Yes. The Consumer Duty applies to the firm, and it judges the outcomes retail customers receive. It does not matter whether a person, a script or an AI agent delivered them. If an agent answers, advises, sells, collects or handles complaints for retail customers of an FCA-regulated firm, its conduct is the firm's conduct under the Duty.
Who does the Consumer Duty apply to?
The FCA Consumer Duty sets a higher standard of care for firms whose products and services reach retail customers in the UK. In substance, firms must act to deliver good outcomes, which means acting in good faith, avoiding foreseeable harm and enabling customers to pursue their financial objectives. The Duty looks at four areas of outcome: the products and services themselves, price and value, consumer understanding, and consumer support.
It applies across the distribution chain, so manufacturers and distributors both carry it. The rule is technology-neutral: it neither permits nor forbids agents. A fuller summary is in what the FCA Consumer Duty is.
This is an explanation of the rules, not legal advice; confirm how they apply to you with counsel.
What changes when an agent is the one acting?
Three things become sharper.
Consistency at scale. One adviser who misreads a policy harms a few customers. An agent with the same misreading harms every customer who asks that question, the same day. The Duty's focus on foreseeable harm means you must look for the systematic error before it spreads.
Consumer support. The Duty expects support to be as easy as buying. An agent that resolves simple questions quickly but makes cancellations, complaints or hardship requests loop back on themselves creates exactly the friction the FCA wants removed. Whether an agent helps or hinders support is examined in Consumer Duty and AI support agents.
Vulnerable customers. An agent must recognise signs of vulnerability it may not be told about directly, and change course: slow down, simplify, or hand over to a person.
What may the agent do, and what stays with a person?
- Agent may: answer factual questions from approved sources, explain product features and charges, take routine requests such as address changes, and draft responses.
- Agent may, within limits: apply goodwill credits or fee refunds up to a set amount, agree standard payment arrangements within policy.
- Stays with a person: any decision where the customer shows signs of vulnerability, complaint outcomes, forbearance outside policy, and anything that amounts to a personal recommendation unless the firm has deliberately designed and governed that service.
Lending raises its own pressures, especially around affordability and arrears, which are set out in FCA Consumer Duty and AI agents in lending.
An illustrative example
Take a UK credit card issuer using an agent to handle chat queries. Customers often ask how to close their card. The agent, briefed to "retain customers where possible", answers every closure request with three retention offers before explaining how to close.
No single conversation looks harmful. Across thousands, the pattern makes closing harder than opening, and customers in financial difficulty are spending more time with the agent than they should. The issuer changes the brief so closure requests get the steps first and one offer at most, routes any mention of money worries to a person, and adds the closure process to its outcome monitoring. Its board report then shows closure times for agent-handled and staff-handled chats side by side.
Which controls show the Duty is met?
| Duty area | Control for the agent | Evidence |
|---|---|---|
| Products and services | Agent's scope limited to products it was tested on | Role scope and test results |
| Price and value | No discretion over pricing beyond set limits | Limits and held exceptions |
| Consumer understanding | Answers grounded in approved, current sources | Source trail per answer |
| Consumer support | Easy routes to a person; no loops on cancellations or complaints | Outcome monitoring by request type |
| Vulnerability | Escalation triggers on vulnerability signals | Escalation log and review |
How the FCA approaches agents more broadly, including accountability under the senior managers regime, is covered in how the FCA views AI agents. For customer service specifically, see FCA Consumer Duty and AI agents in customer service, and for the full regulation map, AI agent regulation, mapped to controls.
Questions people ask
Do we have to tell customers they are talking to an agent?
The Duty's consumer understanding outcome means customers should not be misled about what they are dealing with. Telling them is the safe default, and it should be easy to reach a person.
Can an agent identify vulnerable customers?
It can detect many signals, such as mentions of illness, bereavement or job loss. It should then hand over rather than try to manage the situation itself. Detection must be tested on realistic cases.
Does outcome monitoring have to separate agent and human channels?
The Duty asks firms to monitor outcomes and act on differences between groups of customers. Splitting results by channel is the simplest way to show an agent is not producing worse outcomes.