Fintech Turns to AI as Alison Kay Urges Secure Adoption

Fintech firms and banks are moving artificial intelligence from trials into daily operations, and AWS executive Alison Kay says secure scaling will shape the next phase. At the AWS Financial Services Symposium in London, Kay discussed how financial organizations can use AI while maintaining security and governance.

Examples span financial crime screening, customer service and internal data work. London Stock Exchange Group is rethinking financial crime screening with Amazon Bedrock, while NatWest is using AWS technology to transform customer experiences. The projects reflect a broader push to apply AI to specific financial workflows rather than treat it as a standalone experiment.

Adoption is rising, with data and governance in focus

AWS-commissioned research found that 64% of UK organizations use AI, up from 52% a year earlier. AWS also says one business adopts AI every 40 seconds. Among financial-services AI adopters, 38% were at an advanced stage, compared with 24% of UK businesses overall. The research reported gains for both advanced and basic adopters: 72% for the advanced group and 42% for basic users.

Kay said high-quality data held by financial firms can support faster decisions, risk assessment and personalization. That potential depends on firms applying security and governance controls as they expand use. For leaders weighing similar deployments, wealth firms’ AI partnerships offer a related look at how the sector is approaching adoption.

Several examples illustrate different operating models. Nationwide is modernizing on AWS for 17 million customers, and its Call Checker uses Amazon Connect. Nationwide says impersonation scams accounted for 17% of reports across 2024 and 2025. At TP ICAP, an AI Factory combines an AI and Innovation Lab with a governance framework and a hub-and-spoke operating model; agents are in production.

Scaling agents requires clear limits

Experian’s ProductMate was built by a 10-person team on Amazon Bedrock and deployed domain-specific agents across thousands of users in six regions. The system includes governance, identity controls, memory and observability. Those controls matter as organizations move beyond isolated tools and give agents defined tasks in business systems.

Kay’s guidance is to build security and governance in from the start, secure system data and limit each agent’s access to the scope of its task. She forecasts that within 12–18 months, agents will account for most queries to data systems. That is her outlook, not a reported outcome, and it underscores why access boundaries and monitoring are part of deployment planning.

Competitive pressure is also driving adoption, Kay said. Modernization remains part of that challenge: AWS Transform is intended to update older systems. Capgemini reported that 85 of the top 100 banks use mainframes for mission-critical work and that mainframes power nearly 90% of global transactions. AWS research also found that almost half of UK organizations cite skills shortages as their biggest AI challenge.

Financial firms are testing AI across screening, customer experiences and internal operations, while building governance into the way these systems work. The discussion around agentic AI for finance workflows shows another strand of that shift. As firms define what agents can access and how their actions are observed, questions about agents exceeding instructions remain relevant to the controls they choose.

For Kay, the opportunity rests on combining useful data with safeguards that let financial organizations expand AI use responsibly. The examples show varied stages of adoption, from modernization plans to agents already in production, with governance and task-specific access central to the approach.

Read the original interview with Alison Kay.