AI is already in your firm. The question is whether anyone's operating it.

Every firm owner I talk to is asking some version of the same question about AI.

Are we falling behind?

It’s the wrong question. AI is already in your firm. Someone on staff pasted a client document into ChatGPT this week to save themselves an hour. They didn’t ask, and you don’t have a record of it. That isn’t a hypothetical. It’s the default state of every firm that hasn’t decided otherwise.

So the real question isn’t whether to adopt AI. It’s whether anyone is operating it. The risk was never that your firm misses AI. The risk is that AI enters your firm without an operating model.

Two ways “we’ll figure it out later” goes wrong

It’s already running, and nobody owns it. This is shadow AI, and it’s the common case. Smart people doing good work, reaching for the fastest tool, routing privileged material through systems no one approved and no one can audit. Nothing looks broken. There’s just no answer to the question a client, a regulator, or opposing counsel might eventually ask: what went through what, and who decided that was fine. By the time the question gets asked, the answer is “we didn’t know.”

The pilot that dies after the demo. The other failure looks more responsible and ends the same way. A vendor runs a slick demo, the firm green-lights a project, something gets switched on. Then the demo energy fades and there’s no one whose job it is to run the thing, watch what it does, and answer for the output. An AI with no owner after launch isn’t adoption. It’s a liability with a login.

Both come from the same mistake: treating AI as a tool you install instead of a system you operate.

The operating model is the asset, not the AI

The model itself is a commodity. Everyone has the same ones. What separates a firm that gets leverage from AI from one that gets a near miss isn’t the model. It’s the managed system around it.

That means agents with job descriptions: what this one does, and what it doesn’t. Permissions, so it can only touch what the person on whose behalf it’s working could touch. Review gates on anything that can’t be undone. A record of what it did, in writing, that you could hand to a client. And a person who owns it after launch the way you’d own any hire you trusted with sensitive work.

None of that is exotic. It’s how you already run the parts of the firm that matter. AI doesn’t get an exception just because it’s new.

The one rule

If you hold the firm to one thing as AI comes in, hold it to this:

AI prepares the work. A person decides anything that can’t be undone.

A managing partner can govern the whole firm with that sentence. The assistant can draft, find, summarize, and propose all day. The moment something becomes a decision with consequences, a sent filing, a moved file, a client commitment, a human signs off. Not because the AI can’t be useful, but because accountability has to land on a person. That’s not a brake on speed. It’s the thing that lets you move fast without betting the firm on a model behaving.

What good looks like

• A written answer to “what AI is running here, what can it touch, and on whose behalf,” before anything new gets switched on.
• Human review on every action that can’t be reversed.
• A record you could show a client or a regulator without flinching.
• Someone who owns the system after launch, not just the demo.
• The whole thing scoped to one workflow first, proven, then widened.

That last point is the one most firms get backwards. You don’t need an AI strategy for the whole firm. You need one workflow operated well.

Start with one workflow

Pick the workflow that costs your best people the most time. Map how the work actually moves, decide where a human has to stay in the loop, and prove AI can carry the rest with review and a record you trust. Then widen. One workflow, one AI Team, one measurable outcome.

That’s the work we do in a workflow diagnosis. No tools installed, no data moved. We map how the work moves through your firm, where the lines have to hold, and where AI would actually earn its place. You walk out with the map whether or not you ever hire us.

Book a workflow diagnosis →