FCA Consumer Duty and AI agents in customer service

The FCA Consumer Duty applies to an AI agent serving retail customers exactly as it applies to staff. The agent must give support that meets customers' needs, without unreasonable barriers to getting help, switching, cancelling or complaining. That makes "containment", keeping customers away from people, the wrong target. The right one is whether the customer's problem got solved.

What does the Duty require of customer service?

The Consumer Duty's consumer support outcome asks firms to provide support that meets customers' needs throughout the relationship. In substance, firms should make sure:

• customers can use the product as reasonably anticipated;
• customers can get help, make claims, complain and switch or leave without unreasonable barriers;
• post-sale support is no harder to use than the sales process;
• support is responsive and timely, and takes account of customers with characteristics of vulnerability.

The Duty's cross-cutting rules apply too: act in good faith, avoid causing foreseeable harm, and enable customers to pursue their financial objectives. The FCA has been clear that "sludge", friction that keeps customers from doing what they want, is a concern. Our explainer on what the Consumer Duty is covers the framework.

Who does it apply to?

FCA-authorised firms serving retail customers, including banks, building societies, lenders, payment and e-money firms, insurers and investment firms. If an agent answers your customers, the Duty applies to its service; see whether the Consumer Duty applies to AI agents.

What does it mean when an agent provides the service?

Agents can improve service: no queues, instant answers at any hour, consistent information. They also create specific risks the Duty is built to catch.

Containment as a target. Many service teams measure how many conversations an agent resolves without a person. Pushed hard, that number rewards the agent for deflecting, repeating itself and making the route to a person hard to find. Under the Duty, an agent that keeps a customer from help they need is causing foreseeable harm.

Asymmetric friction. If the agent can open an account in two minutes but cancelling requires a call to a person during office hours, post-sale support is harder than the sale. Agents make that asymmetry easy to build by accident, because sales journeys get automated first.

Retention scripts. An agent asked to "save" cancellations can become persistent in a way a person would not. Repeated offers after a clear request to leave are a barrier.

Actions on the account. Agents that can freeze cards, change payment dates or apply fee refunds can resolve problems on the spot, which is good support. Each such action needs bounds, so that the agent's help does not create a new harm.

Take an illustrative case. A customer tells the service agent they want to close a savings account. A Duty-minded agent confirms what closing means, including any interest or notice period, makes one relevant offer if there is one, and when the customer repeats the request, starts the closure or books it with a person in the same conversation. A poorly designed agent replies with three retention offers, then tells the customer to call during office hours. Both might log the conversation as handled. Only one delivered what the customer came for.

Communications quality is its own outcome; see Consumer Duty and AI agent communications.

How do you show good outcomes from an agent in service?

Measure what the customer got, not what the firm saved.

Vulnerability needs particular care; see vulnerable customers and AI agents. And an agent can help a firm monitor outcomes across its service, with a person drawing conclusions, as our answer on whether an agent can monitor outcomes for Consumer Duty explains.

What will the FCA or internal audit ask?

• What does the agent measure itself against, and does containment drive its design?
• Show us how a customer reaches a person, and how long it takes.
• Show us a cancellation handled by the agent. How many retention offers were made?
• Are outcomes for customers served by the agent as good as for those served by people?
• How often do customers come back with the same problem?
• What did the board see about agent service in its Duty review?

What does the support outcome not cover?

It does not settle whether a product offers fair value or suits its target market. It does not replace the complaints rules, which set their own handling and time limits. It does not cover data protection rights, which follow UK GDPR. And it covers only retail customers. For the second-line principle that compliance agents prepare and people decide, see our guide to AI compliance agents.

This is an explanation of the rules, not legal advice; confirm how they apply to you with counsel.