The AI pitch aimed at lawyers right now is almost always the flashiest, dumbest version: an AI that drafts your briefs, argues your motions, basically practices law. It's the wrong starting point, and it happens to be the one with your bar card attached to the downside. A hallucinated citation isn't a bug ticket. It's a sanction.

I build custom AI for small businesses, and when a firm asks me where to start, I point at the least glamorous door in the building. Intake. The phone call, the web form, the "do we even take this case." That's where AI quietly earns its keep, and where being wrong costs you a follow-up, not your license.

The question for a small firm isn't "can AI practice law." It's "how many good clients walk because nobody got back to them in time." AI fixes the second one today, with no risk to the first.

Where the money actually leaks

Talk to any small firm honestly and the leak is the same: leads come in and sit. A potential client calls during a deposition and gets voicemail. A form comes in at 9 p.m. and nobody sees it until Thursday. By then they've called three other firms and signed with whoever picked up.

That's not a lawyering problem. It's a response-time problem, and it's pure lost revenue — good cases walking out the door because the front desk is one human who's also doing six other jobs. You don't need an AI lawyer to fix that. You need an AI receptionist that never sleeps.

What "intake AI" actually does

An intake system does the unglamorous, high-value work the firm can't staff around the clock:

  • Answers immediately, every time, day or night, in the firm's own voice.
  • Asks the qualifying questions — practice area, jurisdiction, basic facts, conflicts flags — and captures the answers cleanly.
  • Sorts the wheat from the chaff, so the attorney spends time on real cases instead of tire-kickers.
  • Books the consult or routes the live one to a human, with the whole conversation already written up.

Notice what it does not do: give legal advice, predict outcomes, or say a single thing that creates exposure. It's the front desk, not the lawyer. That line is the entire design.

Why this is the safe place to start

Here's the logic that makes intake the right first move and not just the easy one. The blast radius of a mistake is tiny. If the intake AI mishandles something, worst case a person follows up — the same recovery you already have when a human drops a ball. Compare that to an AI that touches work product, where "worst case" has a hearing.

It's the same discipline I run on every AI feature I ship: I'd rather verify a narrow thing works perfectly than trust a broad thing that might. Start where being wrong is cheap. Earn trust. Then widen.

The bigger pattern

Intake is the wedge, not the whole story. Once a firm sees AI reliably handle the front door, the same approach extends to the next safe, high-volume task — document intake, scheduling, status updates to clients who'd otherwise call to ask. Each one boring. Each one a real hour back.

And the principle underneath it is the one I bring to every business, legal or not: own the system, keep the client data in the building, build the asset instead of renting the hype. A firm's client relationships are its entire value. That data does not belong in someone else's chatbot.

If you run a firm and you're tired of good cases slipping out the door after hours, that's the place to start. Here's what that looks like built for a practice — and it starts at the front desk, not the bench.