Test whether your systems can handle AI
An ERP system is rarely just software. It is a digital reflection of how an organisation works. It has been built around every order, every invoice and every process that keeps the business running, sometimes for twenty or thirty years. That is exactly what makes it the backbone of the organisation. But that same longevity means the system was built at a time when nobody was thinking about AI. You notice this the moment you try to introduce new technologies like it.
If you think your organisation is the only one running an outdated legacy system, think again. In a recent study, only 11% of IT professionals surveyed said new technology could be integrated into their existing environment without modifications.¹ Business-critical applications, such as ERP, order and planning systems, simply aren’t designed for the interfaces and protocols that AI requires.
Want to know whether your foundation is ready for AI? These three questions will help you assess whether your business-critical systems can technically take that step.
1. Are the data and actions in your system accessible, and is that access secure?
Many business-critical applications simply weren’t built to make data and functionality available to applications outside the system, such as an AI tool. A common workaround is to build a separate layer around the system, acting as an intermediary between the system and the AI. That layer works, but it exists independently of the application itself. Say the underlying system gets a new rule: invoices above a certain amount must now be approved by a second person. If that rule isn’t also implemented in the intermediary layer, the AI keeps approving invoices the old way, without the extra check. Every change then makes the system more prone to errors, not less.
“Ask yourself: can the core system keep pace with new integrations?”
– Anne Buit
Just as important is whether AI follows the same rules as a human user. Who is allowed to perform which action, and within what boundaries? In many organisations, those roles and permissions exist only in employees’ heads. An experienced employee knows that while they could technically approve an invoice, they’re not supposed to. AI doesn’t know that, unless it is explicitly defined in the system.
New applications often bring new technologies and new protocols with them. Recording the CO2 emissions of a shipment sounds simple, but the way systems exchange that data is constantly changing. That’s why it matters to ask whether the technology your core system is built on can keep pace with these new kinds of integrations.
Think of the rise of electric cars. It took a while before charging connectors became compatible with each other. A system built on the expectations of five or ten years ago isn’t automatically ready for today’s standards.
Applying AI to legacy systems is therefore less about choosing the right tool and more about designing a foundation that can carry new technology. Once that foundation is in place, a business-critical application can evolve. AI just happens to be the latest development that puts demands on your foundation. Get it right now, and you won’t hit the same wall at the next technological shift.
