Satya Nadella just described the biggest risk facing law firms.

He wasn't even talking about legal.

His argument goes like this. Every company now has two kinds of capital.

Human capital, the judgment and pattern recognition of its people. And token capital, the AI capability it builds and owns.

The firms that win won't be the ones that pick the best model.

They'll be the ones that build a learning loop on top of models, where those two forms of capital compound each other.

This hits different in legal.

Law firms are pure human capital businesses. The value proposition is judgment accumulated over decades.

Precedent knowledge. Negotiation instincts. Knowing which argument lands with which judge.

Right now, most firms are renting token capital.

They subscribe to point solutions, feed them matter data, and the learning accrues to the vendor.

The expertise flows out. Very little compounds back in.

Satya's warning is blunt: If a handful of models absorb everyone's expertise and commoditize it, entire industries get hollowed out.

He compares it to what outsourcing did to industrial economies. The GDP looked fine. The displacement was real.

For law firms, the defense is architectural.

Your learning loop has to live inside your walls.

Your data, your tenant, your evals, your accumulated judgment encoded in systems you own.

You should be able to swap out the underlying model without losing the "firm veteran" expertise built on top of it.

That's the test of sovereignty.

Can you change models without losing what makes you different?

Most firms can't answer yes today.

The ones that can in three years will be the ones that started building now.