A small firm's problem is time, not talent. The phone rings during a deposition. The intake form sits until Monday. The client update that should have gone out Tuesday goes out Friday.
AI can take some of that off your plate: intake sorting, scheduling, client update drafts, internal first drafts. It can also get a lawyer in trouble if it's used carelessly. The rules aren't a mystery. In Florida, the Bar has already written them down.
This post walks through what the Bar said, a setup that fits inside it, and what to leave alone. We're not lawyers and this isn't legal advice. You and your ethics counsel decide what's right for your firm. For the wider view on adding AI to any small business, see our guide to AI implementation for small business.
What the words mean
- Generative AI: tools like ChatGPT, Claude or Gemini that write text from an instruction (a "prompt").
- Automation: a rule that runs without you. Consult booked, reminder goes out. Form submitted, a task is created.
- Self-learning AI: a tool that keeps what you give it and uses it to improve. This matters for confidentiality, as you'll see.
- Intake: the first contact with a potential client, by phone, form or website chat.
What the Florida Bar said
The Florida Bar's Ethics Opinion 24-1 is dated January 19, 2024, and it's advisory, not binding. Its summary says lawyers may use generative AI but must protect client confidentiality, provide accurate and competent services, avoid improper billing, and follow lawyer advertising rules. Here's what that means in practice. We read the opinion's text for this post. Read it yourself, and check the Bar's site for anything newer.
Confidentiality. The opinion says lawyers must protect client information by researching the program's policies on data retention, data sharing and self-learning. It warns that a self-learning tool raises the possibility that a client's information may be stored in it and revealed to someone else later. It recommends getting the affected client's informed consent before using a third-party generative AI program if confidential information would be disclosed. It also says concerns may be eased by an in-house AI, and that if nothing confidential goes to a third party, consent isn't required under Rule 4-1.6.
For outside tools, the opinion points back to the Bar's cloud computing opinion. The lawyer should make sure the provider has an enforceable obligation to keep information confidential and secure and will notify the lawyer of a breach, look at its reputation and security, and find out whether it keeps your data or claims rights to it.
Oversight. The opinion compares reliance on generative AI to reliance on inexperienced or overconfident nonlawyer assistants. A lawyer must review the output and verify the accuracy and sufficiency of all research done by AI. It also says a lawyer may not delegate to AI any act that could be the practice of law, such as negotiating claims or anything needing the lawyer's personal judgment. The opinion cites the federal case Mata v. Avianca, where a judge sanctioned two lawyers and their firm for citations an AI made up.
Fees. AI can't be a reason to inflate time. The opinion says a lawyer must not falsely inflate hours, and suggests flat or contingent arrangements so efficiency benefits both sides. If you plan to charge a client the actual cost of using AI, tell them, preferably in writing.
Advertising and chatbots. This is the one most relevant to a small firm's website. The opinion says a chatbot that communicates with prospective clients must tell them it's an AI program, not a lawyer or firm employee. It warns against an overly welcoming chatbot that gives legal advice, fails to identify itself right away, or lacks clear disclaimers, because it could create a prospective client or even a lawyer-client relationship without the lawyer knowing. It suggests screening questions for people already represented by another lawyer. And you can advertise your use of AI, but you can't claim your AI is better than other firms' unless the claim is objectively verifiable.
The opinion says its list is not exhaustive and the technology is changing. Treat it as a floor.
A setup that fits inside it
1. Start with a tool policy, one page
Before any tool: which ones the firm allows, what can go in them and what can't. Our shadow AI policy guide has a plain template, and is ChatGPT safe for business data? covers the basics. The Bar's three confidentiality questions give you the vetting checklist: does it retain data, share data, or learn from it? For anything touching client information, get the vendor's answers in writing, and ask your ethics counsel whether client consent is needed.
A simple rule works for most firms: no client names or facts in any tool you haven't vetted. Use a vetted tool, or work with generic facts.
2. Intake that doesn't create a client by accident
Intake is where AI helps most and where the Bar's warning bites hardest. A prospective client is someone who consults a lawyer about forming a relationship. The opinion points out that the line depends on the circumstances, including whether the firm invites information without clear warnings that limit its obligations.
If you use a chatbot or an automated after-hours reply:
- Say it's an AI, immediately. Not in the footer. In the first message.
- Say it's not a lawyer and gives no legal advice. Put the disclaimer up front and keep it clear.
- Collect facts only. Name, number, a short description, the type of matter, whether they already have a lawyer.
- Screen for represented persons, as the opinion suggests.
- Send everything to a person. A human reads each intake and decides whether to call.
- Don't answer legal questions. The reply says someone from the firm will contact them, and gives an emergency or deadline note you wrote and approved.
Write the approved text once: practice areas, hours, location, how consults work, what to bring. The tool answers only from it. Anything else gets "someone will reply on the next business day."
Our research rates after-hours answering as promising but thin on independent evidence, so test it on after-hours only. Customers also still want a person, and legal clients more than most. See AI receptionist vs answering service for the options. If a phone tool records calls, see AI and call recording disclosure laws.
3. Reminders and scheduling
Reminders carry no confidentiality risk if they say only the time and place. A published review of 26 studies, mostly in healthcare, found people who got appointment notifications were 23% more likely to show up (67% vs 54%), and more than one beat a single reminder. A law office isn't a clinic, so take it as direction. It's the reminder doing the work.
Set up a reminder 48 hours before a consult and another that morning, with a one-tap reschedule. Leave the matter type out of the text. Get consent for texting, and read our texting rules guide. It's general information and not legal advice.
4. Client updates, drafted by AI and sent by a person
A client who doesn't hear from you assumes the worst. A short status update every couple of weeks cuts the "any news?" calls.
Use an approved tool. Give it generic structure ("a two-paragraph update that says what happened, what's next and when") and have a lawyer or paralegal fill in the facts and read it before it goes. The opinion's rule is simple: you own the work product. Nothing goes to a client unreviewed.
5. Internal first drafts
Letters, checklists, engagement letter templates, meeting summaries of non-confidential material. An assistant is fine for a first pass on text with no client facts. A lawyer reviews and finishes it. Never send a draft out as is.
6. Legal research: verify everything
The opinion says to verify the accuracy and sufficiency of all research. In practice, treat an AI answer as a lead, not a source. Every case, quote and cite gets checked in a real legal database before it goes in a filing or a client letter. If you can't find it, it doesn't exist.
Reviews and your Google profile
Ask satisfied clients for reviews, with no incentive and no script. Google's policy lists as violations requiring staff to collect a set number of reviews and asking for specific content. The FTC's rule bans fake and AI-written reviews.
Replies are where firms get hurt. Don't confirm that someone was a client or discuss the matter, even to defend yourself. A short "thank you" is enough. An assistant can draft the generic reply, and a lawyer edits and posts it. Your ethics counsel can advise on how the Bar's advertising rules apply to reviews and testimonials. We haven't verified that and won't guess. See how to get more Google reviews the right way.
What to keep personal
- Legal advice and judgment calls. Including any answer to "do I have a case?"
- Negotiation and anything that commits the firm.
- Every filing and every cite. A lawyer reviews and checks them.
- Client facts. They stay out of unvetted tools.
- The first call with an upset or scared client. A person makes it.
What to skip
- A chatbot that gives legal information that sounds like advice. It's the opinion's biggest warning.
- "Our AI is better" claims on your website unless you can objectively verify them.
- Free AI accounts for client matters.
- Billing for time AI saved.
- Auto-posting review replies. A lawyer reads each one.
Do it in this order
- Week 1: Write the one page tool policy, and ask your ethics counsel what the opinion means for your practice areas.
- Week 2: Vet one tool against the Bar's three questions (retention, sharing, self-learning), in writing.
- Week 3: Write the approved intake text and add the AI disclosure, disclaimer and represented person question.
- Week 4: Turn on consult reminders and start the reviewed client update routine.
- Day 30: Count after-hours inquiries captured, consults kept and hours saved on drafts. Keep what moved the numbers.
To see how many hours a week the repeat tasks take, run the AI time savings calculator. For governance basics, our AI compliance guide covers the wider picture.
Frequently asked questions
Does the Florida Bar allow lawyers to use generative AI?
Yes, with conditions. Ethics Opinion 24-1, dated January 19, 2024, says lawyers may use it but must protect client confidentiality, provide competent services, avoid improper billing and follow advertising rules. The opinion is advisory, not binding.
Do I need client consent to use an AI tool?
The opinion recommends getting the affected client's informed consent before using a third-party generative AI program if confidential information would be disclosed. If no confidential information goes to a third party, consent isn't required under Rule 4-1.6. Ask your ethics counsel how that applies to your tool.
Can I put a chatbot on my firm's website?
The opinion allows it with limits. The chatbot must tell visitors immediately that it's an AI program and not a lawyer or firm employee, avoid legal advice, and carry clear disclaimers. The opinion also suggests screening questions for people already represented by another lawyer.
Can AI do my legal research?
It can help you find leads, but you must verify the accuracy and sufficiency of all research the AI does. The opinion cites a federal case where lawyers were sanctioned over false citations created by generative AI. Check every case and quote in a real database.
Can I bill clients for time AI saved?
No. The opinion says AI-driven efficiency must not turn into falsely inflated time, and it suggests flat or contingent fees so the savings benefit both sides. If you plan to charge the actual cost of using AI, tell the client, preferably in writing.
Where Di-Hy fits
Di-Hy Digital is an AI implementation firm in Ybor City, Tampa. We start with a free AI audit, build in about 30 days, train your staff, and offer monthly upkeep if you want it. We set up the tools and the tool policy. Your ethics counsel decides what the rules mean for your firm. To talk it through, get a free audit.


