Key Takeaways
- The UK Financial Conduct Authority (FCA) is developing a framework for agentic finance that addresses agent identity, delegated authority, accountability, and execution.
- Jessica Rusu, the FCA’s Chief Data, Information and Intelligence Officer, stressed that innovation and consumer protection must be built together from the outset.
- Emerging risks of agentic AI include consumer harm, cyber vulnerabilities, operational failures, and the need for near‑miss reporting as a national‑security and regulatory priority.
- The FCA advocates continuous, conditional assurance because AI systems can behave differently at scale, under stress, or as underlying models evolve.
- PYMNTS Intelligence research shows U.S. financial‑services firms favor AI for structured back‑office tasks, while a growing share of American consumers rely on AI for financial decisions.
- The FCA’s recent guidance paper, “AI in financial services: good practices and areas for improvement,” offers concrete examples such as chatbots that flag vulnerable customers, independent ethics panels, and stress‑testing for operational resilience.
- Well‑deployed AI delivers value to both firms and consumers through better‑run operations, competitive pricing, and faster service.
Introduction and Context
The United Kingdom’s Financial Conduct Authority (FCA) announced plans to craft a comprehensive framework for agentic finance, a nascent field where artificial intelligence acts on behalf of users rather than merely assisting them. Speaking at Bloomberg’s UK Policy Series on Monday, Oct. 19, Jessica Rusu, the FCA’s chief data, information and intelligence officer, outlined the regulator’s ambition to shape the emerging technology in a way that balances innovation with consumer safeguards. “Our aim is to create the conditions in which agentic finance can thrive, with innovation and consumer protection built together from the start,” Rusu declared, setting the tone for a collaborative effort that will involve industry participants, other regulators, and frontier research labs.
Rusu’s Speech at Bloomberg UK Policy Series
During her address, Rusu elaborated on the scope of the forthcoming framework, noting that it will cover four core pillars: agent identity, delegated authority, accountability, and execution. She emphasized that the FCA will not work in isolation; instead, the authority intends to co‑design the rules with market players and supervisory peers to ensure they are both practical and forward‑looking. This approach mirrors the FCA’s broader strategy of fostering regulatory sandboxes and innovation hubs that allow firms to test new AI‑driven products under supervisory oversight while still protecting end‑users.
Risks Associated with Agentic AI
Rusu warned that as AI transitions from a supportive tool to an autonomous agent, new forms of risk emerge. These include potential consumer harm, heightened cyber vulnerabilities, and the possibility of operational failures that could cascade through financial markets. She pointed out that the traditional risk‑management toolkit may be insufficient when algorithms can make binding decisions without human oversight, underscoring the need for regulators to anticipate scenarios where AI behavior deviates from its intended design. “As artificial intelligence moves from helping users understand things to acting on the user’s behalf, there are new forms of risk,” Rusu noted, highlighting the urgency of pre‑emptive safeguards.
Incident and Near‑Miss Reporting Requirements
To mitigate these risks, the FCA intends to extend existing incident‑reporting obligations to cover significant AI failures and near‑misses. Rusu reminded listeners that current rules already compel firms to report major operational incidents, and the same rigor must apply to AI‑related events. She cited the UK’s AI minister, who has flagged near‑miss reporting for agentic AI as a national security priority, adding that it is likewise a regulatory priority for the FCA. By capturing both actual failures and close calls, regulators aim to build a richer data set that can inform preventive measures and improve system resilience.
Continuous, Conditional Assurance
Beyond reactive reporting, Rusu introduced the concept of continuous, conditional assurance as a cornerstone of the proposed framework. She argued that a system deemed safe at launch may exhibit different behavior when scaled, subjected to market stress, or when the underlying models evolve through retraining or data drift. Consequently, assurance cannot be a one‑time checkpoint; it must be ongoing and contingent on changing conditions. “It also means thinking differently about assurance,” Rusu said. “A system that is safe at deployment may behave differently at scale, under stress or as the models beneath it change. Agentic systems therefore require continuous, conditional assurance.” This perspective aligns with emerging industry practices such as model monitoring, automated drift detection, and periodic re‑validation.
PYMNTS Intelligence on AI Adoption in Finance
Supporting Rusu’s remarks, recent PYMNTS Intelligence research illustrates how financial institutions are currently harnessing AI. The report “Financial Service Pulls Ahead in the Enterprise AI Race” (May) found that U.S.-based financial services and insurance firms most frequently deploy AI in structured, auditable back‑office functions—including revenue recognition, credit risk assessment, and sales forecasting. In contrast, the share of firms using AI for customer‑facing tasks remains smaller, suggesting a cautious approach to consumer‑direct applications. This pattern reflects both the maturity of AI tools in operational contexts and the heightened scrutiny surrounding direct consumer interactions.
Consumer Reliance on AI for Financial Decisions
A complementary PYMNTS Intelligence study, “Consumers Stop Sampling AI and Start Relying on It” (February), revealed a shifting tide among American users. A growing proportion of both mainstream and power users of AI are now relying on the technology to make financial decisions, ranging from budgeting advice to investment selections. Rusu highlighted this trend during her speech, noting that as consumers place greater trust in algorithmic recommendations, the stakes for ensuring those algorithms are fair, transparent, and robust increase correspondingly. The findings underscore why the FCA’s focus on agentic finance must extend beyond institutional use cases to encompass the evolving consumer‑AI relationship.
FCA Guidance: “AI in Financial Services: Good Practices and Areas for Improvement”
On the same day as her remarks, the FCA released a publication titled “AI in financial services: good practices and areas for improvement.” Rusu pointed to this document as a practical toolkit for firms seeking to deploy AI responsibly. The guidance aggregates lessons learned from market participants and offers concrete examples of both effective and deficient practices. By providing a clear benchmark, the FCA aims to help firms navigate the complex terrain of AI governance while fostering innovation that does not compromise consumer safety.
Examples of Good Practice in Action
Within the guidance, Rusu cited several illustrative cases of firms putting consumers first. One example involved training a chatbot to detect signals that a consumer may need support from a human adviser, thereby escalating vulnerable cases before harm occurs. Another highlighted the establishment of an independent ethics panel tasked with reviewing proposed AI use cases prior to deployment, ensuring that ethical considerations are embedded early in the development lifecycle. A third example described firms stress‑testing AI systems under various market scenarios to gauge operational resilience and identify potential failure points before they manifest in production. These practices demonstrate how proactive governance can mitigate risks while unlocking the benefits of AI.
Benefits of Well‑Deployed AI for Firms and Consumers
Rusu concluded by emphasizing that the advantages of responsibly deployed AI extend beyond the technology itself. She noted that even AI implemented purely for internal efficiency can pass value to consumers through better‑run firms, more competitive pricing, and faster service. “Overall, our good and poor practice finds that both firms and consumers benefit from well‑deployed systems,” Rusu asserted, reinforcing the FCA’s stance that regulation should enable, rather than stifle, innovation when grounded in robust safeguards. This win‑win outlook underpins the regulator’s drive to create a framework where agentic finance can flourish without sacrificing trust or stability.
Conclusion and Call to Stay Informed
As the FCA moves forward with its agentic finance agenda, the collaboration with industry, other regulators, and research labs will be pivotal in shaping rules that are both effective and adaptable. Stakeholders are encouraged to monitor forthcoming consultations, contribute to the development of the framework, and leverage the FCA’s guidance to align their AI strategies with regulatory expectations. For ongoing coverage of AI developments in the financial sector, readers are invited to subscribe to the PYMNTS daily AI newsletter and stay abreast of the evolving landscape.
https://www.pymnts.com/news/artificial-intelligence/2026/united-kingdom-financial-conduct-authority-plans-regulatory-framework-agentic-finance/