“Ninety per cent of AI investment is going into outputs. Shiny models. Fancy interfaces. The easy part. Most organisations will fail with AI. Not because they chose the wrong model. They’ll fail because their data wasn’t ready for it.”
That was David Rice, Group CEO of Snap Analytics, at a recent Bristol Tech Leaders event. He wasn’t being provocative. He was being accurate.
And SAP just proved him right.
Last week, SAP announced the acquisition of Reltio, a master data management platform. The business press covered it as an enterprise software story. It is bigger than that.
Muhammad Alam, SAP’s board executive for product and engineering, was direct about why it matters. “AI cannot reach its full potential when data is fragmented across business units, platforms and domains, without connection or context,” he said.
That’s not a product announcement. That’s a reflection on the state of enterprise data. And it points at a problem that has been building for decades.
How The Mess Got Made
I have spent more than 19 years inside enterprise software transformations. I know what a bad implementation looks like. It looks fine. The system runs. Reports get produced. Orders get processed. Nobody looks too hard at what’s underneath. Nobody needs to. Not immediately..
Here is how it happens. A large systems integrator wins a transformation contract. Their incentive, whether anyone admits it or not, is complexity. Every customisation is a billable day. Every requirement becomes a change request. Standard configuration, clean data, simple architecture? These are the enemy of a long engagement. Systems get customised. Data gets siloed. It works well enough. The SI moves on.
Tony Baer, principal analyst at dbInsight, noted that when SAP launched its Business Data Cloud platform, the mission was first “about cleaning its own house with a common One Domain Model for all SAP data.” If SAP needed to clean its own house, consider the state of the broader enterprise landscape.
The mess has been there all along. What’s changed is what happens when you point AI at it. The final irony is the systems integrators who over-customised these implementations and moved on won’t fix this problem. They’re going to sell you an AI strategy instead. Same incentives. Different invoice.
AI Is An Accelerant. In Every Direction
Artificial intelligence does not discriminate between good data and bad data. It takes what you give it and acts on it. At speed and at scale.
Point it at clean, consistent, trustworthy data and it makes better decisions than any human team could manage alone. Point it at fragmented, contradictory data and it will make bad decisions — wrong pricing, flawed demand forecasting, compliance failures — with the same speed and confidence.
“Systems integrators won’t fix the problem. They’re going to sell you an AI strategy instead.”
This is what makes the current moment so consequential. Bad enterprise implementations were always expensive. Overruns, delays, frustration. But they were manageable. Firms could paper over the cracks. AI removes that option. It is an accelerant in every direction. And if the underlying structure isn’t clean, it will find the cracks and drive straight through them.
The organisations that spent years over-customising their systems, siloing their data, and accumulating technical debt are not only facing an IT problem. They are facing a business risk. And it is arriving faster than they realise.
What Reltio Fixes
Reltio’s core capability is entity resolution. Across a large organisation, the same customer, supplier or employee may exist as dozens of separate records across dozens of separate systems. Spelled differently, categorised differently, updated at different times. Reltio reconciles all those records into what it calls a golden record. One clean, authoritative version of the truth.
That golden record is the precondition for reliable AI. Without it, an AI making a procurement decision may be working from three contradictory versions of the same supplier. An AI managing customer relationships may not know that two records refer to the same person. The errors compound — silently, until they don’t.
A Better Answer?
At Pivot we have always believed that standard SAP, properly implemented, produces clean data as a natural by-product. Not out of principle. Out of experience. Customisation creates technical debt. Technical debt creates data fragmentation. Data fragmentation creates AI risk.
The global leaders we work with are sophisticated organisations. They made reasonable decisions with the information they had. The fault is not theirs. It is systemic. Built into the incentive structures of the implementation industry.
Reltio is a remediation tool. It fixes a problem after it has been created. The better answer is to build clean from the start. With an implementation partner whose incentives are aligned with yours, not with the complexity of the engagement.
But step back for a moment. Every transformative technology in history has been a multiplier. Electricity didn’t care whether it powered a factory or started a fire. The internet didn’t care whether it connected people or exposed them. The technology itself was neutral.
AI is no different. It will supercharge the organisations that are ready for it. It will supercharge the problems of the organisations that aren’t. Fragmented records, siloed systems, years of accumulated technical debt, all of it amplified and acted upon at a speed no human team can correct in time.
The question for every enterprise leader right now is not which AI to buy. It is whether the foundations beneath it are solid enough to build on.
