AI doesn't replace you. It multiplies whoever learns to use it — and retires whoever doesn't

There is a question that has stopped making sense: will artificial intelligence replace the lawyer? The useful question is another one, and it is uncomfortable — how long before the colleague who learned to use it delivers in two days what takes you two weeks? This piece is about why studying AI is not learning one more tool, but acquiring a multiplier that applies to everything you already know how to do.

1. AI is not a tool for one field — it is a general-purpose technology

Legal software is for practising law. An ERP is for running a business. Every traditional tool is born glued to a single use. AI is not: it is a layer that attaches to any activity made of text, reasoning, analysis, decision and repetition — and legal work is made almost entirely of that.

The consequence is that studying AI does not add one more subject to your training. It gives you a multiplier that applies to everything you already know. Someone who studies tax law gets better at tax law. Someone who studies AI gets better at everything they do.

2. To be good at AI, you are forced to spell out what you know

Here is the mechanism almost nobody notices. An experienced lawyer works largely on accumulated intuition: they can tell “at a glance” whether a clause is risky, where the other side’s brief is weak, which argument that particular court tends to accept. That knowledge is tacit — it works, but it has never been put into words.

To use AI with quality, it has to be put into words. You have to say what you want, in what order, under which criteria, to what standard of quality, avoiding which risk. In other words: using AI well forces the lawyer to become a teacher of their own craft — and nobody comes out of that process the same. Whoever sets up an AI to review contracts discovers, along the way, what their contract review method actually is. Often for the first time in their career.

That is why the gain is double and inseparable: learning to use AI in legal work is, at the same time, a deepening of legal work itself. It is not a pleasant side effect. It is how the thing works.

3. Three layers of competence open at once

The technical layer — what used to belong to IT

Automations, integrations, organised knowledge bases, dashboards, small custom systems. Before, any of these required a budget, a vendor, a deadline and dependence: “IT will look into it”. Today whoever understands AI describes what they need and gets it. It is not that the lawyer becomes a programmer — it is that the language barrier between wanting and getting has fallen. That is autonomy, and autonomy is speed.

The core layer — practising law better

Broader research in less time, analysis of volumes no team would read, testing an argument against its counter-argument, simulating the opponent’s case, checking coherence, drafts produced from a method that is yours. The professional is not replaced: they are amplified. They are still the one who judges — only now they judge with more material on the table and the heavy lifting already done.

The management layer — what separated the large from the small

Deadline control, pricing, cash flow, indicators, client pipeline, quality standardisation, written procedures, marketing, follow-up. None of this is taught in law school, and almost none of it can a small firm afford to buy. With AI, it can build it.

4. The economic argument: the barrier was never talent

The distance between the large firm and the small one was never, primarily, about legal talent. There have always been excellent lawyers working alone. The barrier was installed capacity: a team to absorb volume, a second pair of eyes for quality control, administration, IT, marketing, management. High fixed cost, which only scale pays for — and scale only reaches those who already have structure. A closed circle, nearly impossible to break into for someone starting out.

AI breaks that circle because it performs an unprecedented conversion: it turns fixed structural cost into individual competence. What used to require hiring, training, supervising and paying every month now requires learning. And learning has a decreasing marginal cost: you study once and use it forever, on every case.

In practice, the small firm starts delivering at the standard that was the exclusive preserve of the large one — the same depth of research, the same consistency of documents, the same deadline control, the same client follow-up — without the payroll that sustained that standard. And because the cost is lower, the margin is higher. The small firm does not merely catch up with the large one: in profit per professional, it can overtake it.

5. Refusing AI is not neutrality — it is defeat by absence

This has to be said without hedging, because it is the uncomfortable and true part: on several dimensions of legal work, AI is already better than a human being. Not on a few — on several. Speed, obviously. But also breadth of reading, resistance to fatigue, consistency at three in the morning, the ability to compare five hundred contracts without skipping one, to cross-reference an argument against all available case law without forgetting what it read on the previous page. No professional, however brilliant, competes with that. No team competes with that.

Faced with a fact like that, refusing to use it is neither prudence nor loyalty to the craft. It is choosing to compete in the market without part of your capacity. Whoever refuses to use AI has already lost — not will lose, has already lost, because the comparison is no longer between them and the machine, it is between them and the colleague who adds the machine. And the one making that comparison is the client, every day, looking at deadline, price and quality.

6. But there is a caveat — and it is what makes studying indispensable

Better on several fronts does not mean better on all of them, and the distinction is the core of the value of being prepared. AI is unbeatable at sweeping, comparing, staying consistent and handling volume. It is fragile, and sometimes dangerous, at legal judgement, at weighing a precedent, at reading context and at responsibility — naive use produces well-known disasters, such as citing a precedent that does not exist.

That is precisely why study decides the outcome. If AI were infallible, learning would make no difference: you would just press the button. It is because it fails in specific, predictable ways that the trained professional holds an enormous advantage over the untrained one. Whoever studies extracts the gain and steers around the trap. Whoever does not study chooses between falling behind by not using it and getting burned by using it badly.

7. The sum: what AI will never have, and what happens when the two join

Here we reach the decisive point. You have things AI does not have and probably never will — not because of a passing technical limitation, but because they are attributes of a person, not of a machine. They are strictly personal:

• Responsibility — someone who answers for the mistake, with a name, personal assets and a bar registration.
• Trust built over time — the client who comes to you because they have known you for fifteen years, not because they compared prices.
• Reputation before the judge, the prosecutor, the lawyer on the other side.
• A feel for context — knowing that this case is not about money, it is about resentment; that here a settlement fits better than a ruling.
• The courage to decide and sign your name to it.
• The experience of having lived through it before, with those people, in that courthouse.
• The ability to look the client in the eye and say the truth they do not want to hear.

None of that can be generated. None of that can be downloaded. That is you.

And this is exactly where the reasoning closes. AI alone has a ceiling: it is fast and broad, but impersonal, it answers for nothing, it does not know your client, it has no history. You alone have another ceiling: irreplaceable in judgement, but limited in volume, speed and stamina. Added together, both ceilings fall.

Whoever accepts using it and becomes skilled does not end up “almost as good as the AI”. They end up above it — because they add to the power of the machine what the machine cannot reach, and add to their own experience what no human experience reaches alone. It is the only configuration that beats both parts in isolation. Those are the superpowers: not AI in place of the lawyer, but AI coupled to a lawyer who knows how to steer it.

8. The three destinies

The legal market is splitting into three groups, and the only difference between them is study:

• Those who refuse. They lose by absence, competing with half the capacity against someone who has twice as much.
• Those who use it without learning. They gain speed and lose reliability — and one invented precedent in a brief costs more than all the time saved.
• Those who learn to use it. They add. They deliver faster, with more depth, at a lower cost, running their own firm with the rigour that used to require a team — and on top of that they bring what no machine brings: presence, responsibility and trust.