Are you seeing these new technologies—like those used in loan origination and credit scoring—being fully integrated into traditional banks’ legacy processes, or are they still mostly happening as pilots on the side? What’s the real impact today?
David Ruis (Deloitte): It’s a mix. Integrating new technology directly into legacy systems is tough and requires significant change, so the most successful examples often start as side projects or pilots. These initiatives can then grow and be scaled up, but it’s important to communicate clearly from the start that this is a valid path for the whole organisation. Careful planning and change management are crucial for wider adoption.
Julien Defosse (AWS): I agree—it’s not easy. Most core loan data and general ledger systems are still deeply embedded in banks’ legacy infrastructure. What we often see is banks making “cosmetic” updates on the front end—creating something that looks modern, but without transforming the end-to-end loan journey. True transformation requires courage, because lending is a complex process, much more so than daily banking.
The best approach is to break down the lending journey into smaller, manageable sub-use cases, as David mentioned. Starting with side projects helps build the right foundations, especially around data. Automating document processing, for example, is a great place to begin, since there’s still a lot of paper involved in lending. By digitising these processes, banks can move faster and set themselves up for more comprehensive transformation down the line.
David Ruis (Deloitte): Another important factor is the role of intermediaries, especially in SME and larger-ticket lending. Right now, banks are mainly focused on using AI to improve their own processes, but they’re not yet thinking about how AI could change the way they interact with advisers or intermediaries. In the future, we could see AI agents taking on the role of advisers, directly interacting with customers to find the best offers. Banks will need to adapt their propositions to work with these new AI-driven interaction patterns, which is a key area to watch moving forward.