NatWest launches Venture Banking with AWS partnership - and the data architecture story behind it

Banks are very good at announcing AI ambitions. They are considerably less good at having the data infrastructure those ambitions require. NatWest knows this better than many.

Today the bank launched NatWest Venture Banking – a dedicated unit for high-growth, equity-backed businesses – alongside a new strategic partnership with Amazon Web Services (AWS). But the more interesting story predates both announcements by several years. Lisa Lewison, Head of Global Financial Services Partner Success at AWS, told me in an interview at AWS Summit London about the foundations that had to exist before any of this could. She explains:

“NatWest [is a] 300-plus year old bank, facing challenges not unique to banks in large global financial institutions. It has 20 million customers, roughly, needing a consistent and centralized means of storing customer data, so that they have a single view of their customers.

That problem – the absence of a unified customer view – is what determines whether any AI capability built on top of it actually works. A bank that can’t reliably answer the question “who is this customer, across all our products?” cannot meaningfully personalize at scale, resolve queries faster, or deploy agents with any confidence. The NatWest journey with AWS, alongside Accenture as implementation partner, is aimed squarely at solving that. The architecture is centered around a consolidated data mesh sitting on SageMaker Studio, designed to support AI capabilities that ultimately surface in customer-facing interactions – call center resolution, online product discovery, self-service updates.

There have been some detours along the way – Lewison is candid about the fact that the architecture has evolved. She notes:

NatWest would acknowledge that they changed the view of that architecture as they’ve gone through the journey. 

The catalyst was agentic AI — it arrived on the priority list faster than most institutions had planned for, and the roadmap had to move with it, ensuring NatWest could go large-scale with agents rather than being locked into choices made before that shift.

Governance up front, or stall later

From Lewison’s perspective, governance is not a compliance checkbox – it is a speed enabler. The analogy she draws is to the early cloud migration era, when the industry learned the hard way that customers who skipped foundational guardrails hit a wall. She elaborates:

Customers would hit a point where they would stall, because they then had to go back and effectively unpick what they’d done – rather than if that was established up front – then those controls are built in from the outset.

The same pattern is now playing out with AI adoption. AWS’s response has been to push foundational governance frameworks – including its Agent Core capability for governing agentic deployments – embedded from the outset rather than bolted on after the fact. This is particularly acute in financial services, where regulatory obligations like DORA (the EU’s Digital Operational Resilience Act) require institutions to demonstrate full lifecycle transparency over their technology decisions and permissioning structures.

On DORA specifically, Lewison pushes back on the idea that it represents a big change to how AWS engages with European financial institutions:

We’ve been working with our customers and the regulators for a number of years in preparation for DORA.

Financial services has always operated in a compliance-heavy environment; what DORA has changed is the specificity of the questions customers are bringing to AWS, seeking guidance on how to demonstrate compliance against requirements that are still, she notes, “not completely clear, and not in its final state.”

What most banks underestimate about their workforce

If there is one takeaway Lewison would impress on a bank CIO planning for the next 12 months, it is this: stop treating AI as a use-case-first problem, and start treating it as a people-first problem. She emphasizes:

If you wait for the use case to prove valuable, and then you put that through layers of approval, it’s a very slow way to get into the game. We’ve seen customers move faster if they put tools in everybody’s hands.

The evidence she points to comes from two customers at different ends of the enterprise spectrum. NatWest is building out its consolidated data architecture as the prerequisite layer. Allianz has taken a different approach and made what Lewison describes as an unusually broad commitment to workforce enablement, training approximately 16,000 developers in AI and generative AI tooling. Lewison notes:

To be able to transform at scale, making a commitment to enabling a workforce and enabling the people that are going to have to use the technology is critical.

London Stock Exchange Group (LSEG) offers a third reference point – one focused not on breadth of adoption but on the extreme performance requirements of market infrastructure. With data aggregated across more than 500 sources and low latency a non-negotiable, LSEG’s architecture incorporates AWS Outposts as part of the solution to those requirements. LSEG has publicly declared its intention to run all of its market data on AWS, though Lewison doesn’t go into detail on the specific workload breakdown.

Nationwide – same problem, earlier chapter

Nationwide Building Society gets a brief mention as another institution working through the same foundational challenge – consolidated customer data as a prerequisite for AI – but at an earlier stage than NatWest. “It’s a slightly earlier stage of development to NatWest, but very intentional about what they want to be able to do,” Lewison says, pointing to the same pattern of legacy system complexity and the strategic imperative to build a single customer view before AI capabilities can be meaningfully deployed on top.

Lewison also picks up on a research point that diginomica has tracked through its own data health work: financial services tends to be in a relatively stronger position than other sectors on data governance readiness. She attributes this to a combination of regulatory pressure and competitive disruption. She observes:

With the emergence of all of the challenger banks, it’s pushed some very old financial institutions to have to be able to compete with born-in-the-cloud providers like Monzo. 

That comparison to a neobank that has never run a mainframe is a strong one.

The conversation ends with Lewison’s observations of where most financial institutions actually are: good at running proof of concepts, less good at carrying them through to production. AWS’s AI Innovation Centre is positioned as a mechanism to close that gap – working with customers and partners over a concentrated engagement (as short as 45 days in some cases) to take validated use cases through to real-world deployment with dedicated specialist support from both AWS and the customer’s own team.

Skills are a compounding factor. AWS’s annual research into AI adoption across the UK and Europe, published during the London Summit event, consistently surfaces the same finding: where customers lack internal AI capability, partners are essential to deliver on their ambitions. Training, accreditation, and specialist resource – not just technology access – are what determine whether AI investments compound or stall.

My take

NatWest’s Venture Banking launch, and its partnership with AWS, is one visible output of a multi-year infrastructure investment that started with the painstaking work of consolidating data. The announcement is the easy part to report. Whether the architecture beneath it will hold up as agentic AI makes demands that weren’t anticipated when the journey started – is what will determine if this is a story about a bank that transformed, or one that learned expensive lessons about skipping steps.

Financial services has been living with the data consolidation problem longer than most sectors — regulatory pressure and challenger banks have been putting pressure on the issue for years. What Lewison’s examples make clear is that the institutions making real progress aren’t the ones with the boldest AI announcements. They’re the ones that did the painstaking work first. NatWest’s Venture Banking launch is a visible milestone — but the more interesting story is everything that had to happen before it could exist.