AI for Law Firms: What to Automate on Your Website, and What Not To

Your firm closes. Accidents, arrests, and eviction notices do not.

The after-hours inquiry goes unanswered for ordinary reasons. The receptionist went home. The answering service takes a message but cannot book anything. The form lands in a shared inbox that nobody owns until morning.

None of this is negligence. It is a staffing model built for business hours, applied to a problem that keeps no hours at all.

Meanwhile the person on the other end is scared and shopping. They are not comparing verdicts. They are looking for the first firm that seems to be listening. Our piece on speed to lead makes the general case. For a law firm the stakes are higher, because the person often has a deadline they do not know about yet.

Start with a narrow job. An assistant on your site is a front desk that never sleeps, reading from a script you approved. The job has five parts.

• Answer from approved firm information: practice areas, office hours, locations, parking, languages spoken, and the fee policy exactly as you publish it.
• Tell people what to bring to a consultation: the police report number, the lease, the letter from the court, the insurance card.
• Collect the minimum: name, phone, email, case type, county, and the names of the other parties so your team can run a conflict check.
• Book a consultation on a real attorney's calendar, or a call with your intake team.
• Route urgent matters to a person now. Someone in custody. A hearing tomorrow. A deadline the visitor mentions. Those go to a phone that rings, not a queue.

That is the whole job. It sounds small. It is the difference between a lead that waits nine hours and a consult that is booked before the visitor puts the phone down.

The version of this for any business is in our piece on AI customer service on your website. A law firm needs everything in that piece, plus a stricter set of lines.

Here is the tension. The better an assistant is at sounding helpful, the more likely it is to cross a line that only a lawyer may cross.

• Give legal advice. "You probably have a case" is advice. So is "don't talk to the insurer."
• Imply an attorney-client relationship. No "we've got you," no "your case is in good hands," no "welcome to the firm."
• Promise or predict outcomes. No settlement ranges. No "cases like yours usually win."
• Collect the detailed confidential story before a conflict check. The narrative, the documents, the medical history all wait for a person.
• Pretend to be a person. No paralegal named Sarah who is really a model.

The Florida Bar put this plainly in Florida Bar Ethics Opinion 24-1 (opens in a new tab), approved in January 2024. It tells lawyers to be wary of "an overly welcoming generative AI chatbot that may provide legal advice." It also flags chatbots that "fail to immediately identify itself as a chatbot" or "fail to include clear and reasonably understandable disclaimers limiting the lawyer's obligations."

Overly welcoming. Read that twice. Warmth is not the risk. Warmth that sounds like a lawyer is.

California went further in its 2026 update to The State Bar of California's practical guidance on generative AI (opens in a new tab). It names "autonomously facilitating client intake" as an example of agentic AI, and then draws the line.

“Lawyers must not deploy agentic systems in a manner that allows the system to make substantive legal determinations, communicate legal advice, prepare and file pleadings, or otherwise act in a representative capacity without meaningful lawyer supervision and review.”

Most of the decisions are not hard once you write them down. Here is the list we build to.

Disclosure comes first. In the first message, every time.

Florida's opinion says a lawyer "must inform prospective clients that they are communicating with an AI program." It also says the lawyer is responsible if the chatbot gives prospective clients misleading information or communicates in a way that is "inappropriately intrusive or coercive." That last phrase is worth taping to the monitor of whoever writes your follow-up texts.

Your assistant is a communication about your services. ABA Model Rule 7.1 (opens in a new tab) says a lawyer shall not make a false or misleading communication about the lawyer or the lawyer's services. An assistant that improvises a fee, a result, or a credential is making exactly that kind of communication, in your name.

The disclaimer matters for a second reason. Under ABA Model Rule 1.18 (opens in a new tab), a person who consults with a lawyer about possibly forming a client-lawyer relationship is a prospective client, and the firm owes that person duties even if it never takes the case. The comment to the rule says a consultation is likely when a lawyer specifically requests or invites the submission of information without adequate warnings that limit the lawyer's obligations. A chat window that asks "tell me what happened" is an invitation.

ABA Model Rule 1.6 (opens in a new tab) asks you to make reasonable efforts to prevent the inadvertent or unauthorized disclosure of, or unauthorized access to, information relating to the representation of a client. Rule 1.18 extends a version of that protection to prospective clients. So the chat transcript is not marketing data. Treat it like the first page of a file.

ABA Formal Opinion 512 (opens in a new tab), issued July 29, 2024, is the ABA's first formal ethics opinion on generative AI. It walks through competence, confidentiality, communication, supervision, and fees. One warning applies directly to your website: self-learning tools can expose information relating to a representation to people outside the firm who use the same tool.

California adds a line every vendor should hear: "Reasonable efforts require more than reliance on generalized marketing assurances." Read the terms of use. Read the privacy policy. Ask the questions.

Illinois firms have their own marker. The Illinois Supreme Court Policy on Artificial Intelligence (opens in a new tab), effective January 1, 2025, says the Rules of Professional Conduct "apply fully to the use of AI technologies" and that AI applications "must not compromise sensitive information, such as confidential communications."

What to ask any law firm intake software vendor

• Where are transcripts stored, and who can read them?
• Are our conversations used to train a model that other customers share? Put the answer in the contract.
• Can we limit the assistant to our approved answers and our calendar, with no access to case files or email?
• Can we set a retention period for transcripts from people who never became clients?
• Can the intake flow stop at case type and conflict names, so the full story waits for a person?

ABA Model Rule 5.3 (opens in a new tab) covers a lawyer's responsibilities for nonlawyer assistance. Opinion 512 lists the duty to supervise employees and agents among the duties generative AI touches. California says managerial lawyers should set clear policies on AI use and train both lawyer and nonlawyer staff.

Competence is part of it too. Since 2012, comment 8 to ABA Model Rule 1.1 (opens in a new tab) has told lawyers to keep abreast of "the benefits and risks associated with relevant technology." The assistant on your homepage is relevant technology.

Supervision sounds abstract. In practice it is a short list.

• A named attorney owns the approved answers and signs off on every change.
• Someone reads a sample of transcripts every week.
• Every handoff has an owner and a response time, after hours included.
• Before launch, test it like opposing counsel. Ask for advice. Ask for odds. Ask if it is a lawyer. Paste in a long confidential story. Fix what breaks.
• Test again when the vendor changes the model. California's guidance calls for periodic reassessment as systems change.

Back to the 11:40 pm form.

This time the visitor is told, in the first line, that they are talking to an automated assistant. They learn you handle car accidents in Cook County, that the first consultation is free if that is your policy, and to bring the police report. They give a name, a number, and the other driver's name. They pick 10:00 am. Your intake lead gets a text. A lawyer clears conflicts before anyone hears the story.

Nothing in that exchange is legal advice. All of it is service.

That is how we approach law firm website design at Golden Scope. A Smart Site with an assistant that answers only from approved firm information, intake that stops at what a conflict check needs, an admin portal where every transcript and handoff lives, and a person on the other end when it matters. If you want to see it for your practice, start there.