AI governance in contact centres comes down to four decisions, written down: what your AI is allowed to do, who owns it, how customers reach a person, and how you will know when something goes wrong. On 19 October 2026 the Financial Conduct Authority published examples of good practice that turn those decisions into something practical. It creates no new rules, but any leader running chatbots, agent assist or automated triage now has a clear benchmark to test against.
This piece looks at what the FCA said, why it matters beyond financial services, and where a contact centre leader in London or the South East might start.
What did the FCA publish on AI in financial services?
The FCA’s page, AI in financial services: good practice and areas for improvement, draws on the regulator’s supervisory work and its AI lab initiatives, including AI Live Testing and the Supercharged Sandbox. The examples are described as illustrative and non-exhaustive, and the FCA is clear that they do not create new requirements.
What they do is show what the regulator has seen working. Much of it lands squarely in customer contact: how firms tell customers they are dealing with AI, how chatbots hand over to people, how vulnerability is spotted, and who is accountable when an AI tool gets it wrong. One line deserves a place on every procurement checklist: firms remain accountable for outsourced AI services. Buying a capability from a supplier does not move the responsibility.
Why does FCA AI good practice matter outside financial services?
If you run a contact centre for a retailer, a utility or a hospitality group, it is tempting to file this under someone else’s problem. That would be a mistake, for three reasons.
First, customers do not separate their experiences by regulator. Banks and building societies overtook retail in the Institute of Customer Service’s July 2026 UK Customer Satisfaction Index for the first time, with the overall Index at 78.3. The sectors that set the standard tend to set expectations for everyone else.
Second, data protection applies to all of us. The Information Commissioner’s Office ran a consultation on draft guidance about automated decision-making from 31 March to 29 May 2026, following the Data (Use and Access) Act 2025. Any contact centre using AI to route, prioritise or decide outcomes for customers should be watching for the final version.
Third, the FCA’s examples are simply sensible. Strip out the regulatory language and they describe how any well-run operation would introduce a new tool that talks to customers.
What does good AI governance in contact centres look like?
Reading the FCA’s examples with a contact centre lens, six themes stand out.
Tell customers when they are talking to AI
Good practice included telling consumers that an AI tool was being used and explaining its limits, including when to seek help elsewhere. Chatbots were restricted to approved source material, so they answered from content the firm had signed off rather than from the open web.
Keep a route to a person, without repetition
Where appropriate, customers could choose between AI-enabled and non-AI support for simple tasks, and switch channels without repeating information. For most contact centres the second half of that sentence is the harder test. A handover that loses the conversation history is not a handover.
Use AI to spot vulnerability, not to sort people out of help
Some firms used AI to flag possible signs of vulnerability, such as a customer struggling with a digital journey, and one chose its indicators with frontline staff. The FCA also warns against excluding vulnerable customers from AI journeys by default, as that can leave them without support. The aim is better routing, not a blanket opt-out.
Give every AI tool a named senior owner
Firms mapped AI risks to specific senior managers from the design stage, kept auditable records of decisions and escalations, and gave senior people targeted training. Outside financial services there is no equivalent of the Senior Managers and Certification Regime, but the principle travels: one named person accountable for each tool that touches customers.
Test before go-live, then keep watching
Good practice included testing outcomes across different customer groups, retesting after changes, and running live monitoring with thresholds, spot checks and staff feedback. One firm had reviewers assess cases without seeing the AI’s proposed outcome, which reduced automatic deference to the system. That is a useful idea for any quality team checking AI-assisted work.
Plan for your supplier changing the model
Firms designed services to work with more than one provider, tested contingency plans, and used contracts to get notice of significant model changes. If your AI supplier updates the model underneath your bot next month, would you know, and would you retest?
Where should a contact centre leader start with AI governance?
None of this needs a new department. A practical first pass might look like this:
- List every AI tool in use. Include agent assist, speech analytics, summarisation, chatbots and anything a supplier has switched on inside an existing platform.
- Decide what each one supports, automates or leaves to people. Write it down, with the reasoning.
- Name an owner for each. One senior person, accountable for outcomes, not just uptime.
- Walk the customer route to a person. Try it as a customer would. Count how often you have to repeat yourself.
- Set thresholds and sample reviews. Decide in advance what a bad week looks like and who acts on it.
- Check your contracts. Make sure you will hear about significant model changes before your customers do.
For regulated firms, much of this will already sit somewhere in the risk framework. The gap is usually between the framework and the floor: the people running queues, coaching advisers and handling complaints often have the clearest view of where AI helps and where it quietly makes things worse.
How LCCF is approaching AI and automation
That gap between policy and practice is exactly what peer conversation is good at closing. London Contact Centre Forum (LCCF), part of the Contact Centre Forum network, brings together the leaders who run customer operations, CX, technology and transformation across London and the South East. The network was built by Jane Thomas across the South West, the North and Scotland, and London now has its own home within it.
AI and automation are already on the programme. On Thursday 26 November 2026, the LCCF Eco-System Event offers London contact centre leaders a deep-dive into AI, automation and service design. On Thursday 10 December 2026, the Operational Excellence 2027 Challenges Workshop focuses on practical challenge-solving for members. Details and booking are on the London Contact Centre Forum page.
Between events, members can join monthly virtual user groups for operations, training, planning, HR and team leaders, the Friday Update every Friday at 11am, and Forum on Tour visits to leading contact centres. For more from across the network, see our news and updates, or read how London launched its dedicated contact centre forum.
Frequently asked questions
Does the FCA’s AI good practice create new rules?
No. The FCA says its examples are illustrative, non-exhaustive and do not create new requirements. They show how firms are meeting existing obligations, including the Consumer Duty, when they use AI. Firms remain accountable for the outcomes, including where the AI service comes from a third-party supplier.
Does AI governance matter if my contact centre is not FCA regulated?
Yes. Data protection law applies to every contact centre, and the ICO has consulted on updated guidance on automated decision-making following the Data (Use and Access) Act 2025. Customer expectations also travel between sectors, so the standards set by banks quickly become the standards customers expect from everyone.
Should customers always be able to reach a human?
The FCA’s good practice examples include giving customers a choice between AI-enabled and non-AI support for simple tasks, where appropriate, and letting them switch channels without repeating information. It also warns against shutting vulnerable customers out of AI journeys by default, so the answer is a well-designed route, not a blanket rule.
Who should own AI in a contact centre?
One named senior person for each AI tool that affects customers, with documented decisions and escalation routes. In regulated firms this maps to Senior Management Functions. Elsewhere, the operations or CX director is often best placed, working closely with technology, risk and the frontline teams who see the outcomes every day.
Join the conversation in London
If AI governance is on your agenda for 2027, you will not be the only one in the room. Request an invitation to an upcoming LCCF event, or talk to Jane about membership for your organisation.