Patterns of AI Use in Procurement and Firm Performance
Overview
Companies everywhere are bringing artificial intelligence into their purchasing departments. Software now helps buyers analyze spending, find and evaluate suppliers, manage risk, and even negotiate. Research has followed, but almost always with one narrow question: does adopting AI improve performance?
My current research starts from a different observation. Two companies can both say “yes, we use AI” and still be completely different. One may use AI across half of its purchasing work, deep inside strategic decisions, running autonomously. The other may only use it to sort invoices. Existing research cannot tell these two firms apart, because it measures whether firms adopt AI, not how they use it.
The Three Questions
The project asks three deceptively simple questions that, at the firm level, remain largely unanswered:
- How much of the procurement work does AI actually touch? (extent)
- Where does it sit: spend analysis, supplier selection, negotiation, contracts? (locus)
- How deep does it go: does AI advise while a person decides, or does it decide on its own? (depth)
The wager of the project is that these distinctions decide almost everything about whether AI pays off.
Why It Matters
Early evidence points the same way. A study of 400 physicians found that AI used as an assistant raised productivity and innovation, while the very same technology used as a replacement lowered both. A survey of 408 manufacturers found that simply owning digital technologies had no direct effect on resilience. And a 2026 survey of 2,648 executives found that only 9% want AI leading procurement decisions, while 46% want it kept in tactical tasks under human control.
Practitioners are already debating where and how deep AI belongs. Academia has not yet measured it. This sits squarely within the Industry 5.0 agenda, where the question is no longer how to automate people out, but how humans and AI work together toward more resilient and sustainable operations.
What the Project Will Do
At its core is a firm-level survey of purchasing and supply chain managers, measuring the pattern of AI use across the procurement process and linking it to outcomes that matter: cost performance, supply chain resilience, and sustainability. The design builds on established survey research in the field (structural equation modeling with mediation and moderation analysis), with a planned second phase using individual-level experiments on actual AI-usage behavior.
Where It Stands
As of July 2026, this is an early-stage, unpublished project.
- Systematically read and coded 46 academic papers and industry reports into a structured literature matrix (research questions, methods, variables, measurement, findings, stated gaps).
- Identified and articulated the research gap, and verified through systematic database searches that it remains open.
- Drafted a first conceptual model connecting the pattern of AI use to firm outcomes.
- Next milestones: finalizing the conceptual model and hypotheses, drafting the survey instrument, and targeting an academic conference submission in late 2026.
The aim is to give both researchers and procurement leaders a way to describe, compare, and improve how AI is actually deployed, rather than just counting who has it.