The associate math.

There is a specific arithmetic underneath every leverage-model firm, and it is worth stating without the softening language that usually surrounds it. A firm bills a junior lawyer's time at a multiple of what it pays them. The spread between those two numbers, multiplied across every junior and every billable hour, is what funds the offices, the recruitment pipeline, and the partner draw. The junior is not simply labour. The junior is the margin.

Now list what AI automates first and best: document review, first-draft contracts, research memos, due-diligence summaries, standard correspondence. That is not a random list. It is close to a complete description of what first- to fourth-year lawyers spend their billable hours doing.

Put those two facts together and the conclusion is uncomfortable but unavoidable. In a firm that prices time, every hour AI removes from the junior stack is revenue with no mechanism to recapture it. The model prices time and the technology's central contribution is deleting time. These are not compatible propositions, and no amount of enthusiasm at the innovation committee reconciles them.

This is why the observable behaviour at the top of the market is so consistent across firms that have no other reason to behave identically. They adopt the tools and protect the model. A review tool here, a research copilot there, a drafting assistant for the associates. The pricing stays hourly. The intake still starts with a partner call. The client receives the same experience, delivered slightly faster by people using better software. That is AI-augmented, and it is a rational response to the arithmetic.

It is worth being clear that none of this is a moral failing. Asking a partnership to automate the layer that funds it is asking three hundred people to vote against their own income, and institutions do not do that. It is also not a prediction of collapse - incumbents with strong relationships and bet-the-company work survive structural shifts for a long time.

The narrower and more useful conclusion is this: the AI-native operating model will not be built at the top of the market, because the arithmetic there forbids it. It is being built by practices with no junior layer to protect, where the same automated hour is recovered capacity rather than deleted revenue. Identical technology, opposite economics, and the difference between them is not effort. It is structure.