The UK Financial Conduct Authority (FCA) has announced that it intends to become an “agentic regulator,” using AI in its supervisory activities. This comes at the same time as the FCA is developing its expectations around the use of AI by those it regulates, as discussed in our November 2026 GT Alert, AI in UK Financial Services: FCA Highlights Good Practice and Areas for Improvement.
The FCA’s announcement came in a recent speech delivered by Jessica Rusu, the FCA’s chief data, information and intelligence officer. The announcement follows the publication of the Mills Review earlier in 2026. That review set out how AI might shape retail financial services and made recommendations about how the FCA could respond; one such recommendation was that the FCA “build and adopt an AI-enabled agentic supervisory model.” However, AI is also becoming a regulatory tool beyond the financial services space.
As well as grappling with regulatory expectations around their own use of AI, companies may therefore be required to adapt to UK regulators’ increasing use of AI. Exactly how, if at all, this may change the supervisory relationship remains to be seen.
The Rise of Agentic Regulators
According to Rusu’s speech, the FCA wants to use AI to make its supervisory activities faster and more effective, including by using it to draw intelligence from disparate sources and flag early warning signs. One example of what this looks like in practice is building AI tools to oversee the 60,000 legal and accounting firms coming into the FCA’s anti-money laundering regulatory remit in 2028. Another is the FCA’s market first responder agent which will supervise around 9,000 small wholesale markets transaction reporting firms. The FCA’s Chief Executive, Nikhil Rathi, has described the latter agent as a way to speed up monitoring of wholesale markets, drawing on the FCA’s very large data sets (1 billion rows of data per day) to tackle market abuse faster. According to the recent speech, the aim is to free up supervisors to prioritise more complex supervisory activities. Earlier this year, the FCA also announced plans to use generative AI to review documents received from firms to support faster decisions.
The Mills Review recognises that the FCA will need to calibrate “supervisory tolerances, materiality thresholds and risk appetite” so that increased monitoring “does not inadvertently discourage innovation, competition or proportionate risk-taking.”
Beyond the FCA
The UK’s competition regulator, the Competition and Markets Authority (CMA), has also built data and AI capabilities. The CMA uses that technology in many areas of its work, including tools to detect suspected bid-rigging and breaches of UK consumer laws. For example, its in-house Bid Rigging Intelligence Tool (BRIT) scans public sector tender data for suspicious patterns that may indicate cartel activity. The CMA’s 2026–27 Annual Plan also refers to developing “customised agentic AI to detect consumer harms at scale” alongside an Internet Lab that simulates consumer journeys and captures evidence to support enforcement activity. The CMA has emphasised that its use of advanced technologies “complements, rather than replaces, established legal and economic analysis.”
Other UK regulators are moving in the same direction. For example, the Serious Fraud Office’s (SFO) 2026–27 Business Plan commits to launching a new AI roadmap, and states that “automation, AI and big data all provide ways to radically change [its] approach to intelligence analysis.” The SFO has said that it will invest in proactive intelligence, backed by £8.3m of additional funding.
Regulatory bodies therefore have an increasingly dual focus on AI: using AI to conduct supervisory activities more efficiently, whilst simultaneously conducting oversight over, and developing expectations for, use of AI by those they regulate.
AI Use by Businesses
The FCA wants the UK to be at the forefront of the safe adoption of AI in financial services, acknowledging that AI is moving faster than any regulatory framework was designed to handle. The FCA’s plan includes developing a framework for agentic finance covering agent identity, delegated authority, accountability, and execution.
The FCA has also published examples of good and poor practice for how firms deploy AI. Although not formal guidance, these examples are indicative of what supervision may look for. Further detail can be found in our prior GT Alert. However, the three key themes are as follows:
- Consumer centricity. For businesses directly and indirectly dealing with retail customers, the FCA expects AI to improve outcomes, not simply automate existing processes. Compliance with the Consumer Duty, with its focus on consumer outcomes, remains of central importance. However, the FCA has recently consulted on changes to the Duty to clarify, amongst other things, how it should apply where a firm’s role in a distribution chain is limited or indirect. This scope and the content of this reform may become important to firms that plan to use AI in their distribution chain.
- Accountability and governance. The FCA believes that too many firms are under-investing in oversight of AI. The FCA also expects senior managers to be personally accountable for AI systems deployed by their business. Adding governance after deployment is, in the FCA's words, “the wrong way around.”
- Operational resilience. The FCA expects firms to stress-test AI systems, plan for failure, maintain fallback options, and understand their dependencies, including reliance on a single AI model.
In the competition space, the CMA has also taken the position that any company deploying AI should verify that it tells its customers how AI is being used to serve them and train its AI to comply with consumer laws. The CMA also makes clear that the company is responsible for any breach of consumer law committed by the AI.
Key Takeaways
- Data-driven supervision. Supervisory contact may be prompted by automated analysis, putting a spotlight on the importance of regulatory data (for example, as part of notifications or returns).
- Consider engaging early. The FCA’s agentic finance framework, for example, is at an early stage. Firms developing agent-led products, particularly in payments and investments, may have an opportunity to help shape it.
- Engagement and alignment between regulators may continue as regulatory expectations evolve. The Digital Regulation Cooperation Forum recently published a joint statement from the FCA, CMA, Information Commission’s Office, and Ofcom confirming that organisations deploying AI agents must comply with existing legal and regulatory frameworks and stating that regulatory bodies “will not hesitate to use their powers” where needed.
- Agentic use by regulators is evolving in tandem with increasing use of agentic AI by those they regulate. Examples of good and poor practice are indicative of what supervisors may look for. Consumer outcomes, governance and testing, and assurance continue to be driving principles of responsible AI use across different sectors.