How the Mills Review could reshape AI governance in financial services
On the 6th July 2026, the FCA published its final findings of the
10/07/26
Read more
How the Mills Review could reshape AI governance in financial services
On the 6th July 2026, the FCA published its final findings of the Mills Review, its assessment of how advances in artificial intelligence (AI) could reshape financial services by 2030 and beyond. The review was commissioned by the FCA Board and draws on 140 written submissions from firms, technology providers, consumer groups, academics and regulators – alongside consumer research involving more than 5000 UK adults.
Whilst much of the industry’s discussion has previously focused on AI’s current use cases, the Mills Review takes a longer-term view. It examines what financial services might look like if AI becomes more capable, more widely adopted and increasingly trusted by firms and consumers. Its conclusion is that the industry may gradually move from a world where AI primarily supports human in decision-making to one where it increasingly influences, recommends and carries out actions on behalf of businesses and consumers. As the report author Sheldon Mills (FCA Executive Director) puts it “So, the question is no longer whether to allow AI, it is who is AI going to serve? My own view, as an almost former regulator, is that the answer must be citizens, and that empowering them is one of the UK’s great opportunities for economic growth.”
The growing role of AI in decision-making
Most firms are already using AI in some form, whether to support customer service, identify fraud, analyse data or improve operational efficiency. However, in most cases people still sit at the heart of the decision-making process. The Mills Review explores beyond this current scope. It considers what the future may look like if, in some instances, technology moves closer to making decisions on its own – albeit within determined parameters.
For compliance and risk teams, this is where the Review becomes particularly relevant. The more responsibility that is delegated to AI, the more important the questions around accountability, governance and oversight of customer outcomes become. The challenge is not whether firms use AI – many already do – but about understanding how firms can continue to deliver fair customer outcomes as AI becomes increasingly embedded across processes.
From AI assistance to AI delegation
One of the Review’s central themes is about AI moving beyond being a productivity tool. AI is currently used across most walks of life to summarise information, identify patterns and support with routine tasks – and financial services is no exception. But the review looks to a future in which AI moves from assistance to delegation. In other words, systems do more than provide information. They may recommend actions, initiate processes and in some cases, carry out tasks or transactions within defined limits.
An instance of this might be a customer who initially uses AI to compare savings rates. Over time, they may trust the system to identify a better product and recommend an alternative. Eventually, they may allow AI to make that switch automatically within agreed parameters. Similarly, a firm might begin using AI to identify potentially vulnerable customers before progressing towards using AI to determine the most appropriate intervention or support pathway.
What an AI-enabled financial system could look like
The Mills Review describes several shifts that could reshape financial services during the next decade. The first is how AI could become a fundamental part of firms’ information and operational processes, in areas such as customer support, claims handling, and compliance monitoring.
The second is the emergence of agent-led customer journeys. Consumers may increasingly rely on AI systems to help manage their financial lives. For instance, AI could monitor spending patterns, identify opportunities to reduce debt and alert customers to better savings rates. The FCA’s research found that, in some circumstances, one in five consumers are already open to using AI that can make financial recommendations or take actions on their behalf within pre-defined goals and parameters. This reflects one of the Review’s broader observations, that AI has the potential to address longstanding challenges in retail financial services including advice gaps, low switching rates and poor engagement with long-term financial planning.
That said, the same technologies that support customers can also be exploited. The review highlights the growing threat posed by deepfakes, synthetic identities, AI-enabled fraud and increasingly sophisticated social engineering attacks. At the same time, it also recognises that AI can strengthen a firm’s ability to detect and respond to these threats.
The role of humans is changing, not disappearing
The review does not predict a future in which humans disappear from financial services. Instead, it explains how the human role is likely to evolve. As firms adopt increasingly autonomous systems, people may spend less time performing individual tasks and more time setting parameters, reviewing exceptions, monitoring outcomes and challenging AI-generated decisions.
In turn, that creates important governance questions for the future. Meaningful oversight requires more than simply keeping a human somewhere within a process. Firms need to understand what information that individual receives, what decisions they are expected to make, when intervention is required and whether they possess sufficient understanding of the technology they are overseeing.
This becomes particularly important as AI systems become more adaptive. The further firms move along the autonomy spectrum, the more important it becomes to clearly define where accountability sits and what effective oversight looks like in practice.
Why accountability becomes harder
As AI becomes more capable, responsibility can become harder to trace, the Mills Review outlines. Firms may rely on multiple models, third-party providers and increasingly complex technology supply chains. Different AI systems may influence different stages of a customer journey, making it more difficult to understand how outcomes are being produced.
The Review also highlights the importance of consistency and model drift. If a system evolves over time, firms need confidence that outcomes remain fair, appropriate and aligned with regulatory expectations. A model that performs accurately during testing may behave differently months later if customer behaviour, data inputs or market conditions change. For example, an AI solution used to identify vulnerable customers may initially perform well, but firms still need confidence that it continues to identify customers consistently as circumstances change. Equally, AI-assisted decision-making processes need ongoing monitoring to ensure they are not producing unintended outcomes across different customer groups.
This means AI governance cannot be treated as a one-off approval exercise. It requires continuous monitoring, testing and challenge.
Why the Consumer Duty remains central
One of the more reassuring conclusions from the Mills Review is that it does not call for an entirely new regulatory framework. The review concludes that existing regimes, including the Consumer Duty and the Senior Managers Regime (SMR) provide a strong foundation for an increasingly AI-enabled sector.
The challenge for firms is demonstrating how those frameworks continue to operate as technology takes on a greater role in decision-making. Say, if AI is used to personalise customer journeys, firms need to understand whether those journeys consistently deliver good outcomes across different customer groups. Or, if AI helps identify vulnerabilities, firms need assurance that it does so accurately and fairly. These are not fundamentally new questions. They are existing conduct, governance and Consumer Duty considerations viewed through the lens of increasingly autonomous technology.
The regulator is preparing too
The Mills Review is not solely concerned with how firms use AI. It also considers how regulation itself may evolve. One of its recommendations is the development of an AI-enabled Agentic Supervisory Model, which would allow the FCA to better identify emerging harms, cross-firm trends and system-wide risks in an increasingly interconnected environment.
As firms become more data-driven and AI-enabled, regulators across multiple sectors are likely to become more data-driven and AI-enabled too. Future supervision may become increasingly focused on identifying patterns, monitoring outcomes and understanding risks that emerge across the wider ecosystem rather than within individual firms alone.
Preparing for what comes next
The future described in the Mills Review may still be emerging, but the governance challenges it raises are already here. Firms do not need to wait for fully autonomous AI systems to start considering questions around accountability, oversight, resilience and customer outcomes. As AI becomes more embedded across financial services, the firms that are best prepared will be those that understand where technology influences decisions, how those decisions are governed and how good outcomes can be evidenced over time.
At TCC, we work with firms to navigate regulatory change, strengthen governance frameworks and ensure evolving technologies are implemented in ways that support both regulatory compliance and customer trust. Through our technology partner Recordsure, firms can also access AI-powered tools that help monitor customer interactions, oversee evidence, identify risks, and support better outcomes at scale. Whether reviewing AI governance, assessing Consumer Duty implications or building robust oversight frameworks, we can help firms prepare for an increasingly AI-enabled financial system with confidence.