Title: Where AI Belongs in Your Law Firm, and Where It Doesn’t

A new client calls at 9 PM

A prospective client finds your firm’s website at 9 PM on a Tuesday. They fill out the contact form, describe their situation in three rambling paragraphs, and go to bed wondering if anyone will ever respond.

Your paralegal sees that form at 8:30 the next morning, buried under eleven other emails. By the time someone calls back, the prospective client has already talked to two other firms.

This is the gap AI can close. It is also the gap where firms get nervous, and they should ask hard questions before they let any tool near client intake.

What AI can safely do before a human sees it

Intake is the clearest use case. A well-built system can acknowledge a new inquiry within minutes, collect the basic facts (name, case type, rough timeline, contact preference), and route it to the right person. None of that requires legal judgment. It requires speed and consistency, which is exactly what software is good at and exhausted staff are not.

Follow-up works the same way. A client who signed an engagement letter three weeks ago and hasn’t heard anything is not thinking “my attorney is busy.” They are thinking “did they forget about me.” A scheduled check-in, even a short one, closes that gap without pulling an attorney off billable work to send a status email.

Document work sits in a different category, and firms should treat it that way. AI can draft a first pass, summarize a deposition transcript, or flag inconsistent dates across a filing. What it cannot do is decide the document is correct. Every draft needs a lawyer’s review before it goes anywhere near a client or a court.

The line is supervision, not capability

A lawyer remains responsible for work done on the firm’s behalf by people who are not lawyers, and that responsibility carries over when the assistant is software. If it touches client matters, a lawyer answers for it. Check your own state’s requirements with your bar before you rely on any of this.

That means every AI system in a law firm needs a few guardrails before it goes live:

A human reviews substantive output before it reaches a client or the record. Not a spot check once a month. Every time.

Know where client information goes. If a tool sends client information to a third-party model, you need to know exactly what happens to that data and whether your engagement letters and your duty to protect client information allow it.

Someone can explain what the tool did and why. If you can’t describe the process to a client who asks, that is a sign the process needs more structure before it earns more trust.

A scenario worth sitting with

Picture a mid-size firm that turns on an AI intake tool to catch after-hours inquiries. It works exactly as intended for two months. Then a prospective client mentions, almost in passing, that opposing counsel is someone at a specific firm. The AI logs it as a normal field entry. Nobody flags the conflict until the engagement letter is already signed.

The tool didn’t fail. The process around the tool failed. Intake automation without a conflict check built into the workflow is a liability, not a convenience. The fix is designing the handoff so a human reviews exactly the fields that carry risk, every single time, before anything moves forward.

Start with the lowest-risk task

If you are considering AI for your firm, start where the downside is smallest. After-hours intake acknowledgment and basic scheduling are good first steps. They are visible wins, they reduce the chance of losing a prospective client to a competitor who called back first, and they don’t touch legal judgment at all.

Document drafting and analysis can come next, once you have a review process that catches errors before they matter. The goal is a tool that does the repetitive part well, every time, so your attorneys spend their hours on the parts that actually need a law degree.

What to ask before you adopt anything

Before you bring any AI tool into your intake or document process, ask who reviews its output, where client data goes once it leaves your system, and whether you could explain the process to a bar investigator in plain language. If any of those answers is vague, the tool isn’t ready for your firm yet, no matter how good the demo looked.

In our view, the firms that get this right are the ones that can say exactly what a human checks, and when, whenever a client’s matter is involved.