.png)
When small data errors meet AI and why "Good Enough" no longer is
- davidnelson9
- 2 days ago
- 3 min read
Espoo, Finland - 2.9.2026 Every company running SAP has data quality issues. One assumes this is not going to come as a shock to most of you. For many of our 100 customers in the Nordics and Baltics, we have been helping to eradicate both the bad data and to improve/correct the processes that create them. But in this post I am not talking about those big errors that crash systems or will trigger an audit finding, but the small stuff. The kind that people work around every day without really thinking about it.
Take a fairly ordinary SAP Vendor or Business Partner master. It's quite common to find one supplier showing up with three slightly different names across 3 different plants. One version could have a different capitalization or a slightly different spelling. Another one could have "Inc" or "Ltd" put at the end. Another issue could be a price list might still be carrying currency conversions from years ago that nobody has checked. Or maybe a Vender number retired back in 2019 might still have open records associated with it. Now, on their own, none of these brings the business down. The people working with the data already know about it, and they know which version of the supplier name is the real one, which pricelist to trust and which one to ignore. These issues aren't normally seen as urgent. They are known issues that you have lived with for long time and they have workarounds.
Now, if you point an AI model at the same data and ask it to make the same decisions faster than any person could review them, the workarounds don't exist anymore. So the model doesn't know which vendor name belongs to which account. It just answers confidently based on whatever it's given. We see this with customers regardless of whether it's a chat bot, a forecasting model, a reporting tool feeding into operational or strategic decisions. The output comes faster, and it looks polished and accurate. But whether it's actually worth acting on depends on something far less exciting than the model itself. And that is the data underneath it.
A few things to add here. Gartner's newest data backs up the same pattern. So a survey of 353 data and analytics and AI leaders found that organizations with successful AI initiatives invest up to four times more of their revenue as a share of revenue in foundational areas like data quality, governance and AI ready people. In the same survey, only 39% of technology leaders felt confident their ai investments would actually improve financial performance. The organizations with the most mature AI ready data practices reported up to 65% greater business outcomes, including revenue growth and cost optimization.
None of this should surprise anyone who's been in this space for a while, but it's worth saying AI is going to be a genuine game changer. It will increase productivity and cut down a lot of the manual work, but if the data underneath it isn't solid, the decisions and operations built on top of that will fall short. Data quality and governance have been a part of every serious SAP conversation long before AI arrived. What's changed is who's watching the outcome. If you know the business well enough, you catch a flawed report before it goes anywhere. A flawed input to an AI system just produces a confident sounding answer on the other end. And confident isn't the same as correct. Even more than that, over time, it will erode the trust in the reports and the data quality which is absolutely vital when you're talking about your financial audit or your sustainability audits or decision making capabilities.
Automation and AI both free up time by handling routine work faster, but that saved time doesn't automatically check the data underneath it. Someone actually has to choose to spend time there.
Hydro, a long time Adsotech customer, is a good example of making that choice and automated a large share of its manual SAP data entry years before AI came into the picture. And its team was deliberate about what came next. As they put it
"when we invest more of our resources in quality assurance, we absolutely see improvements in the overall accuracy of our data."
If you'd rather see this in practice than read about it in the abstract, take a look at the Norsk Hydro's case study, or, better yet, give us a call. Always happy to talk through it and look forward to hearing your thoughts!
David Nelson, CEO, Adsotech
Comments