The maturity curve has two halves, not four steps. The architecture that wins the first does not graduate to the second.
TL;DR
- Automation and autonomy sound similar. They are different architectures with different design centers.
- 71% of procurement teams are stuck one stage short of autonomy on a four-stage maturity curve. The next step is not an extension of the last one.
- Automation optimizes individual steps. Autonomy governs whole workflows. One does not compound into the other.
- The architectural gap is data, decision authority, and orchestration. Real Stage 4 requires building for all three, not bolting them onto a Stage 3 stack.
- The cost of staying at Stage 3 compounds in measurable terms over the next 24 months.
- The full Foundry/CIO Market Pulse research is now available. Read the full report.
The Two Words Most Procurement Teams Conflate
The first blog in this series named the gap between near-universal agentic AI ambition and rare execution. The maturity curve explains why.
Automation and autonomy sound like the same thing on different scales. They are not. They are different architectures with different design centers, and the gap between them is the reason 71% of procurement teams are stuck one stage short of autonomy on a four-stage maturity curve they did not realize had two distinct halves.
The function has invested heavily in AI. Productivity is up, cycle times are down, compliance is tighter. And yet the gap between where procurement thought it would be by 2026 and where it actually is keeps widening. The reason is not execution failure. It is framing failure: the path from automation to autonomy is not a curve. It is a step change.
What Automation Actually Is
Automation makes a procurement step faster, cheaper, more accurate. PO routing speeds up. Invoice OCR replaces manual data entry. Contract clause extraction surfaces obligations a human would have missed. Each is a real win. Each delivers measurable ROI.
The defining characteristic is who holds the goal. Under automation, the human does. The system performs the step; the human evaluates the result and chooses what happens next. The system is a force multiplier on human judgment, not a substitute for it.
This is what most of the 71% have built. It is genuine progress over the manual baseline. But it is not autonomy, and the architecture that produces it does not produce autonomy.
What Autonomy Actually Is
Autonomy is structurally different. It does not make individual steps faster. It makes the entire workflow run itself. The system identifies what needs to happen, decides which actions advance the goal within set policy boundaries, executes across multiple systems, and returns to the human only when an exception breaches a threshold.
Consider supplier onboarding end to end: risk screening, validation, approval routing, contract setup, system integration. Under autonomy, the entire chain runs without a human checkpoint at each handoff. The human holds the policy and the exceptions. The system holds everything else.
The defining characteristic is the inversion of who holds the goal. Under automation, the human does. Under autonomy, the system does. That inversion changes what the technology has to be capable of, what the data has to support, and what governance has to look like.
Why One Does Not Compound Into the Other
Three architectural differences explain why automation does not graduate to autonomy through accumulation.
- Data. Automation works on fragmented data per step. Autonomy requires unified context across spend, suppliers, contracts, policy, and history. A workflow that runs itself cannot reason across data the system cannot see.
- Decision authority. Automation needs human approval at each step. Autonomy makes decisions within policy boundaries. The shift is not a configuration change. It is a redesign of where authority sits and what governance covers when machines make decisions.
- Orchestration. Automation runs steps. Autonomy runs workflows. The orchestration layer that holds a procurement workflow together is a separate architectural component, not an extension of the tools that automate steps inside it.
This pattern is not specific to procurement. BCG’s research on closing the AI impact gap has documented that most enterprises struggle to translate AI investment into sustained value at scale, with the cause traced to enterprise architecture rather than pilot capability. The systems built to win the pilot phase are not the systems required to scale beyond it. The 71% is not a procurement-specific condition. It is the procurement expression of a wider enterprise pattern.
Why 71% Is Structural, Not Behavioral
The 71% is not a snapshot of effort. It is a structural position. The bulk of the procurement market is stuck at the same point on the curve not because they ran out of will, but because they hit the architectural ceiling of what they had built. The next step is not what they thought it was.
The report’s four-stage maturity model places most of the market at one specific stage. The two stages on either side of that one define the report’s central diagnostic, letting a CPO see where their own organization actually sits in five minutes of self-assessment.
One detail worth noticing: the research itself flags the Stage 4 figure as more aspiration than measurement, since the supporting data on governance readiness and outcome attribution does not support it. What that means in practice: an even smaller share of the market has reached genuine, governed autonomy than the headline suggests. The 71% stuck in Stage 3 is a larger problem than even the visible numbers admit.
Three questions get to honest placement. Who holds the goal when your AI runs, team or system? Does the system reason across procurement data, or only within each step? When something goes wrong, does the workflow stop, or escalate within policy?
What Staying at Stage 3 Compounds Into
Stage 3 wins are real and measurable today. Organizations at this stage are delivering meaningful productivity gains across the categories that matter to procurement leaders. Those gains are not trivial. They are also not enough.
The trap is the projection. The report also quantifies what AI’s procurement workload share is expected to look like in 24 months, and the math on the gap between today’s measured impact and tomorrow’s required output is the substantive number the report puts at the center of its forward analysis. The takeaway is not that current automation is failing. It is that current automation will not be enough.
Hackett Group’s research on top-quartile procurement organizations has consistently documented their disproportionate outperformance on cost, cycle time, and digital maturity. Every quarter at Stage 3 compounds distance against that cohort, since the architecture that supports autonomy keeps expanding in a way Stage 3 cannot.
What It Actually Takes to Move
The architectural answer follows from the diagnosis. Stage 4 requires building for unified data, system-level orchestration, and execution authority. These are the same three properties whose absence keeps Stage 3 stacks at Stage 3. The shift is not from features to more features. It is from features to architecture.
What changes for the team is not the tooling. It is the operating posture. Governance moves from oversight to design. The audit trail captures machine decisions with the same rigor as human ones. The procurement team holds the policy and the exceptions; the system holds everything else.
These are architectural commitments, not feature additions. Teams that try to bolt them onto a Stage 3 stack end up paying Stage 4 prices for Stage 3 outcomes.
The 71% Will Not Move in One Step
They will move when they recognize that Stage 3 was not the path. It was the plateau. The report names what the move required of those who have already made it, and what the cost of waiting compounds into.
Read the Full research
Most Procurement AI Investments Are Stalling. Here Is Why, and What to Do.
A global study of 240 senior procurement leaders. Research conducted by Foundry (IDG) for CIO Market Pulse. Sponsored by Zycus.
Previous blog in the series: Most Procurement AI Investments Are Stalling. Here’s What 240 Global Leaders Just Told Us
Next blog in the series: Security and Trust Are Stalling Procurement AI. Here’s What Actually Closes Both
Related Reads:
