"The greatest danger in times of turbulence is not the turbulence. It is to act with yesterday's logic." — Peter Drucker

A partner once called me about a new matter that was, on paper, not especially large.

That was the misleading part. In law firms, some matters are small only in the way a lit match is small. This one had the potential to become something much bigger: a chance to prove the firm could be a credible alternative to the client's usual counsel, to enter the relationship through a side door, behave impeccably, and perhaps be invited back through the front. Perform well, win more work. Misprice it, overstaff it, or make the client feel as though they had taken a chance on the wrong firm, and the bridge could burn before anyone had finished admiring the architecture.

There is a magic number we chase in legal pricing. Too high, and you make the client wonder whether you understand the assignment. Too low, and you reveal either insecurity or a worrying lack of discipline, neither of which inspires confidence in people who are about to hand you a sensitive piece of work. The right number should be profitable for the firm, defensible against the scope, comfortable enough for the client, and credible in the market. Occasionally, that number exists on the same planet where attorneys enter their time daily and Outlook finds your email on the first search.

Not always the planet we operate on, but we try.

I did what pricing people do in these moments. I ran analysis designed to show me where the work would be concentrated and translate that time into dollar amounts. The number that came back was not just higher than the partner wanted. It was almost twice as high as the number he thought we could show the client without being politely escorted out of consideration.

Did I suddenly lose my mathematical faculties? I went back through the analysis.

Then I went back through it again.

The model was not obviously broken. The comparable matters were not wildly off point. The team assumptions were not extravagant. I had accounted for the usual suspects: scope, structure, jurisdiction, governing law, regulatory friction, the quiet little variables that seem harmless until they multiply in a spreadsheet and begin demanding their own parking space. Still, you do not walk into a conversation with an experienced partner and suggest his commercial instinct may be wrong unless you are very sure the data is right.

I was not very sure.

So I called someone closer to the machinery of the work and asked the question as plainly as I could: how could the experienced number and the historical number be this far apart?

The answer was AI.

Not "AI doubled the productivity of this deal team." Nothing so tidy. AI had compressed certain parts of the workflow enough that the partner's instinct, half calibration and half assumption, was operating on a different baseline than my historical dataset. Once again, I had kicked a field goal only to see the posts had moved.

The legal pricing industry has a consensus response to this problem: collect more data. Build new baselines. Give it a year, run your AI-assisted matters, see what the new normal looks like, reprice from there. This is reasonable advice. It is also cold comfort to firms being asked to price AI-era work on pre-AI data infrastructure right now.

The gap between when AI changed the work and when you have enough new data to price the changed work is not a rounding error. It is a pricing vacuum that gets filled with guesswork that looks like methodology if you do not press too hard on it.

Historical data is not obsolete. It is differently reliable, and knowing which parts remain trustworthy is the actual skill.

The things AI has not compressed are the objectively more difficult parts: regulatory exposure, cross-border jurisdictional complexity, the novelty of the legal issue, the stakeholder count on a seventeen-party transaction. Complexity does not change shape just because the tools got faster. It is a more durable proxy for what you are actually delivering than hours ever was, because complexity correlates with judgment rather than effort. What has changed is the relationship between complexity and time. The time to do the research, draft the initial memo, synthesize the diligence: that is compressing unevenly across matter types and practice groups. The analytical work is still there. The clock on certain parts of it is not.

Complexity metrics are still valid, but straight hour counts for research-heavy matters may need to be treated as ceiling, not baseline. Outcome-based value assessments seem increasingly essential.

The more interesting question to me is what legal pricing looks like when the concept of a stable historical baseline becomes systematically unreliable across the board. Not "how do we build better benchmarks," but "what if the benchmark model itself is the thing that needs to evolve."

I do not have a clean answer to that, and it doesn't seem like anyone else does either. The closest approximation will likely be "it depends" (a law firm favorite).

The teams that work cross functionally not only in sharing data, but visualizing it, do the best work. I truly believe the firms that figure out this transition together, with pricing and LPM in actual conversation, will price the AI era better than the ones that figure it out separately.