The painter, the emperor and AI: what clients still pay for

The painter and the emperor

A lawyer advises a client during his consultation hours on a question of employment law. He had informed her about the costs beforehand. His invoice comes to 190 euros plus VAT for an initial consultation. The client does not pay, and the lawyer sues for his fee. In the proceedings, the client offers a settlement of 50 euros, which the lawyer does not accept. The Leverkusen Local Court finds entirely in his favour in its judgment of 27 May 2020 (case no. 27 C 135/19). What is remarkable is less the outcome than the reasoning. The court tells a story: 

A Chinese emperor comes to a mountain village on his travels. A painter lives there, and the emperor asks him to paint him a rooster. Then the emperor moves on, and thirty years pass. When he comes through the village again, he remembers the commission and asks the painter about his picture. The painter sits down, takes a sheet of paper and, with a few strokes of the brush, paints a beautiful rooster. The emperor asks what it costs. Three gold pieces, says the painter. The emperor finds that excessive for five minutes' work. The painter replies: the emperor has only seen the five minutes. He has not seen the thirty years in which the painter practised for this rooster.

What is paid for is the knowledge

The same applies to a lawyer's remuneration, the court writes in its judgment. The amount of the fee depends neither on the time the lawyer has spent nor on the success of his work. What is remunerated is the "drawing upon his knowledge". What matters is solely the fee value under the schedule of fees, and the lawyer had calculated that correctly. For non-lawyers: 190 euros plus VAT is the maximum amount a consumer pays for an initial consultation. So the lawyer was not even allowed to charge more for this conversation.

The client saw a short consultation. What she paid for were the years before it: the degree, the traineeship, the cases, the mistakes and the judgments someone had read so that he could answer her question in a few minutes. That the answer came quickly is no argument against the price, because that very speed is the result of practice. Whether the client was happy with the outcome is equally irrelevant. She got what she ordered: access to knowledge she did not have herself.

The parable has only one flaw, and the court could not have known about it in 2020. There is now a painter who paints any rooster in seconds (I mean AI) and did not need thirty years to do so. My thesis: in future, clients will no longer pay for knowledge, but for responsibility.

What AI takes over from the painter

The court's reasoning incidentally explains how the legal market has always worked. Legal knowledge was scarce. It sat in people's heads, in commentaries and in databases to which clients had no access or with which they could do nothing. Anyone who wanted to know the notice period, the limitation period or the prospects of success had to ask a lawyer. Scarcity determined the price, and the years of practice were the reason for the scarcity. That very scarcity is disappearing. Generative AI reads, drafts and answers questions that used to require a certified specialist. The next stage is AI agents that handle entire work steps, from file analysis to drafting. The painter's thirty years of practice now sit in an AI model that anyone can call up on their phone.

Clients have understood this. According to a representative Bitkom survey from January 2025, 26 per cent of Germans can imagine asking an AI rather than a lawyer about a legal question. 14 per cent would do so only for simple questions, 12 per cent in any case. And 21 per cent credit an AI with more expertise than a lawyer. Anyone who has researched the basics with AI beforehand no longer needs information during the consultation. They are looking for an assessment. Nobody will therefore pay 190 euros plus VAT any more to ask about a notice period. The AI's answer costs thirty seconds and is usually right.

What AI cannot paint

At this point it is worth taking the counter-argument seriously. If AI knows more than any lawyer, why should anyone still pay for legal knowledge? The answer lies in the word "usually". In 2024, researchers at Stanford University tested the AI research tools of the major legal database providers. The tools delivered incorrect or invented information in 17 to 33 per cent of cases. The authors conclude from this that lawyers have to check the results. But only someone who has the knowledge themselves can check. The thirty years of practice therefore remain necessary (for now). They are simply no longer a product. They are the precondition for checking the product.

Besides, even a flawless AI would have a problem: it is liable for nothing. Anyone filing a claim needs someone to take responsibility for the deadlines. Anyone signing a contract needs someone to answer for a clause. On top of that comes judgement in the individual case. Suppose the AI delivers five suggestions for how to proceed procedurally. Which of them suits this client, this opponent and this court is a question of experience. Whether to sue at all or rather to negotiate is answered by no commentary and no AI model. For that, someone has to size up the opponent, know the state of the evidence and know what the client really needs in the end. And in the Bitkom survey, 54 per cent say an AI does not sufficiently understand the situation of those affected. The person across the table remains part of the service, all the more so when it is about your own job or your own business.

Why the hourly rate no longer holds up

What follows from this for remuneration? Billing by the hour loses its basis twice over. First, the work takes less time when the AI delivers the draft and the lawyer checks it. Second, what the client paid for with every billed hour is no longer scarce. The hourly rate was never the price for sixty minutes of work. It was the price for sixty minutes of access to thirty years of practice, which is why it rose with experience and reputation. If that access costs nothing any more, the client is only paying for time with the hourly rate. Anyone who becomes three times as fast with AI and carries on billing by the hour punishes themselves for their own efficiency.

My position is clear: clients buy results, not working time. Billing by effort is therefore replaced by fixed prices for clearly described services, digital packages for recurring standard work and ongoing advice at a fixed monthly price. Legally, this works via a fee agreement; in court proceedings, however, a fixed price may not fall below the statutory fees. The law firm's knowledge thus becomes a product: it is prepared digitally once, can be used any number of times and is checked and approved by the lawyer in every individual case.

The price is then no longer based on the time spent, but on how much is at stake for the client and on the fact that a lawyer takes responsibility for the result. Anyone quoting a fixed price benefits from every minute the AI saves them. But they also take on the risk of having misjudged the effort, and they are liable for the result. That is precisely what an AI cannot do. It delivers drafts, but it answers for no mistake. That is why in future, too, the client will pay the lawyer and not the software.

Incidentally, the statutory law on fees does not bill by the hour anyway. The Lawyers' Remuneration Act (RVG) has always assessed fees by the value in dispute, that is, by what the matter is actually about. The oldest fee model in the profession is thus, of all things, the most modern.

The painter must keep practising

That leaves the question of where the thirty years of practice will come from in future. The painter needed five minutes for his rooster because he had painted a thousand roosters before. The lawyer can check the AI draft because he has written a hundred drafts himself. If the AI writes the drafts, the next generation lacks exactly that practice. Anyone training newcomers today therefore has to come up with something, perhaps following the painter's example: first a thousand roosters by hand, then hand the brush to the AI. That is how I picture the law firm of tomorrow: knowledge is the raw material, AI is the tool, and what is sold is the responsibility for making sure the right rooster ends up on the page.