Can an AI agent handle a complaint that mentions legal action?

Only the first step. Once a complaint mentions solicitors, court, a claim or a regulator, the agent acknowledges it, logs it, preserves the record and hands it to a person, who involves the legal team if needed. It makes no admissions, settlement offers or comments on the law. Routine service on unrelated matters can carry on.

Why does a mention of legal action change things?

The words customers use are a weak signal of what they will do. "I'll see you in court" may be frustration; a formal letter from a law firm is not. The agent cannot tell which, and it does not need to. Both mean that anything said from now on may be read later by someone looking for an admission or a promise.

That changes what a good reply looks like. Warm, specific apologies and offers of goodwill are usually good service. Here, they can be read as accepting liability or as settlement offers. A person with legal support should choose the words.

What may the agent do, and what stays with a person?

Preserving records is easy to forget. Conversation logs may be subject to routine deletion; once legal action is in view, they must be kept. How an agent supports that is covered in how to handle legal holds with an AI agent.

Why should the agent never comment on the law?

Because it would be giving legal advice to an opponent, and possibly getting it wrong. A customer asking "would I win?" or "what does consumer law say about this?" gets a polite refusal and a pointer to independent advice. The limits are set out in whether an AI agent can give legal advice.

What does this look like in practice?

Take a home insurer. A policyholder whose subsidence claim was declined writes: "If you don't reverse this, my solicitor will be in touch and I'll go to the ombudsman."

The agent logs a complaint with a legal flag and triggers retention of the claim file and the conversation. It replies with the approved acknowledgement: the complaint is logged, a named team will investigate, and the response will arrive within the time the policy sets. It does not explain why the claim was declined again, does not say the decision was right, and does not hint at a review. In the same message the customer asks for a copy of their policy schedule. The agent sends it, because that is routine and the customer is entitled to it. The complaint team takes the case from there.

What does the formal complaint process still require?

A legal threat does not pause the complaint. In regulated sectors, the deadlines and content requirements for complaint handling still apply, and a final response is still due. The basics are in what complaint handling is, and who may sign the final letter is covered in whether an AI agent can send the final response to a complaint.

Which controls make it safe?

• Detection of legal language in every certified language, routed to a person.
• An approved acknowledgement, used word for word.
• A block on apologies for fault, compensation offers and legal commentary once the flag is set.
• Automatic retention of related records.
• Every message after the flag filed and visible to the complaints and legal teams.

Where a complaint concerns a decision that affects the customer's legal position, such as a declined claim, the reasoning in what a decision with legal effect is also applies. The wider approach is in the guide to AI customer support agents.