right tool for thinking about this is the Lindy effect: the idea that the longer something has lasted, the more likely it is to keep lasting. And in law, certain functions have been around for millennia. Why? Because they answer fundamental human needs: organising trust, settling disputes, taking a position in the face of uncertainty.

Before courts, there were village elders. Before articles of law, there were promises, debts, exchanges. The lawyer’s profession — this stance of a third party able to defend, to negotiate, to advise with responsibility — is far older than our codes. It was not born with legal scholarship. It was born with trade, with conflict, with life in society.

What AI is going to sweep away is not the lawyer’s profession. It is everything that has built up around it. All the tasks that exist only because we had nothing better: sorting, proofreading, copying, reformatting, producing on an assembly line. Those are not “Lindy”. They have only lasted a few decades, and their sole justification was technological: they took time. Today, they take too much.

Let us take a concrete example: contract review. A well-trained AI can now identify risky clauses, detect deviations from a reference baseline, and generate suggested rewordings. No more need to spend hours on it. That is not where the value lies. The question that remains is: should this clause be accepted, given the context? And there, we come back to what endures. Judgement. Experience. The balance of power. The client’s intention. What goes unsaid.

Another example: mass litigation. Semantic analysis tools can already draw out the main trends from hundreds of decisions, extract patterns in case law, and summarise the most frequent arguments. But they will not tell you whether this particular case, with this particular client, is worth pleading. Nor how to plead it, nor what tone to adopt, nor whom to talk to face to face.

The same goes for advisory work: you can ask an LLM to simulate a contract or to summarise a reform. But no model takes a risk in your place. None will say: “Here is what I recommend. And I take responsibility for it.” That sentence lies at the heart of the profession. And it cannot be replaced.

Some firms have already begun this refocusing. Far from grand announcements, they have used AI not to “do AI”, but to get back to the essentials.

One employment law firm, for instance, began by automating its most routine tasks: clause extraction, generation of standard arguments, assisted case-law research. The result: its associates gained several hours a week, not to “do something else”, but to do better — take the time for a conversation with a client, prepare a negotiation, adjust a strategy.

The firm did not try to “transform itself digitally”. It simply reinvested its time where it is most useful, where law is not a stock of knowledge but a lever for decision-making.

What matters today, in the age of the hegemony of AI agents, are the fundamentals. Listening, judgement, clarity, the ability to stand behind a recommendation with commitment. None of this is new. And none of this can be replaced.

As AI reshuffles the cards, it is not technologies that should fascinate us, but invariants. What remains when everything else changes.

The lawyer’s profession does not fit into a task description. It is a role and a function: owning a point of view, defending a position, making a decision amid uncertainty. That role is played by no tool. And it is the one clients come looking for — more and more.

The issue is to find and decide what we are going to deepen. Where does our value lie? Where are we irreplaceable?

The rest can — and must — be automated, and not to follow a fashion. Automate to make room. Do not let technological agitation distract you from what is essential. AI does not kill law. It forces you to ask yourself what you want to make of it. And what, truly, you want to be of service to.