<\/p>\n
The review<\/a> (147-page \/ 2.4MB PDF), undertaken by FCA executive director Sheldon Mills, identified how AI will transform retail financial services by 2030 and beyond, from an industry that operates around \u201chuman-led, episodic financial activity\u201d to one where services are \u201cAI-enabled, continuous and delegated\u201d.<\/p>\n Developments with so-called agentic AI will drive this change, according to Mills who predicted that consumer journeys will become \u201cagent led\u201d, like what is already happening in the world of online shopping<\/a>. In time, he said, \u201ctrusted AI agents\u201d will \u201cact continuously for consumers within agreed limits, providing ongoing financial management and optimising people\u2019s financial lives\u201d.<\/p>\n Experts at Pinsent Masons said the review signals major changes that financial services firms will need to adapt quickly to.<\/p>\n David Heffron<\/a> of Pinsent Masons, who specialises in financial services regulation, said: \u201cThe review recommends a proactive FCA response: adapting the regulatory perimeter, building AI-enabled supervision and laying the foundations for AI systems that can increasingly support and act on consumers\u2019 financial decisions.\u201d<\/p>\n “The review recognises that AI adoption across financial services will create significant opportunities but also new risks. As firms increasingly deploy AI-enabled services, regulators and industry will need to ensure that innovation benefits a broad range of consumers, including those who are unable or unwilling to engage with AI-driven financial services,\u201d he added.<\/p>\n Angus McFadyen<\/a> of Pinsent Masons, who supports financial services firms with their technology contracts, said the review signals that \u201cthe risk checklist approach to third party technology risk management is over\u201d and said firms should further prepare for \u201cdeeper supply chain visibility\u201d.<\/p>\n \u201cTechnology with agentic capability requires an evolution of both BPO and ITO risk management controls, with the AI being treated as an extension of a firm\u2019s own operational structures,\u201d McFadyen said.<\/p>\n \u201cIn the past there was only a focus on outsourced functions; we are now moving to \u2018bill of materials\u2019 level, with visibility over more of the dependencies \u2013 such as model, infrastructure, and data.\u00a0Very few are setup to collate or understand this data at present,\u201d he added.<\/p>\n McFadyen highlighted how the FCA is due to publish more on good and poor practices it has observed from firms\u2019 use of AI later this year. He said this will provide firms with further insight, he said, into what regulators want to see from them in terms of how they manage AI-related risk.<\/p>\n Fintech specialist Luke Scanlon<\/a>\u00a0of Pinsent Masons said there are actions firms can take immediately in response to the evolving AI risk environment.<\/p>\n \u201cFor one, governance, model risk management and related controls should not be static and focused only on initial deployment,\u201d Scanlon said. \u201cThe focus must be on continuous monitoring acknowledging, in particular, the current concerns around the impact model drift can have on outcomes.\u201d<\/p>\n \u201cFirms need to be thoughtful about human oversight: some deployments will need approval before action, others close supervision while AI workflows are actually running, while others still can allow for review by exception in less risky circumstances. It is helpful that the FCA is acknowledging that the controls put in place here should be proportionate to the risk,\u201d he said.<\/p>\n \u201cThe FCA also raised concerns about third party dependencies and made it clear that it expects firms to manage not just models but third-party AI systems vendors and model providers. Data sovereignty and access to models were two specific issues raised in this context which firms will need to focus in on when negotiating with all AI-enabled third-party suppliers,\u201d Scanlon added.<\/p>\n In his review, Mills highlighted the opportunities that agentic AI in retail financial services offers, citing scope for firms to better personalise the products and services they offer consumers and for consumers to obtain better outcomes. One example detailed in the report described how agentic AI might enable consumers to easily switch between providers to the cheapest, most appropriate, products for them.<\/p>\n However, Mills also said the rise of agentic AI changes the nature of regulatory risk. He set out the profound implications it will have for the way firms reach and engage with consumers, handle complaints and govern risks. Those risks include around bias and fraud, as well as in relation to cybersecurity, operational resilience and financial stability. Mills said competitive dynamics in financial services markets will also change, with AI providers expected to play a much more prominent role in shaping how consumers access financial services and financial firms reach and serve those individuals.<\/p>\n Among his recommendations, Mills has urged the FCA to carry out a review of the \u2018regulatory perimeter\u2019, to identify potential regulatory gaps that could arise from the technological shift he anticipates.<\/p>\n He further called on the FCA to adopt an \u201cagentic supervisory model\u201d in what would represent a material change in the way it addresses AI-related risk and regulates the industry. Such a model, he said, is \u201cessential both for firm based supervisory efficiency, and for monitoring and detecting system-wide risks\u201d.<\/p>\n In practice, Mills said, operating an agentic supervisory model would mean the FCA deploying AI-enabled tools for authorisation, supervision, enforcement and administrative processes \u2013 to make application processing and real-time monitoring, among other tasks, more efficient. It would also mean deploying \u201csupervisory agents\u201d that could make regulatory processes more streamlined and responsive while \u201chuman supervisors remain in the loop and retain responsibility for judgement and decision-making\u201d. The model would also boost AI literacy within the FCA\u2019s own workforce and improve the regulator\u2019s system-wide oversight, because AI tools could help it \u201cdetect patterns and risks not visible at firm level\u201d, according to Mills, who said it would allow the regulator to support firms in adopting AI and build consumer trust in the technology.<\/p>\n Mills said: \u201cAI offers a once-in-a-generation chance to close the information asymmetries and frictions that have long left people making poor financial decisions. The right spur, in retail financial services, is ensuring consumers can make healthy ones; regulation, whether supportive or restrictive, should serve that outcome.\u201d<\/p>\n Pinsent Masons recently published a report that explores the implications for financial services firms of their suppliers\u2019 use of AI tools to deliver critical services. In it, Pinsent Masons described how firms must change the way they conduct supplier due diligence to account for use of agentic AI tools<\/a>.<\/p>\n