What does an AI-powered bank look like?
The race is on within the financial services industry to become one, even if no one is entirely sure what the end result will look like. So as we look to the future, how does a bank lead in the dark? One thing’s for sure – we can’t do it alone. Yes, our engineering teams have done amazing things, piloting agentic AI throughout the bank. Their success has only reinforced our desire to find partners who can help us move even faster - preferably, partners with flashlights.
For large enterprises, AI is much more than a technology. It requires, and produces, cultural and organizational changes. And while NatWest may seem to have an aggressive agenda, we don’t see that we have a choice: the technology is here, customers know what can be done, and if we don’t move quickly, our competitors certainly will. Prioritizing speed in a heavily regulated industry may seem counterintuitive, but for us, it’s the definition of boldness.
It’s also something that puts a lot of pressure on leadership at all levels. That means we need leadership wherever we can find it. We have very strong executive sponsorship for AI initiatives, and our CEO talks about the power and potential of AI consistently and persuasively. We also find leadership in our partners – not just in their technological prowess, but in their ambition and ability to catalyze change. And sometimes, we’ve chosen to import leadership, hiring an entire team of AI researchers whose expectations and ways of working are completely different from the rest of the bank.
The power of an early win
One of our first forays into using AI agents came in response to exploding volumes of work for teams that do customer onboarding. Those teams perform due diligence on every new customer. With financial institutions expected to face a 153% rise in fraud by 2030,1 due diligence was becoming more time consuming as staff had to manually ensure credit and background checks were in order. Generally, it was taking 48 hours to verify a customer.
We thought that autonomous orchestrated agents could conduct this due diligence, and that those agents could achieve the same accuracy as humans. And we thought we could do this ourselves. NatWest has an engineer-first strategy, believing that a good deal of the intellectual property of the bank is and will be produced by its engineering talent. We have a bias to build rather than buy. And we aim to have 70% of the code used within the bank written by our own developers. NatWest technology teams therefore built a series of agents that could verify a new customer in nine minutes, with no change in accuracy or in the assumption of risk.
That got everyone’s attention. If there had been an undercurrent of fear, it was replaced by excitement. The benefit of using AI agents was so large, and so obvious, that suddenly, people across the bank understood the opportunity in a new way. It was like the first time you drive an electric car – you’d probably heard about them for years, but as soon as you actually drive one, it suddenly makes sense to go out and buy one.
Prioritizing speed in a heavily regulated industry may seem counterintuitive, but for us, it’s the definition of boldness.
Finding scale with partnerships
The success of this agentic system was a point of pride for the team that developed it, and justifiably so. We pride ourselves on being pioneers, and The Kyndryl Readiness Report 2025 forecasts banks globally will start to use Agentic AI as they begin to understand the wide scale benefits. But no large enterprise can achieve transformation at scale on its own. For us, we chose to use outside partners and it’s a model other large enterprises can use to implement agentic AI quickly and effectively also. However, careful thought needs to be given to choosing a partner. In NatWest’s case, because of the level of difficulty, complexity, and uncertainty surrounding the bank’s AI transformation, the criteria for selecting a partner had to shift.
I’ve often chosen partners primarily for their functionality and technical abilities. In this case, we decided that those qualities, while still important, had to take a back seat to mutuality. There is a big difference between proof of concept and a scaled implementation across multiple geographies and business units, especially one that uses relatively new and untested technologies. When we choose a partner for a long-term project – one that’s building a new capability amidst rapid technological change – we need more than technical chops. Because things are going to go wrong. And when they do, you need to have a bit of slack in the relationship. You need to trust each other. You need to know that you and your partner have the same goals and ambitions.
This was a big factor in our choice of AWS as a partner to help modernize our data capabilities, creating a platform that uses AI to give us a single view of customer data across the bank. As a business, NatWest is very keen to lead in the use and deployment of AI, but we don’t always have the skills or capability to do that. AWS is very keen to have a financial services partner that can show the transformative power of adopting AWS’ AI technologies. So there’s a mutuality there. Our reasons for wanting to make this partnership work go beyond what’s written in our contracts.
Culturally, we were confident we would be comfortable with each other. And if we weren’t? We don’t hesitate to request a shift in the account team if the partnership isn’t working. We’re in this for the long haul, and we need partners who can make the journey with us.
We were also the first UK-based financial services company to collaborate strategically with OpenAI. Again, we’re not looking for partners who can give us a little nudge. We’re looking for transformation, quality, and speed. For us, the big opportunity to work with OpenAI was in product development. Our chatbots use models developed by OpenAI, and those models are going to be increasingly important to the bank. By collaborating closely with OpenAI, we can more quickly deliver excellent products and services that use those models. And the partnership gives us a seat at the table -- we want to know that when we use a product or service from OpenAI, it meets our governance standards and is fit-for-purpose for one of the largest banks in the UK.
Leadership from the inside out
For us to achieve our goals of being the first to deliver superior outcomes to our customers, we require deep technical expertise that most banks don’t have. So we found technology leadership by bringing in a whole team of people who are wildly competent in AI, and who can bring their ways of working to our existing teams.
In June, we hired a team of 30 AI specialists, including NatWest’s inaugural Chief AI Research Officer. The result is what you might expect: We now have this immensely capable team plonked in the middle of an organization that labors under a heavy layer of governance. This team has a very different perspective on the scale of the change we can make, and the pace at which we can do it.
We know we are walking a thin line between keeping this new team inspired and motivated, without being unduly frustrated by the pace of change. We want them to change the organization, and we know it’s a tall order. Our new Chief AI Research Officer is critical in this, constantly appearing on global townhalls, web casts, and anywhere else where she can champion the work of her team and describe what it means for us to change the way we work.
By partnering with AWS, OpenAI, and bringing in significant talent from other tech companies, we’ve gone deep. Now it’s time to go wide. Being an AI-powered bank means that all our employees must be able to use AI to augment their skills. To get there, we have a responsibility to help teach them to use AI, making upskilling a huge part of our overall growth agenda. If we can create a set of adaptable skills among our employees, that might allow front-line staff to spend more time with customers. It might enable our engineers to be more creative. It could change the types of products we build.
It will also require more from our leadership. AI and agents, as powerful as they are, won’t guarantee our success. We need to know how to use those technologies, and we need to know how to evaluate them. More than ever, our leaders will be judged on the quality of their critical thinking, and how they apply that to technology. Will they be able to appraise outcomes, and raise a red flag when an AI isn’t working? For the sake of our customers, our bank, and even our economies, we need to move forward with AI. But to make the best use of this new technology, we’ll need to combine that power with the things that make us most human.
Have questions or want to learn more?
