Brandon Burke Picture a new client who emails your firm the night before intake. Attached is a twelve-page document with headings, a timeline, and a confident theory of liability. It reads like a brief. It also has the accident on the wrong date, and it leaves out the earlier claim the client mentioned on the phone. When your paralegal asks about it, the client admits they only skimmed it.
Richard Frankel, owner of Boss Frankel, wrote about exactly this in a recent LinkedIn post. More of his clients are sending long AI-generated material: questions, case summaries, personal statements, even medical opinions they want a doctor to sign. It reads well. It often misses the facts that matter, and the client often hasn’t read it closely. One line from an agreement he’s considering for new clients says it plainly: “I won’t respond to a ten-thousand word manifesto and don’t expect you to either.”
I agree with him, and I think the point cuts both ways. If you’re thinking about AI in law firm client communication, here’s where I’d draw the line. Software can report what already happened in a case. People should answer the questions that need judgment.
Unchecked AI writing costs trust in both directions
The problem in Frankel’s inbox isn’t just length. Nobody stands behind the words. The client didn’t write them, often didn’t check them, and can’t defend them when a detail turns out wrong. Your team ends up untangling a document instead of talking to the person.
Firms can make the same mistake in reverse. AI-written check-ins, chatbot replies, and generated case summaries sound attentive. They can also guess, smooth over a missing fact, or bury the desirable facts in a sea of words. A client who catches one wrong detail may start to doubt all the rest.
Frankel also wrote that AI has a future. He just doesn’t see it in replacing human communication, or the strategy a lawyer builds over years in a practice area. That’s the part I agree with most. A client with a real question deserves a real person’s time. That time is scarce, and this practice reinforces the client’s manifesto approach, so the firm must be active in protecting it.
Where AI does help clients
Frankel sees real value in AI for clients with disabilities, especially ones that affect cognition and concentration. It can help them understand their case and communicate with the firm. A client can use it to decode a letter from an insurer, sort a stack of bills by date, or turn scattered notes into a short list of questions.
Those uses help the client understand and organize. They don’t put new claims into the case under the client’s name or codify context for the inevitable next automated bot response.
One reply to Frankel’s post made a similar point about firms. The useful AI works behind the scenes, drafting intake summaries for staff to check and following up on missing documents. People still handle the client’s questions. I’d put the line in the same place.
Instrumented updates vs. AI-generated messages
A lot of routine client communication answers one question: what’s happening with my case? You don’t need a model to guess at that. Your case management system and staff already know this.
An instrumented update is a message your system sends because something real happened. The firm filed the complaint. The defense noticed a deposition. The medical records came in. The update says so in plain language, close to when it happened. This creates a human-drafted firm standard for ensuring consistency across a firm’s client experience.
The instrumented update is the one a client can trust, and the one your staff doesn’t have to walk back. If you’re deciding which case events should trigger a message, our guide to automated case status updates walks through it.
What to automate, and what a person should answer
Automate the facts. Tell the client when the case reaches a new stage, the court sets a hearing, or the firm gets a document they sent. Those updates can head off the “just checking in” call before the client makes it, and generally demystify the legal process for client consumption.
Then hand the rest to a person. These questions need someone who knows the case and can stand behind the answer:
- What a settlement offer means for the client, and whether to take it
- Why the case is taking longer than the client expected
- Whether a new symptom, bill, or job change affects the claim
- What the client should say, or not say, to an insurer
- Any question where the client sounds worried, confused, or ready to leave the firm
For more on how those conversations should go, see our client communication best practices.
A client AI policy worth borrowing
Frankel is considering a simple mutual agreement with new clients. Use AI to understand your case and organize your materials. Then restate anything meant for the case in your own words.
It’s a fair policy, and a few lines in your welcome materials can explain it at intake. Use AI to make sense of letters and sort your documents, if that helps. Before you send us a statement, a timeline, or a question, read it and rewrite it so it says what you mean. Short is fine. Everyone would rather have three accurate sentences than ten pages nobody can rely on.
It works better when the firm holds up its end. If you ask clients to stand behind their words, stand behind yours. Every update a client gets should come from a real event in the case or from a person on the team.
It also means fewer manifestos. A client who hears about each step as it happens has less silence to fill.
Where CasePulse fits
This is how we built CasePulse. Our client portal sits on top of your case management system and sends clients automated alerts when key actions happen in the case. Clients see a visual tracker of where their case stands, and their secure client portal shows real-time information from your CMS. Client messages stay in your CMS with the rest of the case. Our features page covers the rest.
We also offer AI Refine, an optional aid in our message composer. It improves the grammar and professionalism of a message a staff member already wrote. It doesn’t write the message, and it can’t send it. The staff member does both.
If you want routine updates to come from your case data, and your team’s time to go to real questions, see how CasePulse works with your case management system.