The Rise of AI in Procurement: Tackling the Bad Data Dilemma
As procurement teams rush to integrate AI technologies, a hidden hurdle looms: poor supplier data. Recent research from apexanalytix reveals that fragmented and outdated supplier records are causing AI systems to “hallucinate,” resulting in missed risks, failed automation, and significant blind spots. In this insightful Q&A, Danny Thompson, Chief Product Officer at apexanalytix, elaborates on why addressing supplier data is crucial for unlocking AI’s true potential.
Understanding AI Hallucinations in Procurement
Supply Chain 24/7: When you refer to procurement AI as “hallucinating,” what does that entail?
Danny Thompson: AI hallucinations occur when procurement AI generates seemingly plausible results without grounding them in accurate supplier data. This stems from a phenomenon where AI fills gaps in fragmented or outdated data—what I term “Lazy Supplier Data”—with educated guesses. The consequences can be dire, manifesting as AI mistakenly endorsing a low-risk supplier without a solid compliance or financial background. This flawed output can lead to misleading risk alerts or automated processes that falter, disrupting operations.
The Roots of Data Fragmentation
SC247: How does supplier data become fragmented or outdated?
DT: Supplier data fragmentation occurs due to several factors, primarily the existence of disconnected systems like ERPs and spreadsheets that each hold partial records. The absence of unified governance processes and the natural decay of data over time further contribute to this issue. Supplier information—names, addresses, compliance documents—often gets scattered across platforms, and the rapid changes in compliance and risk status can lead to inaccuracies that undermine AI’s effectiveness.
The Growing Concern of Bad Supplier Data
SC247: Why is problematic supplier data an increasing concern now?
DT: While bad supplier data has existed for years, the surge of AI and autonomous workflows has accentuated its impact. Previously, manual processes allowed for slower failure rates, enabling corrections. However, AI’s fast-paced decision-making can lead to failures that occur too quickly for timely intervention. Poor data quality has become the primary barrier to AI integration in procurement, transforming what was once a mere annoyance into an existential challenge. As AI finds increased use in sourcing, risk assessments, and supplier interactions, it becomes pivotal that organizations possess complete and validated data.
Danny Thompson
Key Insights from Recent Research
SC247: Were there any surprising findings from your recent research?
DT: One alarming discovery is that many procurement tools claiming to incorporate AI lack the necessary volume of accurate data for meaningful insights. Furthermore, a significant number of organizations lack a clear strategy for addressing their data challenges, which hinders their ability to leverage AI effectively for competitive advantage.
The Risks of Rushing AI Implementation
SC247: What breaks when companies implement procurement AI without rectifying their data first?
DT: According to Gartner, organizations without a cohesive data strategy experience fragmented, reactive supplier risk initiatives. Rolling out AI atop flawed data often leads to automated processes failing or yielding misleading results. Risk assessments become unreliable, further diluting organizations’ confidence in their decision-making. Tackling data issues before adopting advanced technologies is crucial, as neglecting this can worsen existing problems and transform AI initiatives into liabilities.
A Strategic Approach to Supplier Data
SC247: What’s the most common mistake procurement teams make regarding supplier data?
DT: The gravest misstep procurement leaders make is treating supplier data as a mere administrative function instead of a strategic asset. Moreover, failing to establish shared standards or governance around data management fosters fragmentation and redundancy, leaving organizations vulnerable. Instead of resolving issues, AI feeds on poor quality data, amplifying problems exponentially.
Steps Towards Improvement
SC247: If a procurement leader identifies poor supplier data at their organization, what should be their first step?
DT: The immediate focus should shift from manual processes to unified, validated supplier data management. This involves developing a strategy that assigns ownership for supplier records, standardizes data attributes across systems, and implements continuous monitoring and validation. Doing so can improve overall data quality and create a robust foundation for effective sourcing, compliance, and informed decision-making.
The Competitive Edge of Trustworthy Data
SC247: Do you envision clean, trusted supplier data becoming a competitive advantage as AI adoption accelerates?
DT: Absolutely. Clean and trustworthy supplier data is essential for successful AI adoption. Without it, organizations risk automating poor decisions at scale. As AI becomes integral to procurement processes, the effectiveness of outcomes will depend significantly on the quality of the underlying data. Reliable data signals real risk, uncovers savings opportunities, and fosters timely, confident decisions. Companies prioritizing governance and validation of their supplier data will not only enhance efficiency but also leave competitors employing fragmented data behind.
Danny Thompson is Chief Product Officer at apexanalytix
