Short version: Florida lawyers can use generative AI, including chatbots on the firm website, but The Florida Bar's Ethics Opinion 24-1 expects the bot to say it's an AI and not a lawyer, stick to collecting facts, never give legal advice, and protect anything confidential a visitor types in. If you're automating intake, those aren't nice-to-haves. They're the spec.
Most firm owners who ask us about AI want the same thing: a smarter front door. Someone lands on the site at 9:40 p.m. after a car accident or a landlord dispute, and instead of a contact form that sits until morning, something actually responds, asks the right questions and gets the matter into the case management system with a conflict check queued up.
That's a very buildable thing in 2026. But if your firm is in Florida, there's a document you should read before anyone writes a single prompt: Florida Bar Ethics Opinion 24-1. The Bar's Board of Governors approved it on January 19, 2024, and it speaks directly to the chatbot-on-your-website scenario.
This post isn't legal advice (we're an automation shop, not your ethics counsel). It's a builder's read of what the opinion says, and how we translate it into actual requirements when we scope law firm intake automation.
What Opinion 24-1 actually says, in plain English
The headline is permission with conditions. Lawyers may use generative AI, but they have to protect client confidentiality, provide accurate and competent work, avoid improper billing, and follow the lawyer advertising rules. A few lines matter a lot for anyone building intake:
- The bot has to identify itself. The opinion says a generative AI chatbot that talks to clients or third parties has to include a disclaimer that it's an AI program, not a lawyer or an employee of the firm.
- It can gather facts, not give advice. The opinion leans on older guidance (Florida Ethics Opinion 88-6) about nonlawyers doing initial interviews: identify your nonlawyer status, limit questions to getting factual information, and send any legal question back to the lawyer.
- "Overly welcoming" is a risk. The opinion specifically warns about a chatbot that might give legal advice, fail to identify itself right away, or skip clear disclaimers. The concern is that a prospective-client relationship, or even a lawyer-client relationship, gets created without the lawyer knowing.
- Confidentiality includes the vendor. Lawyers are expected to understand a tool's data retention, data sharing and "self-learning" behavior. The opinion recommends getting the client's informed consent before using a third-party generative AI program if confidential information would be disclosed.
- Screen for people who already have a lawyer. The opinion suggests intake chatbots include screening questions that limit what the bot says if the person is already represented.
The American Bar Association followed with its own guidance, Formal Opinion 512, on July 29, 2024. It covers the same ground at a national level (competence, confidentiality, communication, supervision, reasonable fees) and adds a useful point on consent: for self-learning tools, a generic boilerplate line in the engagement letter isn't enough.
Why this is really a systems problem
Here's the thing most chatbot vendors skip. None of these requirements are about the model being "smart enough." They're about the system around the model: what it's allowed to say, where the data goes, what gets logged and who reviews it.
A general-purpose chat widget dropped onto a firm website is optimized to be helpful. Helpful is exactly what gets you in trouble here. A visitor asks, "Do I have a case?" and a helpful bot wants to answer. The opinion basically says: it can't. The answer has to be some version of "That's a question for an attorney, and here's how to get one on the phone."
So when we build intake for a firm, the bot is the smallest part of the job. Most of the work is the guardrails, the routing and the handoff.
The build spec we use for Florida firm intake
This is how we turn the opinion into requirements. Treat it as a checklist to review with whoever handles ethics and compliance at your firm.
1. Disclose first, every time
The first message says plainly that the visitor is talking to an automated assistant, not a lawyer or firm employee, and that nothing in the chat is legal advice or creates an attorney-client relationship. It's not tucked in the footer. It's in the conversation, before the first question. We also log that the disclosure was shown, with a timestamp, for every session.
2. Facts only, by design
The intake flow is a structured interview, not an open chat. Name, contact info, the other parties involved (for conflicts), dates, location, what happened, and what the person is hoping for. When the visitor asks a legal question, the assistant has a fixed response that routes them to a consult. The way we spec it, the build gets tested against a list of "tempting" questions (Is there a deadline? Should I sign this? What's my case worth?) and it doesn't ship until every one of them gets routed instead of answered.
3. Screen early: already represented, emergencies, out of scope
Two questions go near the top: are you currently represented by another lawyer on this matter, and is this an emergency? If someone's already represented, the assistant limits the conversation and hands off. If it's an emergency, it shows the right number to call. Matters outside your practice areas get a polite note and, if you want, a referral list you control.
4. Know where the data lives
The opinion asks lawyers to understand a provider's data retention, sharing and self-learning policies. In practice that means: model providers configured so prompts aren't used for training, data retention set to the shortest workable window, and intake data written to your systems (Clio, MyCase, your CRM), not left sitting in a chatbot vendor's dashboard. We document the full data path in writing so your firm can review it, and so you can answer a client who asks where their information went.
5. A lawyer reviews before anything goes out
The assistant can draft a summary of the intake and a suggested next step for the attorney. It doesn't send engagement letters, quote fees, or tell anyone they've been accepted as a client. Opinion 24-1 is clear that lawyers remain responsible for work product whether a paralegal or an AI drafted it, so a human sign-off is part of the workflow, not an afterthought.
6. Keep the advertising rules in mind
Your bot is part of your marketing. The opinion notes that the advertising rules still apply, including the ban on characterizing a lawyer's skills or record unless it's objectively verifiable. So the assistant doesn't say "we're the best injury firm in Palm Beach" or "we win most of our cases." It describes practice areas and next steps, and that's it.
What the automation actually saves
Done right, a compliant intake system isn't a toned-down chatbot. It's a better pipeline. The parts that eat staff time are the parts after the conversation: retyping details into the case management system, kicking off conflict checks, chasing missing info, sending the scheduling link, and logging who followed up and when.
That's where automation earns its keep. A clean intake record lands in your system, the conflict check request is queued for a human to clear, missing items get a polite follow-up, and the attorney sees a one-screen summary instead of a wall of chat transcript. None of that requires the bot to say anything a lawyer should be saying.
If you want the bigger picture of what we automate for firms (document organization, client communication tracking, matter creation), see our legal automation overview or the broader automation hub.
Questions to ask any intake vendor
Whether you build with us or someone else, these questions separate a compliant setup from a liability:
- Where is the disclosure shown, and can you prove it was shown for a given session?
- What happens, word for word, when a visitor asks a legal question?
- Is any visitor input used to train a model, and where is it stored? For how long?
- Can we turn the bot off for represented parties or certain matter types?
- Who reviews an intake before a response with substance goes back to the person?
- Can we export every transcript and log if we leave?
If a vendor gets vague on the third question, that's your answer.
Frequently asked questions
Is Florida Bar Ethics Opinion 24-1 binding?
No. The opinion itself says advisory ethics opinions aren't binding. It's still the clearest statement of how The Florida Bar views generative AI use, which is why we build to it.
Do we need client consent to use AI in intake?
The opinion recommends getting the affected client's informed consent before using a third-party generative AI program if it would involve disclosing confidential information, and notes consent isn't required if no confidential information goes to a third party. How that applies to your setup is a question for your ethics counsel. Our job is to make the data path clear enough that the question is easy to answer.
Can the chatbot schedule consultations?
Yes. Booking a consult is a logistics task, not legal advice. The assistant can offer available times and send a confirmation, with the attorney deciding what happens in the consult.
Does this only matter for big firms?
It matters more for small ones. A firm with a handful of attorneys usually doesn't have someone watching the website chat at night, so the guardrails have to be built in rather than supervised live.
Want to see what a compliant intake pipeline would look like on your current tools? Book a free fit call and bring one real intake you handled last month. You can also read more about how Zach builds AI systems for small businesses at zacharyvorsteg.com.
This article is general information about automation design. It is not legal or ethics advice. Talk to your own counsel, or The Florida Bar's ethics resources, about how these rules apply to your firm.