The end of September is a busy time for any self-respecting industry analyst, as event season gets into full swing. For me, it started at the AWS (Amazon Web Services) Financial Services Analyst Day in London, which ran under the banner of AWS’s symposium theme: Rethink Everything.

After a day of sessions with AWS, its customers and partners, including OpenAI and NVIDIA, I came away with three themes in mind, and one clear message for banks.

1. AI is playing a growing role in customer-facing processes

Much of the AI activity in banks to date has focused on the middle and back office. What stood out was how quickly those process improvements are now being turned into new and enhanced customer experiences. The two are increasingly inseparable: faster, more automated processes behind the scenes make better customer-facing services possible.

Alvin Huang from AWS, who works closely with many institutions on deploying agentic AI, framed the opportunity for the industry around four areas: accelerated time to insight, 24/7 personalisation, exhaustive coverage and application modernisation. The examples around personalisation really stood out. As we see in our research, UK customers routinely cite the performance of banks in this area as weak, and 66% are comfortable with the idea of institutions using data they hold to deliver more personalised services.

• Rocket Mortgage is using AI to review and approve the documents in a mortgage application, meaningfully shortening the process for customers.
• Lendi Group scans thousands of home loans daily and alerts customers in real time when a refinance makes sense, with the refinance itself completed in around ten minutes.
• Altruist’s Hazel creates a personalised tax plan in ten minutes rather than days, which has potentially important implications for current models of financial advice.

Equally notable was what AI is doing for customer understanding. Where it was once only feasible to review a small sample of customer conversations from the contact centre, typically 5–10% of calls, AI now makes it possible to analyse all of them. That has obvious value for audit and regulatory purposes. More interesting, though, is what it reveals about why customers get in touch, what they need and where service falls short, insight that can feed directly into better and more personalised service.

2. AI is changing the role of assisted channels

AWS described the impact of Amazon Connect Customer, its contact centre platform, is having in the market, with AI often delivering better outcomes than human agents for certain workflows.

We know from our data that this matters because the way customers use these channels is changing. As self-service extends into more activities, a growing proportion of the queries that reach branches and contact centres are increasingly complex. Our data reflects this, with satisfaction (measured by NPS) with the branch experience standing at +28 and contact centre at +19, compared to +42 for mobile (Source: Orion, n=4,000 UK consumers).

Zepz, the group behind WorldRemit and Sendwave, showed what AI can do in practice. Bringing AI into its contact centre has allowed it to resolve customer queries and problems more quickly, and to serve customers in a wide range of languages, which is a core need for a remittance business operating in more than 130 countries.

3. The real value sits in the data banks already hold

Arguably the most thought-provoking session came from NVIDIA, with Georgios Kolovos talking about their work on transaction foundation models. Applying large language models to financial data is harder than it seems, largely because of the wider range of fields in individual transaction records.

The spirit of the session was that the value in the data banks hold on their customers, especially transaction data, is there to be unlocked. NVIDIA expects a new wave of innovation to come from analysing event data – any interaction a customer has with their bank: a payment, a product enquiry, a branch visit, or simply checking a balance repeatedly. Each tells you something on its own. In combination, they say a great deal more. Featurespace, Nubank, Stripe, Adyen, Mastercard, Visa, Lloyds, Atom and Affirm were all referenced as among those already pursuing this approach.

Notably, none of this is being built alone. Partnership and openness to collaboration are central to AWS’s own success, and the same applies to the banks looking to bring new services to market more quickly. OpenAI, which launched its enterprise offering three years ago, is now scaling its work with banks across Europe on use cases spanning fraud prevention, data analysis, customer chat and credit scoring.