It was great to be with AWS last week for their latest EMEA FSI Analyst Day, alongside my Banking teammates Zil Bareisis and Daniel Mayo, and several other colleagues from Celent and GlobalData. As ever, the event was rooted in the principle of ‘show me, don’t tell me’, with the emphasis squarely on letting clients tell their own stories about the ways their partnerships with AWS are helping them to unlock business value.
The sessions and side conversations covered a multitude of themes, but there were three messages that stuck out.
The first was a reminder that any use case for AI is only as good as the data foundation it rests upon. This theme ran through many of the sessions but was best articulated by Karen Dewar, Chief Data and Analytics Officer at NatWest. The bank recently went public with a significant investment it is making with AWS and Accenture to simplify and modernise large parts of its data estate. A key pillar of the initiative is to reduce the number of systems across the bank containing customer data and records from approaching 30 down to one. The bank expects this will significantly enhance its ability to personalise products and services, as well as improving a range of processes.
This transformation of the bank’s data architecture is also expected to catalyse its broader adoption of AI technologies across a range of use cases. For many banks, moving GenAI and Agentic AI projects into production has been a significant challenge, and data infrastructure is a common hurdle (although not the only one). To illustrate, NatWest shared that it has several hundred GenAI or Agentic pilot projects underway with a far smaller number running at scale. This investment is expected to support more of these along the development curve.
This project is also a great example of how the adoption of AI technologies is driving a new wave of legacy modernisation across the industry. The initial hype around GenAI and Agentic is giving way to a deepening understanding of the reality that these models are only as good as the data they can access or work with. Scaling usually requires easy access to the information and business logic sitting in applications potentially decades-old, and not designed to support these use cases, making modernisation a necessary step. Helpfully, there are also many examples that demonstrate how recent advances in GenAI and Agentic can help to accelerate these projects.
The second big message was the adoption of Agentic AI across the industry. As we look forward to 2026, it seems clear that we will see a growing number of institutions move their test and pilot work around Agentic AI towards production. AWS shared many examples of financial institutions and other players in the broader ecosystem that are already live or at advanced stages with the technology and using it to unlock significant benefits from automating (or bringing greater automation into) a range of workflows. These included improving customer onboarding processes (which can be a time consuming and costly process in many banks) and using agents to identify and attempt to retain mortgage customers potentially at risk of churning as promotional rates expire.
The final area was a preview of the AWS European Sovereign Cloud (ESC). Due to launch later this year, the ESC is being created to directly address concerns around data sovereignty in Europe. This is an issue that has grown in prominence over recent years, and yet there can be very different views around what the notion of data sovereignty means in practice.
AWS consulted widely across the industry to understand the common requirements with respect to data sovereignty. This forms the basis of its Sovereign Requirements Framework (SRF), which is a comprehensive set of technical, legal, and operational sovereignty controls which are reflected in the ESC. These go further than the “privacy by design” principles baked into the existing AWS regions and effectively create a European-only infrastructure.
The ESC will be run under a new legal entity created in Germany and will be managed entirely separately from the rest of AWS, meaning it would be unaffected by any geopolitical or other interruptions. Key here is that the infrastructure is fully independent of any systems outside the EU and has its own separate support organisation too (also fully EU-based). In other words, there is no technical route for any interference with the ESC from outside the EU.
Given the changeable nature of today’s geopolitics, a fully European cloud option may appeal to many in the industry. This is particularly important in the case of highly sensitive workloads and may become a feature of future multi-cloud strategies. It may also help entice some institutions to move more of their more sensitive data from on premise environments.
How to sum up? It’s clear from this event, and our other recent interactions across the industry that 2026 will see an acceleration in the scaling of use cases and initiatives driven by GenAI and Agentic. Underpinning all of this will be the deepening of investment in both legacy and data infrastructure modernisation. Indeed, the growing use of AI technologies to release more data from legacy applications is already creating a proverbial flywheel effect that will further accelerate the further adoption of GenAI and Agentic across the industry.
