I’ve been saying for so long now that banks need to replace core legacy systems that I’m boring myself, but here I go again. The reason I’m talking about it again is that, even though some disagree and think they can fudge the issue with plug-ins, I believe that the new competition will decimate banks that don’t replace their core systems.

I say this because I talk to firms like Ant Financial who refresh their complete systems architecture every three or four years, and think that this mindset really hits to the heart of the difference between a technology company that happens to do finance (techfin), and a financial company that happens to use technology (fintech).

If you are tech first, your singular focus is on agility. It’s about fast change cycles in a microservices architecture using a SDK (software developer kit) network of APIs (Application Programming Interfaces). It’s about speed, change, service, updates, vision.

If you are finance first, your singular focus is on stability. It’s about slow change cycles in a monolithic architecture using control systems and sign-off structures that avoid any exposures. It’s about risk, security, stability, control, management.

These two opposites are difficult to marry, and the example of 40-year old legacy systems is a clear example of the latter camp. Keeping a core system operational that was first implemented in 1976 is an indictment of a bank, and it has to change. Some say not, but imagine Amazon or Alibaba having systems that were untouched for 20 years, except by operational maintenance updates. Could they function in their fast change online environment today? I don’t think so.

And this is why I give banks with legacy core systems a maximum of ten years to change. In ten years, we will be in 2027 and not far off the timelines that Ray Kurzweil originally envisioned the Singularity, where machines become more intelligent than humans. In a world where machines are coded to talk, walk, think, see, hear, touch, smell and feel, how would an IBM mainframe (92% of banks’ core systems machines) with a COBOL programme (43% of U.S. banks’ core systems code) feel? Well, it just wouldn’t. But the banks stuck with such machines would have to compete with fintechs and internet giants who are using such machines.

In 2027, where the fintechs and internet giants are analyzing trillions of transactions per second in real-time with artificial intelligence, how would the banks’ core system working in batch overnight updates manage? It wouldn’t.

In 2027, when the competition between all industries and all industry players is about data analytics, how will the creaking old mainframe system deal with the competition? It won’t.

Banks with legacy core systems in a decade’s time will die.

They won’t literally explode, but it will be a slow creeping death by a thousand cuts of code, and the bank will get acquired or folded by regulators and competition. So, this sets the timeframe for a banks’ death: ten years; and their legacy core systems are their ten-year ticking time bomb.

How long does it take to replace these legacy core systems?

I would claim that a bank that wants to refresh its legacy core systems will need a five-year timeframe to do this. The reason I state this is that, especially for a big bank with lots of systems, it is a highly complex task. It is not risky, if done right, but it’s not simple.

First, the bank needs to create an Enterprise Data Architecture. It needs to consolidate, rationalize, analyze and organize its complex data across multiple systems and silos, into a single clear structure.

Second, it needs to move that data to the cloud, and separate content (data) from processing (servers).

Third, it needs to gradually identify what to replace, when. There’s no big change here, but lots of small swap-outs. After all, you don’t eat an elephant in one bite. You eat an elephant one mouthful at a time.

It’s this last part that requires a five-year strategy. Five years.

In other words, a bank that has not determined its core systems change strategy will be starting in 2018 and, by the time they finish that change program, it’s 2023. Four years before the time bomb explodes, assuming that the time bomb I’ve identified explodes in 2027.

Will it?

I think I’m being kind to be honest, as I think the bomb will explode in 2022, in which case you’re already too late. But let’s be kind for the sake of your hearts. It’s 2023+. In other words, whatever timeframe the bomb explodes, we are almost at midnight for those legacy systems.