EP1015: Modernising core banking - the SBS approach

You know, when you open your banking app, um, maybe you're trying to send money to a friend or just check the clearance on a deposit. Right. And the app just hangs there. Aaron Powell Exactly.

You're staring at this interface that feels like it was, I don't know, designed in 1998. And you just wonder why this multi-billion dollar institution feels so ancient. Aaron Powell Well, the answer to that is actually pretty terrifying. Yeah.

Because doing what they call a brain transplant on a bank's foundational software, it usually crashes the entire system. Aaron Powell Which is obviously a massive problem. Yeah. But um for today's deep dive, we actually found a blueprint from a French bank that figured out a sort of cheat code.

Yeah, a way to upgrade a 40-year-old system without rewriting a single line of its core code. Right. So whether you're managing software at work or you're just wondering why your bank app is so clunky, this is a fascinating look at innovation. It really is.

And we're pulling the architecture of this blueprint from uh a really insightful February 2026 interview in the IBSI FinTech Journal. Oh, right. Robin Amlott's interview. Exactly.

He sat down with Jean-Charles Duck, the general manager of SBS Core SAB, to dissect what is effectively a masterclass in high-stakes technological surgery. And I mean, the insights here apply way beyond banking. But let's start with the scale of the legacy problem in finance because it's staggering. It's huge because every time you try to do anything financial today, you expect seamless digitalization.

Trevor Burrus, Jr. Right. You want instant payments, real-time updates, flawless interfaces on your phone. Yeah.

And the institutions providing these services are under this crushing pressure to deliver that modern experience. Trevor Burrus, Jr. Well, running on software that was, you know, written before the internet was even a household concept. Exactly.

But consumer demand is really only half the vice grip here. The other half is regulatory compliance. Oh, sure. The rules are always changing.

Constantly. How money moves, how it's tracked, how it gets reported to authorities. Institutions are forced to adapt, but doing a full core replacement, meaning tearing out the foundational ledger that runs every single calculation in the bank, it's a notorious nightmare. It takes years.

Yeah. Years. It bleeds budgets dry. And it carries an unacceptably high risk of just catastrophic failure.

Aaron Ross Powell Okay. And taking a look at the specific environment of the French banking sector makes this even more daunting, I think. Aaron Powell Oh, absolutely. Because France has these highly mature core systems paired with incredibly strict regulatory constraints.

You're not dealing with some loose experimental sandbox. Aaron Powell No, not at all. A full core replacement in that mature of a market is considered inherently dangerous. Aaron Powell You're essentially trying to replace the foundation of a skyscraper while all the tenants are still living and working inside.

That is a perfect way to put it. One wrong move, and the entire structure collapses. Aaron Powell Right. Which brings us to France Bank Francis A.

This is the institution at the center of the SBS project we're looking at. Right. Established back in 1984 by France Bank Group and Credit Agricole. Trevor Burrus, though BPCE Group replaced Credit Agricole in 2007, right?

Yes, exactly. So their current ownership structure is 79% FranziBank and 21% BPCE International. Aaron Powell Okay, so a solid backing. But their core business isn't it's not issuing checking accounts to college students.

Aaron Powell No, their primary function is financing international trade. They act as the main entry point in Europe for a vast network of correspondent banks. Aaron Powell Right. And okay, let's unpack this for a second because international trade finance sounds kind of abstract.

Does, yeah. Until you realize it's literally the plumbing of the global supply chain. If you're a bank handling international trade, your core system is basically a commercial jet engine mid-flight. Exactly.

We're talking letters of credit, customs clearances, massive shipping manifests, multi-currency conversions, all having to clear simultaneously. And if a consumer banking app goes down, you know, you can't buy a coffee. Annoying, but fine. Right.

But if Franza Bank's core system goes down, cargo ships literally sit idle at ports. Supply chains halt. And the financial penalties run into the millions very, very quickly. The margin for error is absolutely zero.

Because a single glitch shatters international trust. Precisely. So they're caught in this brutal paradox. They can't pause operations to install a new core system.

Aaron Ross Powell You can't just turn off the jet engine mid-flight to install a new one. Right. But if you don't upgrade, you fall out of the sky anyway because you won't be able to process transactions at the speed modern commerce requires. Aaron Powell Or meet those emerging compliance laws we talked about.

Exactly. So if replacing the engine mid-flight is out of the question, how exactly did SBS solve France Bank's problem? I mean, they must be wrapping the old system in something. I've got on the right chain.

Like a translation layer that sits between the outside world and the old system. Aaron Powell Kind of. SBS utilized what we can conceptualize as an exoskeleton approach. An exoskeleton.

So instead of attempting to rip out the massive, complex legacy system, which they call core SAB, they built cloud native functional modules completely around it. Aaron Powell So they just left the core alone? Entirely intact. They left it to do exactly what it has been rigorously tested to do for decades.

Aaron Powell Which is maintain stability. Absolute stability, ensure ledger accuracy, and handle all the heavy regulatory robustness built into its architecture. Aaron Powell And this approach is what won SBS some major industry recognition, right? Oh, big time.

They won the best in class core banking platform and best retail universal bank solution at the 2025 Global FinTech Innovation Awards for this exact solution. Aaron Powell Wow. Okay. But I have to challenge the reality of that for a second.

Sure. Because every time I see a company try to, you know, bolt brand new software onto an ancient system, it just becomes a tangled, clunky mess of workarounds. Trevor Burrus, Jr. It's a dreaded spaghetti code.

Trevor Burrus, Jr. Exactly. The old system drags down the new apps, and the new apps constantly crash because the old system can't feed them data fast enough. How does this not just become a fragile Frankenstein network?

Aaron Powell Well that's the trap of traditional integration, right? Where companies try to build these hard-coded bridges between new interfaces and the old ledger. Right. SBS avoided that by strictly focusing on high-value use cases.

They didn't try to modernize every single function at once. Oh, so they were selective. Very. They targeted digital banking, instant payments, and regulatory reporting.

Just those. High value areas. Exactly. By isolating those specific areas, they balanced modernization, cost control, and execution speed.

While leaving the rest alone. Right. But the most crucial element Dyke emphasizes in the interview is the preservation of technological sovereignty. Trevor Burrus, Jr.

Technological sovereignty. Meaning the bank retains absolute control over its foundational ledger. Yes. Rather than handing the keys over to a completely unproven system built by a third-party vendor.

Trevor Burrus Who might, you know, introduce hidden dependencies or lock them into some proprietary black box. Trevor Burrus, Jr. Right. So they own the core, they trust the core, but an exoskeleton is only as good as a nervous system connecting it to the pilot, if we're sticking with that metaphor.

Aaron Powell That's a good way to look at it. Because if these new cloud native apps are operating at 20 to 26 speeds, processing instant payments and real-time analytics. Yeah. And the core ledger is fundamentally a 1984 architecture built for end-of-day batch processing.

How do they communicate? Aaron Powell Without the modern tech completely overwhelming the old system. Aaron Powell Exactly. To understand why it's not a clunky mess, we have to look under the hood.

We do. And to visualize how these systems interact without causing a massive bottleneck, we really have to talk about REST APIs, containerization, and microservices. Aaron Powell Okay, let's break down the mechanics of that modularity. So instead of a monolithic software program trying to handle every single task, you know, checking balances, routing payments, generating compliance reports, SBS deployed microservices are highly specialized, independent programs that do exactly one job perfectly.

Right. And then containerization places each of those microservices into its own isolated digital environment. Exactly. It has its own memory, its own processing rules.

It cannot interfere with the microservice running right next to it. Which is huge for stability. It is. And the connective tissue between all these containers and the legacy core itself are REST APIs.

Right. These are standardized communication protocols that act as highly efficient translators. Aaron Powell, wait, you still have the speed issue. Aaron Powell Right.

The bottleneck. You cannot have a cloud native microservice hammering a 1984 ledger with 10,000 data requests a second. Aaron Powell The core would just immediately crash. It would.

So there has to be a buffer. Aaron Powell Are they using like a mirrored data lake or a caching layer to intercept those requests? That is the exact mechanism. The APIs don't just blindly route traffic directly into the core.

Okay. They interact with a decoupled data architecture. The modern modules read from synchronized data lakes or caching layers that mirror the core's information in real time. Oh, that makes sense.

So the microservices handle the massive volume of modern traffic, run all their logic, and then just use the API to drop a single settled transaction record into the legacy core. So the core never actually feels the strain of the traffic. Never. It only receives the final verified mathematical update.

Here's where it gets really interesting. It's like building a ship with watertight compartments. More like that. If one of the new experimental rooms springs a leak, you just seal the door.

The main ship, the core banking operations, keeps sailing perfectly fine. And that architectural decoupling, that operational resilience is exactly what makes this work. By strictly isolating the new services from the core, they drastically reduced systemic risk. Right.

If an experimental microservice fails, the incident is trapped. Contained entirely within its isolated module. And they didn't just dump this on Franz Bank and walk away, right? Yeah.

The interview highlights that SBS use a progressive vendor-led model. Yeah. They operated in SOS mode software as a service, which is vital to minimizing disruption. Aaron Powell Because traditionally a bank buys software and their internal IT department spends like five years trying to integrate it.

Usually resulting in budget overruns and compromised features. But here, SBS assumed full responsibility for the integration, testing, and deployment. They pre-integrated the modules with Core SAB and hosted them in a secure cloud. Exactly.

Letting the bank focus purely on business priorities. Okay, so with this heavily armored, compartmentalized ship in place, what kind of cutting-edge cargo is the bank now able to carry? Well, the immediate leap forward is instant payments. Which are heavily regulated now in Europe.

Very. The European Payments Council, the EPC, has incredibly strict standards for real-time processing. Right. Instant payments demand massive scalability.

And the ability to clear funds in seconds, 2047. A legacy core built around daytime hours and nightly batch processing simply cannot do that natively. But the cloud native microservice dedicated to instant payments handles it effortlessly. Exactly.

It executes the transfer, satisfies the EPC compliance, checks in milliseconds, and then asynchronously updates the legacy ledger whenever it's convenient. So what else does this unlock besides just fast payments? Embedded finance is a big one. Oh, right.

Because the architecture is open and API driven, it allows external ecosystems to securely connect to Friends of Bank services. Okay. Let's walk through a real-time scenario for that just to see how it flexes under pressure. Sure.

Imagine a manufacturing client in Paris needs to instantly pay a supplier in Singapore to release a shipment. They initiate the transfer through an embedded finance platform. So the banking service is integrated directly into their supply chain software, not a standalone banking portal. Right.

The supply chain platform sends the payment request via API. The instant payment microservice catches it. Instantly, an artificial intelligence compliance module intercepts the transaction to check for anti-money laundering flags. Wow, okay.

The AI scans a replicated data lake containing historical transaction patterns, cross-references international watch lists, and approves the transaction within a fraction of a second. And then the payment microservice executes the multi-currency transfer via a real-time network, paying the supplier in Singapore. And the legacy core. Only after all of that complex high-speed orchestration is complete does the API send a simple debit and credit instruction down to the 1984 core ledger to finalize the books.

That is wild. But um that flow reveals something massive about the role of data in AI here. Yeah, it really does. Because Jean-Charles Doyck states clearly that data processing and analytics will be the primary drivers of innovation.

Absolutely. But you know, AI is notoriously data hungry. Right. So what does this all mean for the actual data?

If everything is so safely decoupled and walled off in these watertight compartments, how does the AI actually get a comprehensive enough picture to provide these intelligent interactions? Aaron Powell Without accidentally interfering with the legacy core, you mean? Exactly. If the core is completely walled off behind APIs, how does the AI see anything?

Aaron Powell Well, this is exactly why the architecture supports progressive scaling, fine-grained monitoring, and controlled governance. Okay, what does controlled governance mean in this context? It means aligning the technological capabilities with strict operational responsibilities. The AI is never ever given direct rewrite access to the foundational ledger.

When the AI needs to build a predictive model or analyze customer behavior, it doesn't rummage through the live core system. It pulls from the data lakes. Exactly. It pulls anonymized structured data from the synchronized data lakes via those same APIs.

It's a calculated flow of data, not a free-for-all. So it's highly monitored. The AI gets the massive volumes of data it needs to learn, but its outputs, like a blocked transaction, are routed through verification layers before any action is finalized. Right.

It allows the institution to embed intelligence progressively. They can test a new AI algorithm in an isolated container, verify it against historical data, and only let it influence live transactions once it's proven safe. Wow. So you really do get the absolute best of both worlds.

You do. The unshakable reliability of a legacy system that has survived decades of market volatility, layered with the bleeding edge capability of cloud-based AI analytics. And you completely sidestep the existential dread of a total system overhaul. Exactly.

For anyone listening, whether you are in banking or any industry bogged down by legacy tech, the FranziBank and SPS project really proves a major point. Which is You don't have to choose between stagnation and risky total overhauls. Modularity and decoupling offer a completely safe, progressive path to the future. Aaron Powell They do.

And you know, if we extrapolate this methodology beyond Franzi Bank, it leaves us with a really provocative thought to mull over. Oh, what's that? We are so conditioned to believe that legacy systems are a liability, right? Right.

That old institutions will just eventually be crushed by agile startups. Exactly. But if this modular exaskeleton approach allows the most rigid, legacy-heavy systems in the world to safely adopt AI and instant processing, could the oldest, most established companies actually have a hidden advantage? Aaron Powell Wait, really?

How so? Aaron Powell Because instead of being replaced by agile startups, what if these legacy giants use modular upgrades to combine decades of deep historical data with cutting-edge AI? Oh wow. They could become more powerful and insightful than a brand new system ever could be simply because they have a 40-year head start on the data.

That completely flips the script. The old core becomes an indestructible engine block, completely shielded from the outside world, while the cloud native exoskeleton handles all the friction of the modern era. That changes how we view every piece of outdated technology running our daily lives. The next time you encounter an old system, remember that the solution might not be a wrecking ball, but just a better suit of armor.

Well said. Thank you for joining us on this deep dive. Keep questioning how the systems around you are built, and we'll see you next time.