Jon W. Hansen, FCIPS · Procurement Insights | Hansen Models™
This week, ProcureAbility released its 2026 State of Procurement Report, produced with ProcureCon from a survey of 100 senior procurement leaders. Its central finding is what it calls procurement’s strategic readiness gap: the function is now widely seen as strategic, but most organizations are still building the data, processes, skills and governance needed to deliver on that role.
The supporting numbers are telling. Only 11% of leaders said procurement takes part when the business case and requirements are being defined. Seventy percent described their supply chain resilience strategy as still developing. Only 8% have a defined roadmap for AI.
ProcureAbility’s CEO, Conrad Snover, who has been with the firm since 2001, draws the obvious lesson from the timing numbers, and he puts it well: strategic value starts with when procurement is engaged. Bring procurement in earlier, and it can shape requirements, widen supplier options and reduce risk before the critical decisions are made.
None of this is wrong. What struck me is that I have read it before, in September 1983.
What Kraljic wrote in 1983
That month, Peter Kraljic published Purchasing Must Become Supply Management in the Harvard Business Review, the article that introduced the Kraljic matrix. He described purchasing as “wedded to routine,” and warned that many purchasing managers’ skills and outlooks had been formed in an era of stability and had not changed. He called for better information systems, stronger organizational relationships and new skills. He pointed to General Motors bringing suppliers in early in the design process. And he opened with a list of disruptions, political turbulence, government intervention in supply markets and accelerating technological change, that could upset supply and demand virtually overnight.
Put the two side by side:
The issue
Kraljic, 1983
ProcureAbility, 2026
The function’s role
Purchasing must become supply management
Procurement is seen as strategic but is not yet ready to deliver
Foundations
Better information systems, organization, skills
Data, processes, skills, governance still being built
Influence before decisions are fixed
Integrate purchasing with the business; involve suppliers early in design
Only 11% of leaders say procurement is involved at the business-case stage
The same finding surfaced in between. In August 2007, three months after I launched Procurement Insights, I wrote about a 2006 CPO Agenda executive roundtable that asked whether there were any limits to procurement’s role. It brought together senior procurement personnel from Nestlé, Danone, British Airways and Merrill Lynch, and the discussion exposed the tension between procurement’s expanding strategic expectations and its actual standing inside organizations. One participant had moved to Nestlé from a company where procurement was involved in capital projects from the start. He said he now had “very limited involvement to actually influence the buying decision.” Nineteen years before the current survey, the same organizational problem was already being discussed by procurement executives from major global companies.
In the same post, I warned about a “class distinction” between a few strategic thinkers at the top and the buyers doing the day-to-day work, and the disconnect it could create between them. ProcureAbility’s own 2026 CPO Benchmark Study, a separate survey of 160 leaders, offers a contemporary parallel. As Tom Mills pointed out in his review of the study, only 26% of CPOs named late engagement as their biggest barrier, while 55% of directors and category managers in the same organizations did. As Mills put it, “By the time sourcing starts, most of that value’s already locked in.” The leaders see a seat at the table. The people doing the buying see decisions that have already been made.
The diagnosis is right. The instrument is wrong.
The prescription has been repeated since 1983, through ERP, e-procurement, SaaS, digital transformation and now AI, and the surveys keep finding the foundations still unbuilt. A gap that persists that long is not a stage on the way to maturity. Its recurrence raises a question about how the work keeps being approached.
I made this argument in February in The Kraljic Paradox. Kraljic described a world of independent actors, each moving on its own timeline, in which supply could be upset overnight. Then the industry built a static, four-quadrant matrix on top of that diagnosis. The diagnosis was agent-based. The instrument was equation-based. Framework after framework has followed a similar pattern: acknowledge the complexity in the introduction, then deliver a tool built around the factors its designers selected. The limitation is not simply whether a framework can be updated. It is whether updating revisits only the factors already chosen, or can discover that an unrecognized behavior or dependency is what determines the outcome. A method can be iterative and still keep asking the wrong question. A week later, I traced the same error from Kraljic to Salesforce’s new AI metric, which counts the work AI agents perform without asking whether any of it produced a result.
The 2026 report’s diagnosis is right: procurement is expected to do more than it is ready to do. The survey gives us a useful snapshot of that gap. What the published summary does not establish is how its recommended capabilities and roadmaps will change the operating conditions that keep producing late engagement. That requires more than measuring the gap again. It requires tracing how decisions are actually made, testing interventions and following their outcomes. Across the technology eras this archive has tracked, the reported failure rate for implementations has stayed in the 65% to 85% band.
A report that measures the gap again, without asking why it has persisted despite decades of attempts to close it, is a portrait taken one more time. It shows the same face, slightly older.
Two meanings of readiness
Part of the problem is what “readiness” is taken to mean.
As presented in the published summary, readiness is framed primarily as capabilities: better data, more talent, a clearer roadmap, the right partners. Build enough of them, and the gap closes.
In my work, readiness means something earlier. It asks whether the organization’s actual operating conditions can support the outcome it expects: how the work really flows, where the handoffs are, which workarounds and shadow processes nobody has mapped, who actually holds the authority to decide, what each group is rewarded for, who has access to a decision before it becomes fixed, and how suppliers, finance, operations and the people doing the work affect one another. Capabilities built before that is understood tend to automate the gap rather than close it.
The same six functions. One drawn as a sequence. One as it operates. Capabilities are built for the diagram on the left. Outcomes are decided where the strands on the right engage. Illustrative, not measured data.
In 1998, the Department of National Defence contract I worked on was running at 51% next-day delivery while every part of the operation met its own measure. The first question was not about systems or skills. It was what time of day the orders came in. Understanding how the work actually happened came first, and the technology came last. Delivery reached 97.3% and held for seven years.
Visibility is not the same as seeing
The report’s answer to resilience begins with visibility. Visibility matters, but visibility into individual functions is not necessarily visibility into their dependencies. Every function can see its own dashboard, and every dashboard can be green while the outcome is red. That is watermelon reporting. What is missing is not visibility into each part. It is an understanding of the connections between them, which is where outcomes are decided, as this week’s hurricane in the Gulf is showing in real time.
When the storm arrives, the route on the left is not what is crossed. Feedstock, power, pipelines, terminals, people and policy move on different clocks. Landfall is one moment. Power comes back later. Policy grows more active rather than returning to its old line. Illustrative, not measured data.
The missing loop
This week brought another study on the same gap, this time about AI. A new Process Context Study, from a benchmarking firm working with a process-software provider, found that 86% of respondents agree AI agents cannot be deployed reliably without understanding how the business actually works, yet only 22% have comprehensive, real-time visibility into how their processes operate. Its answer is to connect each process to the people, systems, data, rules, risks and controls behind it. That is a genuine step forward. It recognizes that decision rights, dependencies and exceptions matter, not just clean data.
But Kraljic’s matrix, the 2026 readiness report and this new study share the same missing piece: loopback learning. Each builds an understanding of the organization, whether as a categorization, a set of capabilities or a body of process context, and treats keeping that understanding current as if it were the same as getting it right. It is not. A system can refresh its data in real time and still hold the same mistaken assumption about why things happen. Context without a continuous learning loopback creates an illusion of real-world operations.
The DND finance example shows the difference. Recording the part invoices and the courier invoices accurately, in real time, would never have revealed that margin was being lost on every transaction. The data was correct. The interpretation of the purchasing and pricing rules was wrong. What exposed it was the financial consequence feeding back and changing how those rules were understood. In 1998, every order, delivery and post-delivery result was fed back into the system automatically, and today I continue to track the outcome of each decision long after it is made, in what I call the long tail assessment.
The loop has five steps, and it never stops:
Observe how people, suppliers and systems actually behave.
Trace the consequences across every strand they touch, not just the next handoff.
Challenge the current explanation of why things are happening.
Revise both the understanding and the intervention.
Verify the next outcome, and feed what it shows back into the first step.
That is why these instruments keep missing the operating reality. The operating reality is not a picture to be captured once and kept up to date. It moves, and the only way to keep up with it is a loop in which actual outcomes keep changing the understanding of the human and AI agents working inside it. Without that loop, every new framework, survey or context model becomes another portrait of a subject that will not hold still.
The question worth asking
Snover’s question is a good one: how do we get procurement involved earlier? Earlier engagement remains a worthwhile objective. But after 43 years, it may not be the deepest question.
The deeper question is why organizations keep leaving procurement out in the first place. Kraljic called for purchasing to be integrated with the business in 1983. Executives at a 2006 roundtable described procurement shut out of the buying decision. The 2026 report finds the same problem again. So the question I would put to Snover, and to all of us, is this: what combination of authority, incentives and access to decisions keeps producing late procurement involvement, what must change in that combination before additional capability can deliver the value it promises, and what loop will show whether the change actually worked?
Capability-building alone may not change those conditions. The next report should not only measure the gap. It should show which interventions changed it, and whether the improvement held.
Six minutes
In this short conversation, I walk through the operating reality on the right-hand side of the graphic, using the 1998 example: the four o’clock orders, the customs cutoff and the finance procedure that lost margin on every transaction. After this six-minute video, you will understand why technology initiatives fail more than they succeed.
Truth Is Believing. Accuracy Is Knowing. Outcome Is Proof.™
Snover Asks a Good Question, but After 43 Years, Is It the Right Question?
Posted on October 9, 2026
0
Jon W. Hansen, FCIPS · Procurement Insights | Hansen Models™
This week, ProcureAbility released its 2026 State of Procurement Report, produced with ProcureCon from a survey of 100 senior procurement leaders. Its central finding is what it calls procurement’s strategic readiness gap: the function is now widely seen as strategic, but most organizations are still building the data, processes, skills and governance needed to deliver on that role.
The supporting numbers are telling. Only 11% of leaders said procurement takes part when the business case and requirements are being defined. Seventy percent described their supply chain resilience strategy as still developing. Only 8% have a defined roadmap for AI.
ProcureAbility’s CEO, Conrad Snover, who has been with the firm since 2001, draws the obvious lesson from the timing numbers, and he puts it well: strategic value starts with when procurement is engaged. Bring procurement in earlier, and it can shape requirements, widen supplier options and reduce risk before the critical decisions are made.
None of this is wrong. What struck me is that I have read it before, in September 1983.
What Kraljic wrote in 1983
That month, Peter Kraljic published Purchasing Must Become Supply Management in the Harvard Business Review, the article that introduced the Kraljic matrix. He described purchasing as “wedded to routine,” and warned that many purchasing managers’ skills and outlooks had been formed in an era of stability and had not changed. He called for better information systems, stronger organizational relationships and new skills. He pointed to General Motors bringing suppliers in early in the design process. And he opened with a list of disruptions, political turbulence, government intervention in supply markets and accelerating technological change, that could upset supply and demand virtually overnight.
Put the two side by side:
In July 2024, I wrote that Kraljic’s opening description of the supply environment could apply as easily to 2024 as to 1983. It now applies to 2026 as well.
The same finding surfaced in between. In August 2007, three months after I launched Procurement Insights, I wrote about a 2006 CPO Agenda executive roundtable that asked whether there were any limits to procurement’s role. It brought together senior procurement personnel from Nestlé, Danone, British Airways and Merrill Lynch, and the discussion exposed the tension between procurement’s expanding strategic expectations and its actual standing inside organizations. One participant had moved to Nestlé from a company where procurement was involved in capital projects from the start. He said he now had “very limited involvement to actually influence the buying decision.” Nineteen years before the current survey, the same organizational problem was already being discussed by procurement executives from major global companies.
In the same post, I warned about a “class distinction” between a few strategic thinkers at the top and the buyers doing the day-to-day work, and the disconnect it could create between them. ProcureAbility’s own 2026 CPO Benchmark Study, a separate survey of 160 leaders, offers a contemporary parallel. As Tom Mills pointed out in his review of the study, only 26% of CPOs named late engagement as their biggest barrier, while 55% of directors and category managers in the same organizations did. As Mills put it, “By the time sourcing starts, most of that value’s already locked in.” The leaders see a seat at the table. The people doing the buying see decisions that have already been made.
The diagnosis is right. The instrument is wrong.
The prescription has been repeated since 1983, through ERP, e-procurement, SaaS, digital transformation and now AI, and the surveys keep finding the foundations still unbuilt. A gap that persists that long is not a stage on the way to maturity. Its recurrence raises a question about how the work keeps being approached.
I made this argument in February in The Kraljic Paradox. Kraljic described a world of independent actors, each moving on its own timeline, in which supply could be upset overnight. Then the industry built a static, four-quadrant matrix on top of that diagnosis. The diagnosis was agent-based. The instrument was equation-based. Framework after framework has followed a similar pattern: acknowledge the complexity in the introduction, then deliver a tool built around the factors its designers selected. The limitation is not simply whether a framework can be updated. It is whether updating revisits only the factors already chosen, or can discover that an unrecognized behavior or dependency is what determines the outcome. A method can be iterative and still keep asking the wrong question. A week later, I traced the same error from Kraljic to Salesforce’s new AI metric, which counts the work AI agents perform without asking whether any of it produced a result.
The 2026 report’s diagnosis is right: procurement is expected to do more than it is ready to do. The survey gives us a useful snapshot of that gap. What the published summary does not establish is how its recommended capabilities and roadmaps will change the operating conditions that keep producing late engagement. That requires more than measuring the gap again. It requires tracing how decisions are actually made, testing interventions and following their outcomes. Across the technology eras this archive has tracked, the reported failure rate for implementations has stayed in the 65% to 85% band.
A report that measures the gap again, without asking why it has persisted despite decades of attempts to close it, is a portrait taken one more time. It shows the same face, slightly older.
Two meanings of readiness
Part of the problem is what “readiness” is taken to mean.
As presented in the published summary, readiness is framed primarily as capabilities: better data, more talent, a clearer roadmap, the right partners. Build enough of them, and the gap closes.
In my work, readiness means something earlier. It asks whether the organization’s actual operating conditions can support the outcome it expects: how the work really flows, where the handoffs are, which workarounds and shadow processes nobody has mapped, who actually holds the authority to decide, what each group is rewarded for, who has access to a decision before it becomes fixed, and how suppliers, finance, operations and the people doing the work affect one another. Capabilities built before that is understood tend to automate the gap rather than close it.
The same six functions. One drawn as a sequence. One as it operates. Capabilities are built for the diagram on the left. Outcomes are decided where the strands on the right engage. Illustrative, not measured data.
In 1998, the Department of National Defence contract I worked on was running at 51% next-day delivery while every part of the operation met its own measure. The first question was not about systems or skills. It was what time of day the orders came in. Understanding how the work actually happened came first, and the technology came last. Delivery reached 97.3% and held for seven years.
Visibility is not the same as seeing
The report’s answer to resilience begins with visibility. Visibility matters, but visibility into individual functions is not necessarily visibility into their dependencies. Every function can see its own dashboard, and every dashboard can be green while the outcome is red. That is watermelon reporting. What is missing is not visibility into each part. It is an understanding of the connections between them, which is where outcomes are decided, as this week’s hurricane in the Gulf is showing in real time.
When the storm arrives, the route on the left is not what is crossed. Feedstock, power, pipelines, terminals, people and policy move on different clocks. Landfall is one moment. Power comes back later. Policy grows more active rather than returning to its old line. Illustrative, not measured data.
The missing loop
This week brought another study on the same gap, this time about AI. A new Process Context Study, from a benchmarking firm working with a process-software provider, found that 86% of respondents agree AI agents cannot be deployed reliably without understanding how the business actually works, yet only 22% have comprehensive, real-time visibility into how their processes operate. Its answer is to connect each process to the people, systems, data, rules, risks and controls behind it. That is a genuine step forward. It recognizes that decision rights, dependencies and exceptions matter, not just clean data.
But Kraljic’s matrix, the 2026 readiness report and this new study share the same missing piece: loopback learning. Each builds an understanding of the organization, whether as a categorization, a set of capabilities or a body of process context, and treats keeping that understanding current as if it were the same as getting it right. It is not. A system can refresh its data in real time and still hold the same mistaken assumption about why things happen. Context without a continuous learning loopback creates an illusion of real-world operations.
The DND finance example shows the difference. Recording the part invoices and the courier invoices accurately, in real time, would never have revealed that margin was being lost on every transaction. The data was correct. The interpretation of the purchasing and pricing rules was wrong. What exposed it was the financial consequence feeding back and changing how those rules were understood. In 1998, every order, delivery and post-delivery result was fed back into the system automatically, and today I continue to track the outcome of each decision long after it is made, in what I call the long tail assessment.
The loop has five steps, and it never stops:
That is why these instruments keep missing the operating reality. The operating reality is not a picture to be captured once and kept up to date. It moves, and the only way to keep up with it is a loop in which actual outcomes keep changing the understanding of the human and AI agents working inside it. Without that loop, every new framework, survey or context model becomes another portrait of a subject that will not hold still.
The question worth asking
Snover’s question is a good one: how do we get procurement involved earlier? Earlier engagement remains a worthwhile objective. But after 43 years, it may not be the deepest question.
The deeper question is why organizations keep leaving procurement out in the first place. Kraljic called for purchasing to be integrated with the business in 1983. Executives at a 2006 roundtable described procurement shut out of the buying decision. The 2026 report finds the same problem again. So the question I would put to Snover, and to all of us, is this: what combination of authority, incentives and access to decisions keeps producing late procurement involvement, what must change in that combination before additional capability can deliver the value it promises, and what loop will show whether the change actually worked?
Capability-building alone may not change those conditions. The next report should not only measure the gap. It should show which interventions changed it, and whether the improvement held.
Six minutes
In this short conversation, I walk through the operating reality on the right-hand side of the graphic, using the 1998 example: the four o’clock orders, the customs cutoff and the finance procedure that lost margin on every transaction. After this six-minute video, you will understand why technology initiatives fail more than they succeed.
Truth Is Believing. Accuracy Is Knowing. Outcome Is Proof.™
-30-
Share this:
Like this:
Related