Procurement teams are racing to roll out AI, but many are running into a problem they don’t see coming: bad supplier data. According to new research from apexanalytix, fragmented and outdated supplier records are causing procurement AI systems to “hallucinate,” leading to missed risks, failed automation, and costly blind spots. In this Q&A, Danny Thompson, Chief Product Officer at apexanalytix, explains why fixing supplier data is now critical before AI can deliver real value.
What’s Related
Supply Chain 24/7: When you say procurement AI is “hallucinating,” what does that actually look like in practice?
Danny Thompson: When procurement AI hallucinates, it produces results that seem plausible but aren’t actually supported by real supplier data. This phenomenon is tied directly to the well-established “hallucination effect”, where AI fills gaps in fragmented, outdated, or inconsistent source data (in the Procurement space, “Lazy Supplier Data”) with guesses rather than facts, producing false confidence and flawed outputs. In practice, this can show up as AI recommending a supplier as low risk when an accurate compliance or financial profile would show otherwise, generating risk alerts that don’t match missing supplier facts, or automated actions that fail or cause operational disruption.
SC247: How does supplier data end up so fragmented, duplicated, or outdated in the first place?
DT: Supplier data can end up fragmented or outdated for a number of reasons, the main being multiple, disconnected systems (ERP, procurement tools, spreadsheets) each storing fragments of supplier records, a lack of unified governance or validation processes, and data decay over time. Supplier details like names, addresses, tax IDs, compliance docs, or bank details can be scattered across systems, and compliance and risk status can change faster than traditional assessment cycles, leading to inaccurate data that isn’t just inefficient, but directly undermines AI and automation.
SC247: Why is bad supplier data becoming such a big issue now, instead of five or ten years ago?
DT: Bad supplier data has always existed, but the rise of AI and autonomous workflows has magnified its impact. In the days of manual human processes, Procurement could “fail slow”, allowing stakeholders to recognize and correct mistakes. AI and autonomous workflows based on poor “fail fast”, and without the necessary oversight to catch most errors in time to prevent significant impact. As a result, poor data quality is the number one barrier to AI adoption in procurement, making it not just a back-office annoyance but a fundamental roadblock to transformation. Now that organizations are investing heavily in AI for sourcing, screening, risk checks, and automated supplier interactions, their systems depend more heavily than ever on complete, reliable, validated data to function properly. Making the issue even more prominent is the fact that AI accelerates decision cycles and scales automation, exposing data quality gaps that were previously hidden.
Danny Thompson
SC247: What was the most surprising or concerning finding from this new research?
DT: A pretty significant finding is that most procurement solutions that claim an AI component don’t have the volume of accurate data necessary for the AI to make meaningful insights.
“Another is that most companies don’t have a line of sight on how to fix their data problem, they don’t know that solutions exist that solve the biggest challenges associated with filling their data gaps and keeping it up to date. So they are frozen, not knowing how to reach a point when AI can have the transformative effect required to compete.
SC247: What tends to break when companies roll out procurement AI before totally fixing their data?
DT: Gartner research shows that without a cohesive data strategy, supplier risk initiatives tend to be fragmented and reactive, rather than proactive and predictive. This means that when procurement AI is rolled out on top of flawed data, several things are at risk of breaking. Automated processes can fail or generate misleading outputs, while risk assessments provide false confidence or miss real signals. Failing to fix your data before adding on additional technologies can exacerbate the problem and compound risk, turning AI investments into sources of operational exposure rather than competitive advantage.
SC247: What’s the biggest mistake procurement teams make with supplier data?
DT: The biggest mistake procurement leaders make with supplier data is treating it as an administrative task rather than a strategic asset. Those who do not enforce shared standards, allow siloed versions of truth, or fail to build governance around master data management leave fragmentation and duplication entrenched, hoping that the technology will fix it. But that is not the case. AI built on poor data amplifies problems rather than resolves them.
SC247: If a procurement leader realizes this is a problem at their company, what’s the best first step to take?
DT: If bad supplier data is an issue at your company, the first priority should be transitioning away from manual, disconnected processes towards unified, validated supplier data records. Creating a strategy that establishes clear ownership for supplier records, standardizes data attributes, and centralizes master data for a consolidated view across systems, and processes to continuously monitor, validate and automatically update supplier data can help eliminate inconsistencies, improve data quality, and create a scalable foundation for smarter sourcing, compliance, and decision-making.
SC247: Do you see clean, trusted supplier data becoming a real competitive advantage as AI adoption grows?
DT: Trusted supplier data is the competitive advantage for AI adoption. Without it, organizations risk automating bad decisions at scale. As AI becomes more embedded in procurement across the entire lifecycle, the quality of the outcomes will only be as good as the data feeding those models. Clean, trusted supplier data enables AI to surface real risk signals, prevent disruptions, uncover savings opportunities, and drive faster, more confident decisions. Companies that invest in governing and validating supplier data won’t just be more efficient, but they’ll be able to outpace competitors who are trying to layer AI on top of fragmented, unreliable information.
Danny Thompson is Chief Product Officer at apexanalytix
