What Makes Starling Bank Different?

Phil Hobden: Jason, welcome to the podcast!

I guess my first question has nothing to do with AI. I just want to talk about Starling for a moment because it’s such a fascinating example of a bank and a business in the UK.

I’ve worked in financial services, particularly in legacy banking. What has Starling done differently from all the traditional banks, especially from a technology perspective?

Jason Maude: Sure! From a technology perspective, what we’ve done differently is take a look at best-in-class technology practices—not just from the financial sector, but from the world in general.

We asked: How do people create software quickly while keeping it reliable? Because, as a bank, reliability is critical.

One of the key principles we realized in software development—and this applies especially to risk management in banking—is that the best way to minimize risk isn’t to release new code every three or six months. Instead, the best approach is to release small, frequent updates all the time.

At Starling Bank, we release dozens of software updates every day. We are constantly changing and improving our systems.

You might think, Doesn’t that increase the chance of something going wrong? The answer is yes, but—when something does go wrong, the impact is minimal.

It’s much easier to maintain a stable and reliable system when you’re dealing with small, manageable issues rather than one massive system-wide failure every few months. The latter is far harder to prevent and control.

The Dangers of Large-Scale System Failures

Phil Hobden: Yeah, and we’ve seen that before, right? A few years ago, one of the high street banks released a software update that completely crashed all their ATMs. People couldn’t withdraw cash, transactions weren’t going through—it was chaos.

On the scale of “worst things that can happen to a bank,” that has to be right up there.

Jason Maude: Exactly. One of the most telling things that came out of one of these major IT disasters—it may not have been the one you mentioned, but a similar one—was that, just days before the failure, someone involved in the project boasted about how they were about to release 2,500 man-years worth of work.

If I had heard that as an engineer, I would have been on the first plane out of the country before the fallout hit! That is an incredibly dangerous approach.

Releasing such a massive amount of code at once is risky for multiple reasons.

• First, you have sheer volume—thousands of changes happening at the same time, any of which could cause issues.
• Second, you risk conflicting changes—two separate updates might work fine on their own but break when combined.

The bigger the release, the harder it is to test every possible scenario where things could go wrong.

This is why big releases are a problem. Doing lots of small, incremental updates is a much safer approach—it keeps the risk profile under control.

The Hidden Complexity of Banking Systems

Phil Hobden: Yeah, it’s such an interesting way of looking at it. And you’re completely right.

I once worked at a bank where we tried to remove a single legacy system—but we didn’t realize just how many other systems depended on it.

When we took it out, everything that relied on it collapsed. We had no idea how deeply intertwined everything was until it was too late. It was one of those “Oh, damn” moments.

Of course, another thing that makes Starling interesting—and we’ll move on to AI in banking soon, but I really want to explore this—is that you built your banking platform from scratch.

Not only that, but you built it as a platform that could be white-labeled and repackaged for different territories and countries, adapting to different regulations.

That’s a really innovative approach to banking. What was the thought process behind that?

Building a Bank from Scratch & White-Labeling

Jason Maude: Yeah, absolutely. It was an evolving process.

We didn’t start out thinking about world domination. Our initial goal was simple: Let’s make banking better in the UK.

To do that, we needed to build our own banking platform—one that would give us the flexibility to innovate.

A lot of the major changes that Starling has helped drive in the banking industry wouldn’t have been possible if we were relying on someone else’s platform.

Why? Because when you use someone else’s system, you’re limited by their rules and constraints.

Before digital banks like Starling came along, the idea of giving customers more control and insights into their finances simply wasn’t a thing.

For example:

• Freezing your card instead of canceling it.
• Dividing your account into separate pots for budgeting.
• Real-time spending notifications and insights.

These features—now copied by high street banks—weren’t available before.

If we had built Starling using someone else’s legacy banking platform, we wouldn’t have been able to develop these features as quickly—or at all. We needed to own our technology.

Once we built it, we thought: Wouldn’t other banks also benefit from this?

That’s where Engine comes in—our white-label banking platform. We’ve now started offering it to other financial institutions who need a modern, flexible banking system.

Expanding into Other Markets Without the Red Tape

Phil Hobden: And in a way, that’s a really smart strategy.

Expanding into different countries comes with huge regulatory challenges.

We’ve seen banks like Revolut struggle for years to get banking licenses in different countries. You can’t just say, “Let’s launch a bank in Portugal!” It’s incredibly complex.

You need to:

• Understand local regulations
• Work with local regulators
• Build trust within that market

That’s why the approach of selling your platform to local institutions makes so much sense. They already have regulatory approval, so you get the benefits of scaling globally without the nightmare of securing licenses everywhere.

And let’s be honest—dealing with regulators is never the fun part of launching a business!

Jason Maude: Yeah, exactly.

Navigating regulations is always a challenge, but it’s necessary.

By partnering with existing banks that already understand their local regulations, we can provide the technology while they handle the compliance side of things.

At the end of the day, the fundamental needs of banks are the same across different countries:

• They need a resilient, reliable banking platform.
• They need a system that never misses a payment.
• They need a system that can recover quickly from failures.

That’s what we provide.

Pragmatism & AI in Banking

Phil Hobden: And I guess that pragmatism brings us to AI. How does that same approach factor into how Starling looks at AI in banking?

What role is AI playing in Starling’s journey today, and how do you see it evolving moving forward?

Jason Maude: Absolutely. AI is a fascinating tool with many different applications.

But I use the word tool deliberately. We need to constantly remind ourselves that AI isn’t magic—it’s a tool, just like any other. It’s a complex tool, a difficult tool, more like an industrial lathe than a hammer, but a tool nonetheless.

To get value from AI, you need to focus on solving real problems rather than simply adopting it for the sake of it. Instead of asking, How do we use AI?, the real question should be, What problem are we trying to solve, and is AI the right tool for it?

And in many cases, the answer is no.

The Reality of AI Adoption in Business

Phil Hobden: Yeah, I think that’s an important perspective. AI isn’t a magic fix for every business challenge.

In accounting, for example, firms are excited about AI, but many are still using legacy desktop software with on-premise servers. AI isn’t going to help much in those cases, because the foundations aren’t there for it to work effectively.

I guess when people think about AI in financial services, their first concern is usually risk.

Let’s be honest—historically, financial services in the UK haven’t always covered themselves in glory, especially when you think about the 2008 financial crisis.

Even today, there’s still a lot of skepticism about how financial services operate. So, when it comes to AI, what safeguards and parameters need to be in place to ensure it’s used responsibly?

And do you think safeguards are even necessary?

The Importance of AI Safeguards in Financial Services

Jason Maude: Undoubtedly.

Any powerful tool requires safeguards to ensure it’s used responsibly—and AI is no exception.

Many of the safeguards needed for AI are actually the same ones we already use for software in general.

• You need to know what’s being released, when it’s being changed, and why
• You need proper review processes
• You need to ensure your data is high quality and meets GDPR and other regulatory standards

All of these good software development practices must also apply to AI.

But here’s the thing… That alone will only get you most of the way there.

So, what else is needed to ensure AI is truly safe and reliable in financial services?

Tune in to find out.

AI at Starling Bank: Where It’s Actually Used

Phil Hobden: That’s really interesting. So, let’s talk about AI at Starling.

What are some of the areas where AI is actually being implemented? I think most people assume customer service and chatbots are the first places where banks would use AI—has Starling explored that?

Jason Maude: Actually, no.

I think that’s a false assumption that many businesses are making. AI isn’t always the right tool for customer interactions.

At Starling, we’ve chosen to keep human customer service at the heart of our operations. That’s why we have a 24/7 call centre staffed by real people—so that when customers need support, they can talk to someone who understands their concerns and can help efficiently.

Instead of focusing on AI for customer support, we see bigger opportunities in areas where AI can enhance our systems rather than replace human interaction.

The Real AI Challenges: Data and Fragmentation

Phil Hobden: That makes a lot of sense.

One of the big challenges we see in accounting is data fragmentation—multiple disconnected systems that don’t talk to each other.

How critical is data standardization when it comes to AI adoption in financial services?

Jason Maude: I’d say it’s critical, but I’d go further than that—it’s critical even if you’re not using AI.

The financial industry has major issues with data fragmentation. Many traditional banks have disconnected systems due to mergers, acquisitions, or legacy technology. This creates a situation where:

• The same customer might be stored in multiple systems
• Different databases might contain conflicting customer data
• Banks aren’t always sure which system holds the most accurate information

At Starling, we don’t have this problem because we built our platform from scratch. Our system ensures data is consistent, easily transferable, and there’s always a golden source of truth.

If a bank wants to implement AI without solving its data issues first, it’s going to be a nightmare. AI models need clean, structured data, or they’ll generate unreliable outputs.

AI: A Useful Tool, Not a Magic Fix

Phil Hobden: I think what I’m taking away from this conversation is that for Starling, AI is a tool to enhance banking, but it’s not the driving force behind everything you do.

Jason Maude: Exactly.

We’re not an AI company—we’re a bank. We use AI where it makes sense, just like we use any other technology.

Some businesses position themselves as AI-first, but I’d be cautious about that. In many cases, AI is just one part of a much bigger system. If a bank’s entire identity is built around AI, that might mean their core systems aren’t as strong as they should be.

Will AI Transform Banking More Than Open Banking?

Phil Hobden: Would you say AI will be more significant in financial services than open banking?

Jason Maude: AI will definitely have a big impact, but I wouldn’t say it’s transformational in the way some people expect.

AI is evolutionary rather than revolutionary. It will help banks improve efficiencies, detect fraud, enhance customer experiences, and develop smarter risk models. But it’s not going to completely redefine banking overnight.

Open banking had the potential to be more disruptive, but it hasn’t fully delivered on its promise yet. Initially, it was meant to increase competition by reducing the market share of major banks, but that hasn’t really happened. Now, it’s shifting toward becoming a new payment rail, but there are still big challenges—like fraud protection and compliance.

The Future of AI in Financial Services

Phil Hobden: Alright, final question—get your crystal ball out. What do you see as the long-term trends for AI in financial services?

Jason Maude: As the saying goes, predictions are difficult, especially about the future.

That being said, AI will continue to evolve and be used across banking in fraud detection, automation, credit risk modelling, and operational efficiencies.

For some banks, AI will be built in-house, but for most, it will be provided by specialist technology vendors. AI will become a standard tool, just like cloud computing and automation have.

But ultimately, AI will not replace traditional banking—it will simply be another tool in the toolbox.

Phil Hobden: Jason, this has been a fascinating conversation. Thank you so much for your time.Jason Maude: Thank you for having me.