In 2025, Gartner projected that by 2028, 90% of B2B buying interactions would be mediated by AI agents, channeling more than $15 trillion in spend through automated exchanges. That prediction shaped vendor roadmaps, conference narratives, and executive expectations. Autonomous AI became the headline.
Then reality started to surface.
Recent autonomous agent developments — OpenClaw being the most documented example — demonstrated how quickly these systems can scale, and how quickly they can expose underlying risk. The capability is real: agents executing tasks across systems, interacting with APIs, and operating with minimal supervision. The exposure is equally real: security concerns around prompt injection, data exfiltration, and control boundaries emerged just as quickly as the adoption curve.
At the same time, Gartner signaled a sharp correction: more than 40% of agentic AI projects are expected to be canceled by the end of 2027, citing cost, unclear business value, and inadequate risk controls. Their own analyst called most of these projects “early stage experiments driven by hype and often misapplied.”
This is not a forecasting failure. It is a readiness failure.
Across seven technology cycles, the pattern has remained consistent. The constraint was never the technology. The capability works. What doesn’t exist in many organizations are the pre-conditions the technology requires to function as designed — clear decision ownership, data that reflects operational reality, mapped behavioral dependencies, and incentive structures aligned with the outcome being automated. Without those conditions, the technology does not fail. It succeeds at the wrong thing.
If you skip Phase 0™ and don’t have a complete picture of how Human Agents and Agentic AI Agents interact in your actual business structure, you’re not automating processes — you’re automating mayhem. And at AI speed, mayhem does more harm than good.
This is the pre-commitment gap — the stage where organizations determine whether the system can support the outcome before deployment begins. It is also the stage most consistently skipped. Which is why the pattern repeats.
Gartner’s forward-looking projections and its later caution are not contradictions. They are two sides of the same reality: capability is advancing faster than organizational conditions are being validated.
Recent HFS™ assessments have already identified certain agentic AI platforms as high risk without readiness, with capability-to-outcome gaps widening as vendors move toward multi-agent architectures. The question for buyers is whether readiness is being genuinely integrated — or layered on after the fact. The archive has 18 years of evidence on how that question resolves when it goes unanswered.
The Procurement Insights archive has documented this dynamic across 3,300+ independently produced entries spanning more than 18 years. Zero vendor sponsorships. Zero paid analyst relationships. The conclusion has remained consistent: organizations do not fail because the technology doesn’t work. They fail because they deploy it into conditions that cannot sustain the outcome.
That is why RAM 2025™ matters. Its multi-model, multi-level structure exists to assess how decisions actually behave under real-world conditions — not just how they are designed to perform.
But There Is a More Fundamental Question the Market Is Not Asking
The conversation around autonomous procurement assumes the goal is to make procurement fully autonomous. That is the wrong question.
The right question is: under what conditions can procurement safely become autonomous?
There is a domain where autonomy genuinely works — invoice matching, PO processing, catalog buying, contract compliance checks. Stable, repeatable, bounded processes where conditions are predictable and interpretation is minimal. Within that scope, automation delivers real value.
The moment you move into sourcing strategy, supplier selection under uncertainty, behavioral incentives, or real-world variability, you encounter a problem no agentic system has solved: the system doesn’t know what question to ask.
The Department of National Defence proof case from 1998 established this exactly. The breakthrough was not better data, better process, or better technology. It was asking a question the system didn’t know it needed: what time of day do orders come in?
That question was not in the data. It was not in the workflow. It was not in the process. It was a reframing of reality — and agentic AI cannot reliably do that today. These systems operate on available signals, not missing logic. They cannot identify hidden variables, incentive distortions, or behavioral patterns that exist outside the data they are given.
This is the critical inversion the market is missing.
The assumption is that Phase 0™ slows down autonomy. The reality is that Phase 0™ determines whether autonomy is even possible.
You do not bolt Phase 0™ onto an autonomous system. You use Phase 0™ to define the system — mapping reality, identifying constraints, exposing hidden variables — and then apply automation selectively where conditions allow.
The industry is asking: how do we make procurement autonomous?
Phase 0™ asks first: what is the real system this decision operates in?
It is the gatekeeper of autonomy.
What is different this time is that AI itself is forcing a pause. The same capability that accelerated adoption is now exposing risk, cost, and control gaps quickly enough that organizations are being forced to slow down. That changes how Phase 0™ is perceived. It is no longer seen as delay or interference. It becomes the activation point — the step that determines whether speed creates value or simply accelerates failure.
Because the real issue is not whether autonomous AI works.
It does.
The question is whether the organization it operates within does.
Phase 0™ is the pre-commitment organizational readiness diagnostic that addresses that question — before the commitment is made, when the outcome is still changeable.
→ Book a 30-Minute Readiness Conversation
Procurement Insights 
When Gartner’s Own Predictions Collide, It’s Not a Forecasting Problem — It’s a Readiness Problem.
Posted on April 7, 2026
0
In 2025, Gartner projected that by 2028, 90% of B2B buying interactions would be mediated by AI agents, channeling more than $15 trillion in spend through automated exchanges. That prediction shaped vendor roadmaps, conference narratives, and executive expectations. Autonomous AI became the headline.
Then reality started to surface.
Recent autonomous agent developments — OpenClaw being the most documented example — demonstrated how quickly these systems can scale, and how quickly they can expose underlying risk. The capability is real: agents executing tasks across systems, interacting with APIs, and operating with minimal supervision. The exposure is equally real: security concerns around prompt injection, data exfiltration, and control boundaries emerged just as quickly as the adoption curve.
At the same time, Gartner signaled a sharp correction: more than 40% of agentic AI projects are expected to be canceled by the end of 2027, citing cost, unclear business value, and inadequate risk controls. Their own analyst called most of these projects “early stage experiments driven by hype and often misapplied.”
This is not a forecasting failure. It is a readiness failure.
Across seven technology cycles, the pattern has remained consistent. The constraint was never the technology. The capability works. What doesn’t exist in many organizations are the pre-conditions the technology requires to function as designed — clear decision ownership, data that reflects operational reality, mapped behavioral dependencies, and incentive structures aligned with the outcome being automated. Without those conditions, the technology does not fail. It succeeds at the wrong thing.
If you skip Phase 0™ and don’t have a complete picture of how Human Agents and Agentic AI Agents interact in your actual business structure, you’re not automating processes — you’re automating mayhem. And at AI speed, mayhem does more harm than good.
This is the pre-commitment gap — the stage where organizations determine whether the system can support the outcome before deployment begins. It is also the stage most consistently skipped. Which is why the pattern repeats.
Gartner’s forward-looking projections and its later caution are not contradictions. They are two sides of the same reality: capability is advancing faster than organizational conditions are being validated.
Recent HFS™ assessments have already identified certain agentic AI platforms as high risk without readiness, with capability-to-outcome gaps widening as vendors move toward multi-agent architectures. The question for buyers is whether readiness is being genuinely integrated — or layered on after the fact. The archive has 18 years of evidence on how that question resolves when it goes unanswered.
The Procurement Insights archive has documented this dynamic across 3,300+ independently produced entries spanning more than 18 years. Zero vendor sponsorships. Zero paid analyst relationships. The conclusion has remained consistent: organizations do not fail because the technology doesn’t work. They fail because they deploy it into conditions that cannot sustain the outcome.
That is why RAM 2025™ matters. Its multi-model, multi-level structure exists to assess how decisions actually behave under real-world conditions — not just how they are designed to perform.
But There Is a More Fundamental Question the Market Is Not Asking
The conversation around autonomous procurement assumes the goal is to make procurement fully autonomous. That is the wrong question.
The right question is: under what conditions can procurement safely become autonomous?
There is a domain where autonomy genuinely works — invoice matching, PO processing, catalog buying, contract compliance checks. Stable, repeatable, bounded processes where conditions are predictable and interpretation is minimal. Within that scope, automation delivers real value.
The moment you move into sourcing strategy, supplier selection under uncertainty, behavioral incentives, or real-world variability, you encounter a problem no agentic system has solved: the system doesn’t know what question to ask.
The Department of National Defence proof case from 1998 established this exactly. The breakthrough was not better data, better process, or better technology. It was asking a question the system didn’t know it needed: what time of day do orders come in?
That question was not in the data. It was not in the workflow. It was not in the process. It was a reframing of reality — and agentic AI cannot reliably do that today. These systems operate on available signals, not missing logic. They cannot identify hidden variables, incentive distortions, or behavioral patterns that exist outside the data they are given.
This is the critical inversion the market is missing.
The assumption is that Phase 0™ slows down autonomy. The reality is that Phase 0™ determines whether autonomy is even possible.
You do not bolt Phase 0™ onto an autonomous system. You use Phase 0™ to define the system — mapping reality, identifying constraints, exposing hidden variables — and then apply automation selectively where conditions allow.
The industry is asking: how do we make procurement autonomous?
Phase 0™ asks first: what is the real system this decision operates in?
It is the gatekeeper of autonomy.
What is different this time is that AI itself is forcing a pause. The same capability that accelerated adoption is now exposing risk, cost, and control gaps quickly enough that organizations are being forced to slow down. That changes how Phase 0™ is perceived. It is no longer seen as delay or interference. It becomes the activation point — the step that determines whether speed creates value or simply accelerates failure.
Because the real issue is not whether autonomous AI works.
It does.
The question is whether the organization it operates within does.
Phase 0™ is the pre-commitment organizational readiness diagnostic that addresses that question — before the commitment is made, when the outcome is still changeable.
→ Book a 30-Minute Readiness Conversation
Share this:
Related