A quiet but profound debate is unfolding inside the United Kingdom’s financial system. While much of the public conversation around artificial intelligence has focused on chatbots, image generators and workplace automation, regulators are increasingly confronting a more consequential question: what happens when AI stops merely assisting humans and begins acting on their behalf in financial markets? The UK’s Financial Conduct Authority (FCA) believes that moment is approaching faster than many expect, and its response may help shape the future of finance not only in Britain but across global markets.
At the centre of the discussion is a concept now gaining traction throughout the industry known as “agentic finance.” Unlike traditional AI tools that analyse information and offer recommendations, agentic systems are designed to take actions. An agent could move money between accounts, rebalance investments, manage liquidity, execute trades, make payments, negotiate financial products, or coordinate increasingly complex financial decisions according to objectives established by the user. In theory, such systems promise efficiency, lower costs and greater personalisation. In practice, they challenge centuries-old assumptions about accountability, responsibility and trust in financial services.
The FCA’s recent comments suggest that regulators no longer view agentic finance as a distant possibility. Instead, they are beginning to construct the foundations of a supervisory framework before autonomous financial agents become commonplace. According to remarks by Jessica Rusu, the FCA’s chief data, information and intelligence officer, the regulator plans to work with industry participants, other regulators and frontier AI laboratories to develop a trusted framework covering agent identity, delegated authority, accountability and execution.
The significance of this effort should not be underestimated. Modern regulation is largely built around human decision-makers. Financial advisers provide recommendations. Portfolio managers make investment choices. Bank employees approve transactions. Compliance officers oversee conduct. Even where algorithms have been utilised extensively for years, there has generally been a human accountable somewhere in the process. Agentic finance introduces something fundamentally different: systems that may continuously make decisions, interact with other systems, and execute actions at machine speed while operating with increasing autonomy.
The FCA’s concern is not theoretical. Adoption of AI across financial services has accelerated dramatically. A joint survey conducted with the Bank of England found that 86% of firms are now using AI in some form. Consumer behaviour is changing just as rapidly. Research cited by the regulator found that four in five less-experienced investors aged between 18 and 40 have already used AI tools for investment assistance. Yet only 40% of consumers understand that they may have no right to redress if information from a general-purpose AI model proves inaccurate or financially harmful.
These figures reveal a widening gap between technological adoption and public understanding. Consumers are embracing AI tools with increasing confidence, but many remain unclear about where responsibility lies when things go wrong. If an AI agent misunderstands instructions and executes a catastrophic trade, who bears responsibility? The customer? The software provider? The financial institution? The model developer? These questions sit at the heart of the FCA’s emerging work and represent some of the most challenging policy issues facing financial regulators globally.
Interestingly, the FCA is not responding by creating an entirely new regulatory regime. Instead, the regulator has consistently argued that existing frameworks remain largely fit for purpose. Consumer Duty, governance requirements and the Senior Managers and Certification Regime continue to form the backbone of the UK’s supervisory approach. The emphasis is not on creating AI-specific legislation but on ensuring established principles apply regardless of whether decisions are made by humans, algorithms or autonomous agents.
This approach reflects a broader philosophy emerging within British financial regulation. Rather than attempting to write detailed technological rules that may become obsolete within months, the FCA is favouring principles-based regulation focused on outcomes. Firms remain responsible for consumer outcomes delivered through AI-enabled products and services. Senior managers remain accountable for risks within their areas of responsibility. Operational resilience obligations continue to apply whether an incident originates from human error or machine behaviour.
Yet maintaining existing rules does not mean regulators are standing still. The FCA is actively adapting its own supervisory capabilities to an AI-powered environment. The organisation has publicly described its ambition to become an “agentic regulator” and is developing AI tools to support its own oversight activities. Among these initiatives are systems designed to supervise tens of thousands of anti-money laundering firms expected to come under its remit and tools intended to monitor wholesale transaction reporting across financial markets. Following the launch of the FCA’s Handbook API, which allows firms to ingest machine-readable regulations, the regulator is also working toward machine-readable data collection systems.
This dual transformation is perhaps the most intriguing aspect of the story. Regulators are not merely preparing to oversee AI-powered institutions; they are increasingly becoming AI-enabled organisations themselves. In effect, a future may emerge in which autonomous financial agents interact with AI-enhanced regulators through machine-readable rules, continuous monitoring systems and automated reporting frameworks. The relationship between regulator and regulated entity could become far more dynamic than traditional models of supervision ever envisioned.
There are compelling arguments on both sides of the debate. Advocates believe agentic finance could democratise sophisticated financial services, reducing costs and improving outcomes for consumers who have traditionally lacked access to wealth management or professional financial advice. Automated agents could continuously monitor accounts, optimise savings, manage investments and respond to market conditions in ways previously available only to institutional investors and high-net-worth clients.
Sceptics, however, warn that increased autonomy may introduce new layers of systemic risk. AI models can behave unpredictably outside tested environments. Multiple agents interacting simultaneously could create unforeseen market dynamics. Concentration risk may emerge if large portions of the financial sector rely on the same underlying models, infrastructure providers or data sources. Accountability could become blurred precisely when clarity is needed most during periods of market stress. These concerns explain why the FCA has emphasised governance, operational resilience and incident reporting as central pillars of its approach. Firms will be expected to report significant AI failures and near misses, while agentic systems may require continuous assurance rather than one-time validation.
