Embedding Consumer Duty in an AI governance framework

In July 2026, the FCA published the Mills Review assessing how AI could reshape retail financial services by 2030. The review considers how the balance between the uses of artificial and human intelligence may change in the next few years. Mills predicts a five-step autonomy spectrum: At the operator level, humans use AI as a tool, and as a collaborator they work with it directly. As a consultant, AI recommends actions and a human decides which course to take. But as AI use becomes more sophisticated, AI will prepare actions and humans will act as approvers. Finally, AI will be able to act autonomously, working within agreed boundaries while humans monitor actions as observers.  
 
This transformation will affect compliance activity under the Consumer Duty. The FCA believes AI can and should operate in accordance with the Consumer Duty aims and principles. It believes that firms can continue to carry out Consumer Duty activities effectively during the first three steps of the autonomy spectrum. But once humans act as approvers, the regime may have to operate differently. At this point, as AI recommends or takes actions, it could become tricky for firms to demonstrate consumer consent and understanding, and it could become more challenging to document fair value and product suitability. At the final stage, where humans are observers, the regime will come under further pressure, making it difficult to evidence consumer outcomes. 

New customer journeys and experiences under autonomous AI 

The shift towards autonomous AI means that many customers journeys will start with AI rather than a product search or financial adviser. Consumers will be able to describe their needs to an AI agent, which will research the market and act on their behalf. As AI moves along the autonomy spectrum, this could eventually result in AI agents that switch products or manage savings, within agreed limits. This could revolutionise how consumers access financial services, potentially increasing consumer engagement, improving customer outcomes and supporting growth. 
 
Consumer research, commissioned by the FCA, found a real appetite for this shift, with 20% of consumers saying they were open to AI making autonomous decisions based on pre-set instructions or goals. That number rose to 28% among adults already using AI. Key use cases include debt advice, investments and pensions, and demand is strongest for areas that are complex, high stakes or difficult to navigate. However, trust is fragile, with 67% concerned about a lack of protection for products bought this way, 68% worried about misuse of data, and 65% worried about a concentration of power among a small number of providers. 
 
The ‘mapping’ of customer journeys, and firms working to clear articulations of ‘good outcomes’ and ‘foreseeable harms’ that can be measured, is the foundation layer of a good Consumer Duty framework. Firms must revisit and update their infrastructure to account for the role of AI, including both their own and customers’ uses of it.  

Article

Mills Review: enhancing AI governance in financial services

Read more

AI brings additional risk 

As the Mills Review sets out, effective use of AI could reduce the advice and protection gap, and improve investment decisions. But successful uptake and use will rely on tools that are accessible to customers, and trusted and well controlled. The Consumer Duty obligations to ensure that customers understand what they need, when they need it, and have access to appropriate support, remain in place. As such, firms will need to design customer journeys and experiences with AI in mind, and embed appropriate risk management and an AI governance framework.  
 
This includes establishing how a firm might interact with an AI agent on a customer’s behalf, or how to operate product lifecycles using autonomation or AI. Firms must establish the associated risks, ensuring they design and operate effective controls that align with the firm’s risk appetite. Firms’ risk management and governance frameworks should also include clarity and precision regarding which decisions are made by AI and which ones by humans, and where human accountability sits.  

Applying Consumer Duty in an AI-enabled market 

The FCA isn’t currently planning new regulations for AI adoption, so firms need to factor AI activity into their existing frameworks and Consumer Duty-related activities. The Mills Review emphasises that responsibility for customer outcomes remains with firms even where AI systems act autonomously. 

In time, the FCA is planning to produce further guidance as its supervisory approach and expectations evolve. This will help firms embed effective governance, accountability and consumer protection. In the meantime, firms need to actively assess how AI can support the four Consumer Duty outcomes to help individuals achieve their financial goals.  

Products and services

To ensure products and services meet the needs of the target market, firms need to map the customer journey including any AI-enabled elements. They need to define what counts as a good outcome and what foreseeable harm looks like, including harm from the technology itself. These must be supported by robust metrics that can inform product reviews, fair value assessments and ongoing outcomes monitoring. These are all ongoing expectations for firms, so the use of AI isn’t creating new expectations, it’s simply another factor to consider.

Price and value 

AI offers greater personalisation and scope to reduce firms’ costs, but they need to continue to ensure that prices are justified and reflect value, including any genuine difference in risk profile, cost or consumer benefits. There is a risk that different pricing could veer into extraction, where consumers are charged what they’re willing to pay, without any perceivable differences in the product or services. This can be hard to spot, but is a key area that the FCA will undoubtedly be vigilant on. 
 
As firms realise efficiencies from AI adoption, they should assess whether their pricing and fair value assessments continue to deliver appropriate value to customers.  

Consumer understanding 

When AI tailors a journey and makes lots of small decisions over time, it gets harder to demonstrate customer understanding under Consumer Duty. Firms may find it challenging to evidence consumer understanding where AI agents act under broad delegated authority, including that consumers understand the product or service features, and the associated risks. Firms using AI should enhance their capabilities to ensure they are clear on what constitutes genuine consumer understanding, that customers continue to receive information at the right time, and they are informed to make good decisions.  

Consumer support 

Firms should identify the type and level of support consumers need, or can reasonably expect, from a given product or service. They also need to provide that support, be able to show it, and demonstrate that it has enabled good consumer outcomes.  
 
AI is already playing a significant role , by offering immediate customer support, but human involvement is still needed for complex or sensitive issues. AI is also widely used to scan and review large volumes of customer interactions to extract information on the extent and quality of consumer support. These can inform Consumer Duty management information and support outcomes monitoring, but must be subject to a robust AI governance framework and appropriate oversight. 

Article

Top themes for the financial services sector in 2026

Read more

What firms should consider now 

The Mills Review doesn’t introduce new rules, but firms need to consider how their AI usage could affect Consumer Duty compliance. That means actively identifying the impact of AI on current activities and procedures, and how it could lead to foreseeable harm throughout the customer journey. To get started, firms can: 

• Map customer journeys that include AI, and/or are enabled or controlled by AI, and identify exactly where decisions shift from human to AI control.
• Review pricing models to confirm they reflect genuine value given AI involvement, not just what a consumer will pay.
• Keep cost-to-serve data current, and update prices when AI-driven efficiencies bring costs down.
• Build ways to evidence consumer understanding and consent, and the provision of support, especially as AI models become more autonomous.
• Embed appropriate risk management processes with a robust AI governance framework to monitor AI use and ensure it continues to support good customer outcomes. 

For further information on Consumer Duty and creating an AI governance framework, contact Supriya Manchanda or Peter Lovegrove.