The AI Adoption Paradox
Procurement leaders are highly enthusiastic about artificial intelligence, but a significant gap exists between experimentation and real-world execution. According to new research from Ardent Partners published by CPO Rising, approximately 90% of procurement organizations are currently exploring, piloting, or deploying AI. However, fewer than one in four have actually progressed past the trial phase into live operations.
This execution gap stems from a mismatch between technology capabilities and operational readiness. While AI tools are advancing rapidly, actual procurement operations are struggling to keep pace, leaving many teams unable to scale their pilots.
The Heavy Burden of Tactical Work
A primary obstacle to AI readiness is the persistent volume of administrative tasks. The Ardent Partners study, which surveyed more than 300 global CPOs and procurement leaders, revealed that 87% of teams still spend a meaningful portion of their time on tactical work.
This manual workload includes copying requisition data from spreadsheets and emails into ERPs, generating purchase orders, and chasing internal approvals for low-dollar items. Trying to layer advanced AI on top of these inefficient, manual processes without redesigning the underlying workflows limits the technology's impact.
Ambition Versus Readiness
To understand this dynamic, the research mapped organizations across two dimensions: AI ambition and operational readiness (including data quality, processes, governance, and people).
The findings show that roughly two-thirds of procurement departments fall into the Overreaching category. These teams have high AI ambitions and active pilots, but lack the foundational data and governance to support them. Only 23% of organizations sit in the AI-First quadrant, where their operational readiness matches their technological goals.
For mid-market leaders, the takeaway is clear: successful AI deployment requires cleaning up data, defining clear processes, and establishing strong governance before expecting automated systems to run autonomously.
