ChamberOne

AI chatbots on law firm websites: benefits, risks and rules

By the ChamberOne team 4 minute read

A visitor lands on your site at eleven at night, worried about a notice they received. They type a question into the chat box. What happens next is a decision your firm made, whether or not anyone at the firm remembers making it. A chatbot can take the visitor's name and the type of problem, and promise a call in the morning. It can also start explaining their rights, and that is where the trouble begins. This article is for marketers and principals who are weighing a chatbot, and it sets out what to decide first.

Key points

  • Decide the chatbot's job narrowly: information, booking and capturing enquiries, not discussing the visitor's legal problem.
  • Say plainly on screen that it is automated, gives no legal advice and does not create a client relationship.
  • Check where transcripts are stored, who can read them and whether they are used for model training.
  • Name a person who reads enquiries, runs the conflict check and replies within the time you promise.

What a chatbot is actually for

Most firms want one of three things. They want to answer simple questions about opening hours, fees and which areas of law they cover. They want to capture an enquiry outside office hours. Or they want to point visitors to the right page or the right person.

None of those needs the chatbot to discuss the visitor's legal problem. A tool that does the first three well is useful and low-risk. A tool that tries to be a junior lawyer on the website is a different product, with different risks, and it should be judged as one.

Where accidental advice comes from

People do not type neat questions. They describe their situation in detail and ask what they should do. A chatbot built on a general language model will often try to help, and its answer can sound confident and personal.

That matters for two reasons. The visitor may act on the answer, which can cause harm if it is wrong for their jurisdiction or facts. And a reply that looks like advice to a specific person can blur the line between marketing and a professional relationship. Whether a relationship has formed is a question for your regulator's guidance and local law, so check what your bar or law society says rather than assuming.

Confidentiality starts at the chat box

Visitors will type sensitive facts into a chat window because it feels private. It may not be. Before you launch, find out where chat transcripts are stored, who at the vendor can read them, whether they are used to train models, and how long they are kept.

Also think about who is on the other side of the chat. A person may be writing about a dispute with an existing client of the firm, or about a matter where the firm already acts for the opposing party. If that text goes into a general marketing inbox, you may have received confidential information from someone you cannot now act for.

Design rules that reduce the risk

  • Open with a plain statement that the tool is automated, that it cannot give legal advice and that no lawyer-client relationship starts from the chat.
  • Limit the scope in the instructions you give the tool: firm information, how to book, what documents to bring. Tell it to decline questions about the visitor's rights or what to do.
  • Ask for the minimum: name, contact details, the type of matter and how urgent it is. Ask visitors not to enter detailed facts or confidential material.
  • Show an obvious route to a human: a phone number, a form, or a call-back request.
  • Add a short warning for urgent situations, such as court deadlines, telling people to contact a lawyer or the court directly rather than wait for a reply.
Do
  • Say plainly that the tool is automated and gives no legal advice
  • Limit it to firm information, booking and what documents to bring
  • Ask only for name, contact details, matter type and urgency
  • Show an obvious route to a human
  • Warn visitors with urgent deadlines to contact a lawyer or the court
Don't
  • Let it answer questions about the visitor's rights or what to do
  • Ask visitors for detailed facts or confidential material
Setting up a website chatbot

The handoff matters more than the answer

The strongest use of a chatbot is as a front door. It collects a few fields and hands them to a person. What happens next decides whether the tool helps or harms.

Suppose a four-lawyer firm uses a chat tool that drops every enquiry into one shared mailbox. Nobody owns the mailbox. An urgent enquiry about a deadline sits there over a weekend. The chatbot did exactly what it was set up to do, and the visitor was still let down. Decide who reads enquiries, how quickly they respond and what the chat says about timing, and then make the promise on screen match what the firm can deliver.

Conflict checks before conversation

An enquiry is not a client. Treat the details a chatbot collects as an enquiry that still needs a conflict check, and make sure nobody starts giving substantive help until that check is done. Names entered in a chat box are often incomplete or misspelt, so a check by exact match will miss some. A person should look at the result and decide.

Testing before you go live

Treat the launch like any other risk-bearing change. Have several people at the firm try to break it. Ask it about their own made-up legal problems. Ask it to predict the result of a case. Tell it you are in a different country. Paste in a long, emotional message. Read the transcripts and tighten the instructions until it declines properly.

Test the handoff as well as the words. Submit an enquiry as a visitor would, and time how long it takes before a person responds. If the answer is longer than what the chat window promises, change one or the other.

Then keep checking. Review a sample of real transcripts every month at first. Look for answers that sound like advice, for confidential details that should not have been collected, and for visitors who asked for a human and did not get one.

  1. 1Try to break it with awkward questions
  2. 2Read the transcripts
  3. 3Tighten the instructions
  4. 4Time the handoff to a person
  5. 5Review a sample of real transcripts
  6. Then back to step 1.
Keeping a chatbot honest

When to say no to a chatbot

A chatbot may not be worth it for a small firm that gets a handful of enquiries a week. A clear enquiry form and a prompt reply may do the same job with less exposure. If you cannot say who will read the transcripts, who will update the instructions and who will handle a complaint, you are not ready.

If you want an enquiry route without a chatbot

  • Each firm has a public enquiry form with its own link.
  • Enquiries arrive in an inbox with a name-based conflict warning.
  • Nothing from the enquiry form becomes a client until someone at the firm converts it.

What it does not do

  • It is a form and an inbox, not a live chat or an advice tool.
  • The conflict warning is based on names. The decision stays with a lawyer.

This article is general information about the law, legal practice and technology, current when it was published. It is not legal advice. Laws and guidance change: check the sources linked above and take advice on your own circumstances from a lawyer qualified where you are.

Read next

Law firm website chatbot: benefits, risks, rules