Our new article series, “The Path to AI-First Procurement: Closing the Gap Between AI Ambition and Execution,” based on new research from Ardent Partners and sponsored by Ivalua, looks at how procurement’s AI journey is being constrained less by a lack of interest in AI than by the tactical workloads, fragmented processes, legacy systems, and data challenges consuming organizational capacity. The path to AI-first procurement therefore begins with reducing that operational burden, thereby creating capacity for higher-value work and enabling greater speed, intelligence, and strategic impact.
Of all the findings in our recent Path to AI-First Procurement research, one of the numbers I keep coming back to is 11%.
That is the percentage of procurement organizations that report operating with a unified data model across Source-to-Pay. After more than two decades of researching procurement transformation, I have seen the industry’s data problem from just about every angle, including poor spend visibility, disconnected supplier records, contracts separated from transactions, and sourcing data that never makes its way downstream. Procurement teams have spent years working around these limitations through spreadsheets, manual reconciliation, and institutional knowledge. Those workarounds allowed organizations to function despite fragmented information environments, but their limitations become much more consequential as AI assumes a larger role in procurement operations.
 The 11% Problem
Our research asked 311 CPOs and senior procurement leaders to describe the current state of their procurement data across Source-to-Pay. Only 11% operate with a unified procurement data model, while 41% have integrated data that still depends on manual reconciliation across systems. Another 40% operate with fragmented data across sourcing, contracts, suppliers, and transactional activity, and 8% report that much of their procurement intelligence remains trapped in unstructured documents, PDFs, and emails.
The immediate takeaway is that 48% of procurement organizations remain in clearly fragmented or largely unstructured data environments. I think the more interesting number, however, is 89%.
Nearly nine out of 10 procurement organizations lack a truly unified data model. Even within the 41% that describe their environments as integrated, people are still reconciling information manually across systems. That represents meaningful progress from where the industry stood a decade ago, but it remains well short of the connected information environment that advanced AI increasingly requires.
AI Needs Context
For years, fragmented data primarily limited procurement’s visibility and decision-making. CPOs struggled to see spend clearly, understand supplier relationships across the enterprise, connect contracts to transactions, and establish reliable performance benchmarks. Organizations compensated by having analysts pull information together, category managers maintain their own records, and procurement teams reconcile information across systems. In effect, people supplied much of the context that the technology lacked, an operating model that becomes increasingly difficult to sustain as AI takes responsibility for more complex work.
Consider an AI-generated supplier recommendation. Its quality may depend on sourcing history, current contracts, supplier performance, risk information, pricing, transaction history, and category strategy. If those records reside in different systems, use different supplier identifiers, or require manual reconciliation, the recommendation is being developed from an incomplete picture.
The same issue appears throughout Source-to-Pay. Contract intelligence loses value when terms cannot be reliably connected to purchasing activity, while supplier risk analysis suffers when supplier records are incomplete or stale. Category recommendations are similarly constrained when spend, sourcing, market, and supplier information cannot be assembled in context. This is where procurement’s longstanding data challenge begins to create an AI trust problem.
When Incomplete Data Looks Complete
The greatest risk may not be an obviously poor AI output, since those are relatively easy to identify. The more difficult problem is a credible answer built on incomplete information.
A procurement professional reviewing an AI-generated supplier assessment may have no obvious way of knowing that the system lacked an important contract, relied on an outdated supplier record, or could not access relevant performance data. The output can appear complete even when the underlying context is not.
That matters because trust will help determine how quickly procurement organizations expand their use of AI. If users routinely need to verify recommendations, check source systems, or reconstruct the context behind an answer, expected efficiency gains diminish, and confidence in the technology erodes. Once procurement professionals become skeptical of AI-generated outputs, broader adoption becomes considerably more difficult.
Connected Beats Perfect
For CPOs, the practical objective should be connected data rather than some elusive state of data perfection. Few organizations will ever reach a point where every supplier record is pristine, every contract is structured, and every system is perfectly synchronized.
What matters is whether the information required to understand a business decision is accessible when and where that decision occurs. Supplier information should connect to contracts, contracts should connect to purchasing and invoices, sourcing history should inform category strategy, and relevant risk signals should be available when supplier decisions are made.
This distinction becomes increasingly important as procurement moves from AI that summarizes and recommends toward AI that coordinates processes, makes decisions, and executes work.
Data as Operational Infrastructure
The procurement industry has historically treated data as an enabler of reporting, analytics, and visibility. In an AI-first operating environment, data becomes part of the infrastructure through which procurement work gets done. That should influence how CPOs approach their AI roadmaps. They need to understand what information each AI capability requires, where that information resides, how reliably it can be connected, and whether the system has enough business context to use it appropriately.
For all the attention paid to models, copilots, agents, and the latest AI capabilities, the 11% finding points to a more fundamental issue. Procurement’s AI ambitions have advanced much faster than its underlying data environments. Closing that gap will be essential to turning today’s experimentation into tomorrow’s operating model.
This article is part of CPO Rising’s ongoing analysis of findings from Ardent Partners’ The Path to AI-First Procurement: Closing the Gap Between AI Ambition and Execution, based on the perspectives of 311 CPOs and senior procurement leaders across 25 industries worldwide.
Download this exclusive report from Ardent Partners.
RELATED ARTICLES
The Path to AI-First Procurement, Pt.1: What Does AI First Mean?
The Path to AI-First Procurement, Pt.2: The Execution Gap
The Path to AI-First Procurement, Pt.3: The Operational Reality Gap
The Path to AI-First Procurement, Pt.4: The Foundation Problem
The State of Source-to-Pay Transformation in 2025: Navigating Uncertainty Through Technology and AI
