“The AI made a mistake.”

Imagine hearing that from your accountant after a bad return. You would not accept it, and you should not. You paid the accountant. Same with your lawyer, your banker, your financial advisor. The signature, the license and the fiduciary duty belong to a person, and that is exactly why the person costs what they cost.

One of my partners at IQ Source put it in three words on WhatsApp on October 30, a little after 6:30 in the morning. Judgment and consequences.

A minute earlier the thread had opened with a single Greek word. Phronesis. We were talking about the fear professional firms keep raising, that AI is coming for their jobs. That word is the best answer I have found, and it is older than the Roman Empire.

AI makes knowledge cheap. It makes judgment more valuable. A firm that sells knowledge will feel the price pressure. A firm that sells judgment has more room than it had before.

What Aristotle already sorted out

Aristotle separated three kinds of knowing. The Stanford Encyclopedia of Philosophy entry on his ethics goes deep on it. The short version I use with clients looks like this.

Layer	What it means	In a firm	Where AI is
Episteme	Knowing what is true in general	Tax code, case law, accounting standards	Retrieves and summarizes it in seconds
Techne	Knowing how to produce something	A contract, a return, a reconciliation	Drafts it; a person reviews
Phronesis	Knowing what to do for this person, now	Advice under uncertainty, with consequences	Stays with the professional

The first two rows are where junior hours used to go. They are also where clients are starting to ask why the invoice still shows hours a model could do in minutes.

The third row looks different up close. A model does not sit across from a widow deciding whether to sell the family business. It does not choose between two term sheets with the founder who has to sign one. It does not catch the client who says one thing about risk and does another.

The license is the product

Lawyers have carried this idea in their vocabulary for two thousand years without noticing. The Romans translated phronesis as prudentia. Jurisprudence is iuris prudentia, practical wisdom about law.

Courts are already drawing the accountability line in 2026. In United States v. Heppner, decided in the Southern District of New York in February, what a defendant typed into a consumer AI tool on his own was not privileged. I covered the ruling in why your AI chats can end up as evidence. The tool takes on no responsibility. Whoever decides how it gets used takes all of it.

So the competition in professional services runs between two kinds of professionals. One hands the codifiable work to the machine and spends the recovered hours deciding better and staying close to clients. The other keeps billing for work the client already knows a machine can do.

Where the argument breaks

There is a hole in this, and any managing partner will find it fast.

Aristotle also said young people can be excellent at geometry and still lack practical wisdom, because phronesis comes from experience with particular cases. For decades the experience came from the grunt work. A junior drafted a hundred contracts, a senior marked up forty of them, and somewhere along the way the junior started seeing what the senior saw.

If AI drafts the hundred contracts, that school closes. I wrote about this in who trains the next senior, after the McKinsey Berlin vote.

The fix is to train judgment deliberately, and adult learning research has said how for a long time. Malcolm Knowles found that adults learn on real problems, from their own experience, once they understand why it matters. In a firm that means real resolved cases, the junior’s recommendation written down before they see the senior’s, and a conversation about the gap. The material already exists. It is the firm’s past decisions with their reasons, if anyone kept them.

What we do about it in AI Maestro

AI Maestro starts with each person’s week, and the tool comes later.

Every professional sorts their own work into three piles. Knowing the rule, producing the work, deciding the case. They do the sorting, because they know which parts of their day take judgment and which are repetition. Then we build a prompt with them that carries two or three of their own past decisions as examples, so the output reasons like they do instead of like a textbook.

From there we build the firm’s map. Every task lands in one of four buckets: automate now, automate with review, assist only, or keep human. We spend the most time on the fourth, because the client relationship lives there.

Sometimes the recommendation for a task is to leave it alone. That gets written down too.

If you run a law firm, an accounting practice or an advisory shop, try this before you talk to anyone. Pull up last week’s calendar and tag every block with one word. Rule, procedure or judgment. If most of your billable hours landed in the first two, send us the breakdown and we’ll go through it with you. If most landed in the third, you are in good shape, and you are probably charging too little.