AI in financial services, so what’s the latest?

The latest EVIDENT AI OUTCOMES report just hit the inbox. So, what's the most recent news from the select 50 leading banks tracked by Evident?

1. AI use cases have surged “fueled by expanding deployments of GenAI tools”;
2. Outcomes reporting is still very limited; 
3. Reported outcomes are noted in more diverse areas, including revenue, sales, cross-selling and customer acquisition;
4. Most use cases still relate to internal functions;
5. Externally-facing use cases are concentrated in retail banking, where there is evidence of upgraded chatbots, virtual avatars and experimentation with agentic AI;
6. 40 out of 50 tracked banks have deployed AI assistants to their workforce but only 10 have disclosed any detail about their deployment;
7. Most of the reported use cases are in “Front Office and IT & Security”; and
8. Nine of the tracked banks are reporting agentic AI use “in production or in pilot stage”.

There are many challenges around reporting, the reasons for which are well put in this quote from the report: “Much of this comes down to the lack of precision in aggregating the overarching benefits of each discrete AI initiative, while disarticulating the underlying cost". Breaking this down further, the report notes the following four key issues:

1. The improvements delivered by GenAI are hard to measure in isolation;
2. It is challenging to translate improvements into monetary value;
3. There is a time-lag between the upfront investment required and the accrual or realisation of benefits; and
4. The cost of AI is hidden in wider and interconnected technology overheads.

Turning to the latest from the FCA, its recent output on AI Live Testing: The use of AI in financial markets - from promise to practice reminds us of the AI Live Testing initiative, which is currently still open and taking applications until 20 August. The FCA emphasises that the initiative is intended to be a partnership with industry firms, and “a practical, hands-on way to build trust, reduce risk, and accelerate safe and responsible innovation”, enabling all parties to:

1. Understand how AI performs in practice and in live market conditions;
2. Identify and adapt to risks;
3. Explore market-specific assurance methods; and
4. Share insights that will future-proof the industry's approach to AI.

Significantly for Live Testing participants, the initiative offers serious advantages to tap into what the FCA is really looking for in AI deployment, develop confidence in the AI that is subject to test in safe but real conditions, and cover-off all the fundamentals (intended outcomes, key risks, mitigations, and risk assessment etc.) with regulatory expertise on-hand. This is where the current “guesswork” that many might be feeling around what the regulatory position on AI is or might look like, can be clarified, not just for the lucky few Live Testing participants, but also for the rest of the industry, which will be hotly anticipating the delivery of detailed and informative insights. On this note, feedback from the FCA is expected in September.  

You can read more updates like this one about the AI journey in financial services by clicking the hyperlink, and you can subscribe to our monthly financial services regulation update by clicking here, clicking here for our AI blog, and here for our AI newsletter. 

If you would like to discuss how current or future regulations impact what you do with AI, please contact me, Tom Whittaker, or Martin Cook. You can meet our financial services experts here and our technology experts here.

68% of use cases with reported outcomes are concentrated in just two areas (Front Office and IT & Security) – functions where sales advisor assist, developer augmentation, and fraud prevention applications are starting to yield clear benefits for banks

https://evidentinsights.com/reports/evident-ai-outcomes-report-2025-excerpt?id=37ce42fa99