Why Procurement Lags in Agentic AI Despite Clear ROI

The Procurement Paradox

Procurement should be leading the enterprise shift to agentic AI. Instead, it's trailing badly. While software development teams have reached 35% adoption of agentic AI systems and IT operations hit 31%, procurement languishes at just 9%, according to a recent survey of 385 organizations.

The gap is puzzling because procurement possesses three characteristics that make it exceptionally well-suited for autonomous AI: economic visibility with direct bottom-line impact, repeatable workflows at scale, and persistent friction points that resist traditional automation yet require judgment rather than simple rule-following.

Why it matters

Procurement's slow adoption of agentic AI represents a significant missed opportunity for enterprise value creation. Organizations that redesign procurement operations around autonomous systems are already capturing measurable EBIT contributions and productivity gains exceeding 45%, while competitors remain stuck in pilot purgatory. The window to build compounding advantage is narrowing.

Production Systems Delivering Results

Several organizations have moved beyond pilots to production-scale deployments. A major European energy company deployed AI agents to manage 30,000 to 40,000 annual supplier applications, each involving up to 25 documents and more than 20 qualification steps. The system now handles category classification (where suppliers err 30% of the time), queue prioritization, compliance scanning across multiple data sources, and executive briefings.

Six months after launch, the backlog dropped 20%, supplier satisfaction increased, and accuracy climbed from 60% to between 70% and 80%, with a trajectory toward 95%. A team of 25 people now manages work that would otherwise require significantly more staff.

A major automotive OEM took a different approach, rebuilding its entire procurement model around commercial outcomes rather than process efficiency. After discovering that time saved through early GenAI pilots couldn't be monetized within legacy processes, the company pivoted in 2024 to redesign the full procurement lifecycle. The result: 70% of procurement work now flows through a single agent-orchestrated interface. A controlled three-month A/B test produced measurable negotiation uplift and material EBIT contribution.

Three Organizational Barriers

The constraint isn't technological—it's organizational, according to research spanning more than a decade of procurement technology transformations. Three barriers consistently block adoption:

Accountability and control. Procurement sits at the intersection of legal, finance, and operations, with decisions carrying contractual consequences. Organizations respond by limiting agent autonomy so strongly they neutralize the benefits.

Fragmented ownership. IT or digital teams often drive AI initiatives with limited procurement input, creating a mismatch between system design and operational reality. Pilots succeed in controlled conditions but stall when meeting messy, exception-rich workflows.

Data as afterthought. Organizations attempt agent deployment before addressing inconsistent supplier master data, incomplete spend categorization, and fragmented legacy ERP systems. When systems fail, leaders blame the technology rather than inadequate foundations.

Seven Actions for Leaders

Organizations capturing value make distinct choices about ownership, sequencing, and governance:

1. Make procurement the owner, not the beneficiary, with end-to-end accountability for outcomes
2. Target the greatest pain points rather than low-risk use cases, focusing on direct materials and complex services
3. Treat data as infrastructure that must be addressed upfront, not retrofitted after failure
4. Design for graduated autonomy from advisory to human-in-the-loop to full execution under oversight
5. Build internal capability factories with reusable components rather than isolated projects
6. Engineer governance from the start with clear investment cases, approval rights, and audit trails
7. Measure commercial outcomes like captured savings and EBIT contribution, not adoption metrics

The research, detailed by Heiner Himmelreich, Ilan Oshri, Paolo Scala, and Anas Zaidani in Harvard Business Review, draws on in-depth interviews with procurement leaders and technology providers across multiple sectors. The technology is ready—the question is which organizations will build the advantage first.