What is complaint handling?

Complaint handling is the process a firm uses to receive, record, investigate and resolve expressions of dissatisfaction from customers, and to tell them the outcome and their options if they disagree. In regulated sectors it runs to deadlines and rules set by the regulator. AI agents can log, classify and draft; the final response belongs to a person.

Why does complaint handling matter once agents act?

Two reasons. Agents will receive complaints, because customers complain to whoever is in front of them. An agent that treats "this is the third time I've asked and I'm fed up" as a routine query has failed to record a complaint, and the deadline clock may already be running.

Agents can also make the process much better. They can spot a complaint inside an ordinary message, log it the moment it arrives, classify it, pull the account history and draft the investigation file. That leaves the complaint handler more time for the judgement: what went wrong, what is fair and what to offer.

How does complaint handling work in practice?

Most regulated processes share these stages:

1. Identify and record: recognise the complaint and log it with the date received.
2. Acknowledge: tell the customer it has been received and what happens next.
3. Classify: product, cause, severity and any vulnerability. A complaint classification matrix gives the categories.
4. Investigate: gather facts, consider the customer's account and the firm's records.
5. Respond: a final response with the outcome, any redress and the customer's right to escalate to an external body where one exists.
6. Learn: analyse causes across complaints and fix them.

Deadlines usually apply to acknowledgement and final response. Tracking complaint deadlines with an AI agent is one of the safest high-value jobs it can do.

An example from a broadband provider

Take a broadband provider whose support agent receives a message: "My internet has dropped every evening for two weeks, I work from home and I've lost a client because of it." The agent recognises a complaint, logs it with the timestamp and sends the approved acknowledgement. It classifies it as a service fault with possible financial loss and attaches the line's fault history.

A complaint handler reviews the file, confirms the fault, decides on a refund for the affected period and considers the claim for lost business. The agent drafts the final response. The handler edits it and sends it. The agent did not decide the outcome or the compensation.

How is a complaint different from a service request?

A service request asks the firm to do something: change an address, send a statement, fix a fault. A complaint expresses dissatisfaction with something the firm did or failed to do, and usually expects a remedy.

The same message can be both, and the difference has consequences: complaints have deadlines, records and escalation rights that service requests lack. Firms often lean towards treating ambiguous messages as complaints. An agent should be tested specifically on that boundary, with real messages where the line is blurred.

Which terms sit next to complaint handling?

• Final response: the firm's formal answer to a complaint, which an AI agent should not send on its own.
• Vulnerability: a customer's circumstances that call for extra care, covered in the vulnerable customer handling checklist.
• Root cause analysis: finding why complaints happen, so fewer happen.
• Our guide to AI customer support agents that know when to say no covers handover rules generally.