The United Kingdom’s Financial Conduct Authority (FCA) has outlined plans to integrate agentic artificial intelligence into its regulatory oversight workflows to monitor financial markets at scale. Detailed across a speech by the FCA on October 16, 2026, and a good-and-poor-practice publication released on October 19, 2026, the initiative marks a shift in how financial watchdogs oversee complex institutions. Rather than relying solely on manual reviews to police a sector where 86% of firms now use AI, according to a joint survey by the regulator and the Bank of England, the agency is building automated systems to triage market data, verify regulatory filings, and flag early warning signs for human supervisors.

The push comes as global spending on artificial intelligence is expected to exceed $2.6 trillion this year, according to remarks by Jessica Rusu, the FCA’s chief data, information, and intelligence officer, at Bloomberg’s UK Policy Series. By deploying its own automated tools and establishing technical interfaces for machine-readable reporting, the British regulator aims to manage an expanding supervisory workload while preparing the financial sector for autonomous software agents that execute payments and investments.

Inside the FCA’s Agentic Supervision Strategy

At the core of the regulator’s internal technology roadmap is the development of specialized AI tools designed to handle high-volume oversight tasks. According to the FCA, the agency is deploying automation and constructing a “market first responder agent” to supervise approximately 9,000 small wholesale markets transaction-reporting firms. The stated objective of this system is to automate routine monitoring and free human supervisors to focus on more complex investigations.

In addition to wholesale market monitoring, the watchdog is building AI oversight tools to prepare for a major expansion of its regulatory perimeter. Approximately 60,000 anti-money-laundering (AML) firms are expected to come under the FCA’s remit in 2028. To manage the massive influx of supervisory data, the agency plans to use AI to collect evidence, link intelligence across disparate sources, and identify early indicators of compliance failures.

The sheer volume of market data is a primary driver behind this automation push. During an October 13, 2026 appearance before the UK Treasury Committee, FCA chief executive Nikhil Rathi noted that the regulator receives roughly 1 billion rows of data each week through its market surveillance operations, according to TokenPost. An early application of agentic tools involves automatically checking regulatory returns and flagging discrepancies directly to the reporting firms.

However, accounts differ regarding what “agentic” supervision means in day-to-day practice and how autonomous these systems actually are. While the FCA’s October 16 speech highlights the creation of a “market first responder agent” to oversee 9,000 firms and new AI tools for the 60,000 incoming AML entities, coverage of the October 13 Treasury Committee hearing by Cryptopolitan and FStech indicates that AI currently assists staff by processing unstructured text across authorization, supervision, and enforcement workflows, with a human official remaining responsible for every decision. The public announcements describe AI as a support mechanism to gather and reconcile information rather than a transfer of final regulatory or enforcement authority to autonomous software.

Preparing for Agentic Finance and Machine-to-Machine Reporting

Beyond its internal supervisory upgrades, the regulator is addressing the rise of “agentic finance” within the commercial sector—a shift exemplified by recent industry developments such as the launch of Meta’s Muse and general-purpose large language models like ChatGPT reaching over a billion active users. Following the July publication of the FCA-commissioned Mills Review, which examined how AI could reshape retail financial services by 2030, the watchdog announced it is immediately embedding 6 of the review’s 7 recommendations into live regulatory work.

To address the technical and legal challenges of autonomous consumer and enterprise software, the FCA stated it will collaborate with financial institutions, fellow regulators, and frontier AI labs to build a framework for agentic finance. According to the FCA and reporting by Finance Magnates, this framework will focus on four critical pillars: agent identity, delegated authority, accountability, and execution. The regulator has not yet provided a delivery timetable for this framework.

In parallel, the agency is modernizing its technical infrastructure to support automated, machine-to-machine regulatory exchanges. The FCA launched a machine-readable Handbook API in August 2026 and is currently developing machine-readable data-collection APIs. These interfaces are designed to allow regulated firms’ compliance systems to interact directly with the regulator’s databases, streamlining how rules are ingested and how compliance data is submitted.

Compliance Expectations and Incident Reporting

For financial institutions deploying their own AI models, the regulatory messaging presents some nuance regarding whether new AI-specific obligations are being established. In its October 16 speech, the FCA stated that its existing rules requiring major incident reporting “must extend” to significant AI failures and near-misses, though the remarks did not establish a separate formal reporting rule or an implementation date. Meanwhile, trade coverage from Finance Magnates and MLex notes that the FCA continues to rely on its existing regulatory framework—such as the Consumer Duty guidance (FG22/5) and vulnerable consumer protections (FG21/1)—rather than issuing a dedicated AI rulebook.

The regulator’s October 19 practice guidance explicitly notes that its examples do not create new requirements or prescribe a single technical model. Instead, the document outlines practical expectations for firms deploying machine learning, generative AI, or agentic AI. Key areas emphasized include:

  • Governance and Accountability: Maintaining trained human oversight and clear escalation routes across the entire AI lifecycle.
  • Ongoing Monitoring: Utilizing performance dashboards and post-implementation monitoring to track model behavior in live environments.
  • Operational Resilience: Working with banks on frontier AI security to understand how advanced models affect cyber defense, vulnerability triaging, and system patching, especially in light of industry events like the Hugging Face security incident.

How the UK Approach Compares to US Banking Scrutiny

The FCA’s push to build its own agentic supervisory tools contrasts with the current posture of financial watchdogs in the United States, where public focus has centered on auditing how commercial banks control their internal AI deployments. According to an October 13, 2026 report by Reuters, key US banking regulators—including the Federal Reserve, the Office of the Comptroller of the Currency (OCC), and the Federal Deposit Insurance Corporation (FDIC)—have ramped up scrutiny of advanced AI across the banking sector due to mounting financial stability concerns.

Rather than announcing regulator-operated autonomous agents, the US agencies have been questioning banks directly about their internal safeguards. According to Reuters, US examiners are pressing financial institutions on whether software agents are being used in cyber defense, how the actions of those agents are restricted and monitored, whether automated decisions can be audited after the fact, whether human managers approve key actions, and whether reliable shutdown mechanisms exist if a system malfunctions.

Key Questions Surrounding Agentic Supervision

What specific thresholds will trigger human intervention in the FCA’s systems?

While the FCA has confirmed that its “market first responder agent” will monitor roughly 9,000 wholesale transaction-reporting firms and flag discrepancies in regulatory returns, the agency has not yet published the exact technical thresholds that will trigger an escalation to a human supervisor. Similarly, the regulator has not announced whether it will release evaluation metrics covering false positives, missed compliance risks, or model drift prior to scaling these tools to 60,000 AML firms in 2028.

How will regulated firms verify the authority of an FCA software agent?

As the FCA expands its machine-readable APIs and begins using automated agents to query firms about reporting discrepancies, institutions will need secure methods to verify that an incoming automated request is genuinely from the regulator. The FCA has identified agent identity and delegated authority as core components of its upcoming agentic finance framework, though specific technical protocols for dispute resolution and automated authentication remain under development.

Will AI near-miss reporting require new compliance workflows?

The FCA has made clear that it expects significant AI failures and near-misses to be captured under major incident reporting. However, because the agency has framed this as an extension of existing operational resilience rules rather than a standalone regulation, compliance teams are still awaiting granular guidance on the exact definitions, reporting channels, and deadlines that will apply specifically to algorithmic near-misses.

Sources

Read More