Chatbots in client communication: useful or risky?

What a chatbot can do in a law firm

The typical enquiry does not reach a small law firm during office hours. The employee found the dismissal letter in the letterbox after work and now wants to know whether anyone will take care of it. Law firm chatbots are marketed for precisely this moment: they take the enquiry around the clock, ask for the name, area of law and the matter at hand, answer standard questions about opening hours or initial consultations and accept documents. One provider in the legal market puts the running costs of simple solutions at 600 to 1,000 euros a month, plus 4,000 to 6,000 euros for setting them up.

The benefit therefore does not lie in legal advice, but at reception. If you already have the first details before the phone call, you can assign enquiries to a lawyer more quickly and spot hopeless cases early. For a law firm without a secretary on the night shift, that is a real advantage. The only question is what else the chatbot says.

What clients expect from the machine

At the end of 2024, the digital association Bitkom surveyed a good 1,000 people in Germany. 61 per cent named round-the-clock availability as the greatest advantage of AI for legal questions. At the same time, only 21 per cent believed that an AI has more expertise than a human. More than half took the view that an AI cannot put itself in the position of those affected, and the vast majority still wanted to ask a lawyer about a legal problem.

For use in a law firm, that is a useful map. People value availability, but they want the advice from a human. A chatbot that takes enquiries and arranges the call back meets that expectation. A chatbot that assesses the prospects of an unfair dismissal claim at night does not, and creates liability questions into the bargain.

When the chatbot invents titles

A decision of the Hamm Higher Regional Court from May 2026 shows how quickly that can happen. A provider of aesthetic treatments had an AI chatbot on its website that booked appointments and answered questions. When asked, it described the two managing directors as specialists in plastic and aesthetic surgery. The doctors did not hold those titles. The consumer advice centre issued a warning letter and brought proceedings.

The court attributed the false answers to the company as its own commercial act, just like an advertisement. The chatbot was not a third party whose statements only trigger duties to check. Nor did it help to argue that the system had been fed only correct data. An appeal to the Federal Court of Justice has been allowed. For a law firm, the case is easy to translate: if the chatbot declares the owner to be a certified specialist without being asked, or promises a free initial consultation that does not exist, the firm is answerable for it. Language models formulate freely, and that is precisely the problem under advertising law.

Confidentiality from the first word

The second risk is less obvious. Anyone who types into a chat window at night that they have been dismissed and that their boss has been harassing them for months is entrusting the firm with a secret. A lawyer's duty of confidentiality also applies to this preliminary stage of a matter. If the chatbot runs on a provider's servers, that provider is a third party gaining access to client secrets. Professional law permits this only under certain conditions: the service provider must be carefully selected, bound to confidentiality in writing and limited to what is necessary. Where servers are located outside the EU, a comparable level of protection is also required, and according to an assessment from 2023, some US services fail on that point.

At the end of 2024, the German Federal Bar published guidance on the use of AI. It sets out the principle that requirements increase with the degree of automation. A heightened standard of care applies to systems that communicate directly with clients, and the chatbot must not create the impression that a lawyer's review of the individual case is unnecessary. It is also recommended to be open about the use of AI, even though professional law does not yet expressly require this.

The chatbot must identify itself

What the Bar recommends has been mandatory under the European AI Regulation since 2 August 2026. Anyone who lets people talk to an AI system must inform them clearly, at the latest at the first interaction, that they are dealing with a machine. A note in the terms of use is not enough; the information belongs in the chat window or in the greeting. The only exception is where it is obvious anyway to an attentive user.

For a law firm, that is not a hurdle but rather a clarification. A chatbot that introduces itself as one and says what it can and cannot do raises fewer false expectations. Anyone who instead creates the impression that a human from the firm is at the other end breaches the regulation and risks the very trust that, according to the Bitkom survey, clients place in people.