AI can't handle legal intake. Wrong.

We hear it on almost every call.

A lawyer says it before we've finished the sentence.

They're not wrong to be skeptical. They're skeptical about the wrong year.

Here's what's actually going on.

The objection is stuck in 2021

Say "AI" to a law firm and a movie plays in their head.

A robot voice. Stilted, obvious, three beats behind the conversation. A phone tree from 2009. A confused client hanging up.

That version was real. It also no longer exists.

Voice AI has moved faster in the last twelve months than in the five years before it. The jump is hard to overstate. A year ago the systems paused awkwardly, missed interruptions, and gave themselves away in the first sentence. Today they answer in under three seconds, handle a caller talking over them, ask real follow-up questions, and hold a conversation that flows.

The lawyer's objection isn't wrong about the old technology. It's just judging a version that's already been replaced.

The bias arrives before the evidence does. Lead a demo with the word "AI" and the firm spends the whole call hunting for the 2021 tell. Lead with the outcome and the same firm leans in and asks how it works.

Same product. Different order. Opposite reaction.

What intake actually has to beat

Here's the part skeptics skip.

Most firms aren't choosing between AI and a great human receptionist.

They're choosing between AI and voicemail.

The phone rings after 6pm. Nobody answers. The caller — a real person with a real legal problem and a credit card — leaves a message, or doesn't. By morning they've hired the next firm on Google.

That client was never lost to bad AI. They were lost to no answer.

So the bar isn't "sound perfectly human."

The bar is "beat a voicemail box." That bar is low. The technology now clears it by a mile.

What "working" looks like

Strip the label off and judge intake on results.

• Every call answered in under three seconds.
• Every caller asked the right qualifying questions: case type, urgency, the details that tell a firm whether this is a client worth calling back.
• Every conflict checked before anything gets booked.
• Every consultation dropped straight onto the calendar.
• Every call summarised and sitting in the firm's inbox before the coffee's gone cold.

That's not a demo trick. That's a Tuesday.

Firms that run intake this way stop chasing missed calls. They start their mornings with booked consultations instead of a voicemail backlog.

Why the skeptics are half right

Bad AI intake does exist.

Generic bots that don't know a firm's practice areas. Systems that can't ask a follow-up. Tools built for pizza orders and repurposed for law.

That stuff deserves the skepticism.

The mistake is deciding the whole category is broken because the generic version is — and because the version you tried eighteen months ago was.

Intake built for law firms, trained on the firm's practice, its qualifying questions, its conflict rules, running on this year's voice models, is a different animal. The skeptic who tried a generic bot in 2024 is judging a category that has already moved past them.