The Bank of England (BoE) and the Financial Conduct Authority (FCA) have jointly released a survey report titled "Artificial Intelligence in UK Financial Services - 2024." This document provides an in-depth analysis of the current state and future projections of artificial intelligence (AI) and machine learning within the UK financial sector. The report builds upon previous surveys conducted in 2019 and 2022, aiming to offer continuous insight into the adoption, implementation, and implications of AI technologies by firms regulated by the BoE and/or FCA. Here we look at the findings through a consumer finance lens.
Key findings on AI adoption and use
The 2024 survey indicates a notable increase in AI utilisation among financial service providers, with 75% of firms currently employing AI technologies, up from 58% in 2022. Additionally, a further 10% of firms plan to adopt AI within the next three years.
Benefits and risks
Firms report significant benefits from AI usage. Some of the largest perceived benefits include in combatting fraud, customer engagement and services, and customer care. The survey indicates that respondents expect that these benefits will continue to grow over the next three years. Responsible AI use in these areas could assist firms in complying with their requirements under Consumer Duty. For example, by deploying AI models to help to provide tailored support for customers. Additionally, leveraging AI to combat fraud could be invaluable in light of the mandatory reimbursement rules for authorised push payment scams that have recently been introduced.
However, the adoption of AI is not without its challenges. Some of the top perceived risks of using AI include data bias and representativeness, explainability, and bad consumer outcomes. Respondents expect each of these risks to increase over the next three years. These risks reflect challenges to the fair treatment of consumers, a core aspect of the Consumer Duty and firms will have to carefully consider how they implement AI.
Implementation and governance
The report underscores the importance of governance and accountability in AI applications, with 84% of firms having an identifiable accountable person for their AI frameworks. Despite this, 46% of firms have only a 'partial understanding' of the AI technologies they use, primarily due to the reliance on third-party models. The survey reflects an increase in reliance on third-party implementations, which constitute a third of all AI use cases, marking an increase from 17% in 2022.
To ensure compliance with outcome focused regulation, such as the Consumer Duty, it is vital that firms adopt robust AI frameworks which provide for regular monitoring and benchmarking of AI models, particularly when models are used in relation to retail customers. This should include policies for contracting with third party suppliers of AI models to ensure an appropriate level of transparency of the underlying model and how it operates.
Transparency and Explainability
The survey reveals a high adoption of explainability methods among firms using AI. These are methods that enable humans to understand the underlying logic and data that informs an AI model's behaviour. This is crucial in the context of the Consumer Duty, which promotes transparency and comprehensibility in firms' dealings with consumers, particularly with regard to the consumer understanding outcome. By employing explainable AI, firms can offer consumers clearer insights into how decisions affecting them are made, enhancing transparency and trust. However, respondents still view issues relating to explainability and transparency as one of the largest constraints on adoption of AI.
Constraints on AI adoption
Some of the most significant perceived non-regulatory constraints included:
- Safety, security and robustness;
- Appropriate transparency and explainability; and
- Bias and fairness in decision making.
These are considerable constraints not only for producing effective AI models but also to ensure fair outcomes for consumers. These must be overcome in order to comply with the Consumer Duty and firms should consider these risks appropriately before deploying AI models in relation to consumer finance.
Consequently, it is no surprise that the Consumer Duty was the third most identified regulatory constraint to the adoption of AI after data protection and privacy rules, and resilience, cybersecurity and third-parties rules. 13% of firms considered that there was a lack of clarity in relation to the Consumer Duty and 23% considered the Consumer Duty a high regulatory burden.
Conclusion
It is clear that if firms are able to deploy AI safely, there could be significant benefit to consumers. Deployment of AI models may assist firms in complying with their requirements under the Consumer Duty and help to create a safer, more tailored and engaging environment for consumers. However, this survey reflects that as the perceived benefits grow, so do the perceived risks. It is vital that firms consider these and implement AI compliance policies and outcomes monitoring to ensure that they do not fall short of the Consumer Duty and other relevant regulations.
