FCA and PRA Research note highlights increasing use of AI in financial services.

The FCA and the Bank of England are very active in the AI discussion. Way back in in October 2022 they jointly published a discussion paper (DP5/22) titled "Artificial Intelligence and Machine Learning". The paper was part of a wider programme of work including the AI Public Private Forum which released a report in February 2022 on the take up of AI. The October Discussion Paper led to a Feedback Statement (FS 2/23 and FS23/6) published on 26 October 2023. In Likely there will be more research and discussion opportunities as the technology becomes more prevalent. But there is unlikely to be any direct regulation of the use of AI in the short to medium term. The FCA has made clear that it sees itself as technology-agnostic and ill equipped to regulate technology. The FCA's focus is on how firms can safely and responsibly adopt AI technology, and understanding what impact innovations in AI will have on consumers and markets. This includes looking carefully at the systems, processes and controls firms have in place to ensure regulatory expectations are met.

On 10 October 2024 the PRA and the FCA published a research note on the uptake and use of AI in financial services. Its an interesting read and I'd recommend it. Below I have tried to summarise the research findings.

The use of artificial intelligence (AI) within UK financial services is on the rise, with 75% of surveyed firms already incorporating AI into their operations. The figure for those already using AI is significantly higher than that reported in the 2022 Bank of England and Financial Conduct Authority survey, which recorded 58% of firms using AI.

In the 2022 research, 14% of firms were planning to adopt AI within 3 years. In 2024 that figure dropped to 10%. You could argue that there is significance in the 4% difference in potential adopters. Perhaps it signifies that adoption is slowing, or that the firms that had the capability to adopt early have already done so. I offer no view.

As with any research note, its important to understand the sample population. For this research note regulators surveyed 118 firms. 50 UK banks were surveyed alongside19 insurance providers; 18 international banks; 13 investment and capital markets institutions; 11 non-bank lenders and 7 financial markets infrastructure firms.

As you will see there has been an acceleration of the adoption of AI with certain sectors within financial services being more bullish than others.

Adoption Patterns

Based on the survey results, international banks (90%) and insurance firms (90%) are far more likely to be currently using AI than any other cohort. Perhaps disappointingly, UK banks were second last in adoption (c. 65%) to infrastructure providers (c. 60%).

The highest proportion of AI use cases are in Operations and IT, followed by Retail Banking then General Insurance. Within the use cases respondents reported that 55% had some degree of automated-decision making. However, only 2% of AI uses had fully autonomous decision making ability.

Foundation Models Drive Rapid Adoption

Foundation models, those large-scale machine learning or deep learning models, now account for 17% of all AI use cases, underscoring their growing role in the industry. These advanced machine learning systems support anecdotal evidence that these particular models are being rapidly adopted across various sectors of financial services.

Increasing Third-Party Reliance

Outsourcing continues to expand, with third-party implementations comprising 33% of AI use cases, a notable increase from 17% in the 2022 survey. As model complexity increases and outsourcing and technology costs decrease, third-party dependencies are expected to grow. Notably, the top three third-party providers dominate the market, accounting for 73%, 44%, and 33% of cloud, model, and data services, respectively.

Materiality and Understanding of AI

While 62% of AI use cases are considered low materiality, 16% are rated as highly material, reflecting the varied impact of AI systems. Notably, 46% of firms report only a partial understanding of the AI technologies they employ, with third-party models cited as a significant factor. In contrast, 34% of firms claim a complete understanding, primarily when models are developed in-house.

Benefits Outweigh Risks

The most significant benefits of AI include enhanced data analytics, anti-money laundering (AML) capabilities, fraud prevention, and cybersecurity. Over the next three years, firms anticipate increased benefits in operational efficiency, productivity, and cost reduction. However, risks remain, particularly those associated with data—privacy, quality, security, and bias.

Looking ahead, the most concerning risks are likely to come from third-party dependencies, growing model complexity, and the proliferation of hidden models. Despite these challenges, firms project a 21% rise in perceived benefits from AI over the next three years, compared to a 9% rise in perceived risks. Cybersecurity remains the most critical systemic risk, with third-party dependencies expected to see the largest increase in risk during this period.

Constraints to AI Development

Regulatory constraints, particularly around data protection, resilience, cybersecurity, and the FCA’s Consumer Duty, are seen as the primary barriers to AI adoption. Non-regulatory challenges include ensuring the safety, security, and robustness of AI models, along with a shortage of skilled talent.

Governance and Accountability in Focus

Governance frameworks for AI are evolving, with 84% of firms reporting an accountable person for their AI initiatives. Most firms implement a combination of governance components, with over half employing nine or more. While 72% of firms attribute accountability to their executive leadership, roles are often shared among three or more accountable individuals or bodies.

Conclusion

As AI adoption accelerates, UK firms are balancing its transformative potential with the challenges of governance, risk management, and regulatory compliance, underscoring the need for robust frameworks to harness its benefits effectively.

Wherever your firm is on its AI adoption journey, RSM UK can support you in developing robust governance frameworks underpinned by effective controls that will ensure you draw benefit from AI whilst minimizing regulatory risk.