Jon W. Hansen, FCIPS · Procurement Insights | Hansen AI Labs™ · October 2026
Executive summary
In 2026, the industry’s explanation for disappointing AI results is that organizations have not changed enough. McKinsey reports that only 6% of organizations qualify as AI high performers, and that nearly three-quarters of those high performers have redesigned their workflows around AI, compared with about a quarter of everyone else. Gartner reports that 78% of CHROs say workflows and roles must change to get the most out of AI investments. BCG recommends putting 70% of the effort into people and processes, 20% into technology and data, and 10% into algorithms.
None of this is new. The same remedy has accompanied every technology era I have tracked: ERP, e-procurement, suites, cloud, analytics, and now generative and agentic AI. Only the vocabulary changes. It was buy-in, then user adoption, then change readiness, then process reengineering, and now workflow redesign, role redesign and AI literacy. Underneath every version is the same assumption: the technology and the future state have already been chosen correctly, so success depends on moving the organization toward them.
That assumption is why the remedy has not changed the outcome. Change management, as it is practiced, is equation-based. It treats the organization as a single body that can be moved from one state to another if the forces for change exceed the resistance. A real organization is not one body. It is a network of agents (people, departments, suppliers, couriers, customs brokers, regulators and now algorithms), each acting rationally on its own measures. What is needed is an agent-based approach within a Metaprise™ framework: trace how those agents actually behave and where their measures cross, and only then decide what should change and whether technology belongs there.
I have the receipts. This is not a position I arrived at in 2026. It is on the dated public record, beginning in 1998:
What the industry is saying in 2026
The call for change is coming from several directions at once, and each means something slightly different by “change.”
The management consultancies mean redesigning workflows around AI. McKinsey’s 2026 State of AI found that nearly nine in ten organizations use AI in at least one business function, but only 6% qualify as high performers, and only 37% report that AI has contributed positively to earnings. McKinsey’s conclusion is that the challenge is shifting from encouraging individual AI use to redesigning workflows so that individual productivity gains become measurable financial results. BCG frames the same idea as its 10-20-70 approach, which is a recommended allocation of effort rather than a measured result.
Gartner means redesigning roles and jobs. In a December 2025 Gartner survey of 110 CHROs, 78% agreed that workflows and roles will need to change. In procurement specifically, a Gartner survey of 101 chief procurement officers found that only 36% were very confident in their ability to redesign roles and processes around AI, and Gartner observed that many procurement organizations are layering AI tools onto existing jobs.
The change-management practitioners mean individual adoption: awareness, desire, knowledge, ability and reinforcement, the stages of Prosci’s ADKAR model.
These are different activities. But they share a starting point. In each case, the technology has already been selected, and the question is how the organization should change around it.
The same remedy under a new name
Every technology era I have tracked since 1998 has followed the same sequence. The technology is presented as transformative. Implementations multiply. The results fall short. The explanation then moves from the technology to the people, the processes and the culture, and the organization is asked to adapt more thoroughly to the implementation.
The implementation failure rate has not moved with the technology. In my tracking, it has held at around 80% across ERP, e-procurement, suites, cloud, analytics and generative AI. If the remedy had worked, the rate would have changed somewhere along the way. It has been applied in every era, and the outcome has stayed the same.
That is the point of this post. The 2026 conversation about change is not a discovery. It is a recycling of a measure that has been tried, era after era, without changing the outcome of technology implementations.
Why it keeps failing: change management is an equation
The best-known model of organizational change is literally an equation. It developed from David Gleicher’s change formula in the 1960s, was published by Richard Beckhard and Reuben Harris, and was later restated by Kathie Dannemiller in the form most people know today:
D × V × F > R
Dissatisfaction with the current state, multiplied by a vision of the future, multiplied by concrete first steps, must be greater than the resistance to change. Kotter’s eight steps are a sequence to be completed in order. ADKAR is a sequence to be completed in order. Each treats change as something you calculate and then push people through, toward a future state designed in advance.
I named this directly in September 2024: the emphasis on change management resulted from the insistence on an equation-based implementation model. That is why the remedy keeps returning. If you begin with an equation, every shortfall looks like a variable that was not pushed hard enough, so the answer is always more of the same remedy.
To work, the equation has to assume several things:
- that the dissatisfaction is shared, and aimed at the real problem;
- that the vision describes an operation that will actually work;
- that the first steps address the actual causes;
- that resistance is a force standing in the way of progress;
- that the organization is one body that can be moved as a whole.
The last assumption is the one that breaks. An organization is not one body with one level of dissatisfaction, one vision and one level of resistance. It is many agents, each with its own objective, its own incentives, its own information and its own definition of success, and each agent’s behavior changes the conditions the others face. The equation asks whether the combined force for change is greater than the resistance. It never asks whether one agent’s success is defeating another agent’s outcome.
The 1998 receipt
In 1998, the services provider managing the Department of National Defence’s MRO procurement came to me saying it needed to automate. The first question I asked was what time of day the orders came in.
The service technicians were measured on how many calls they responded to. Stopping after each call to submit a part order interrupted the next call, so they held their part orders until about four in the afternoon. The late orders reached suppliers too late to make the customs cutoff, so next-day delivery failed. A part that cost $100 when ordered at nine in the morning could cost $900 when ordered at four. Next-day delivery, and with it the technicians’ ability to close their calls, stood at 51%, against a contract that required 90%.
The service provider had described it as a procurement problem, and its answer was to automate. Nobody had connected the technicians’ call-response measure to next-day delivery, because each was measured separately. Automating the procurement segment would have processed the four o’clock orders faster, and they would still have missed the customs cutoff. The same failure would simply have arrived sooner.
The technicians were not resisting anything. They were succeeding against the measure they had been given. Their local success was producing the operation’s collective failure. The agent-based question, about the time of day the orders came in, found the crossing between their measure and everyone else’s. Once that crossing was corrected, delivery performance moved from 51% to 97.3% in three months and held for seven years. The technology came after the operation had been fixed, not before.
Resistance is often evidence
In May 2007, I wrote about what organizations across Canada and the United States had told me at conferences: in the majority of cases, they had “modified its operations to accommodate the technology that was being implemented,” and they believed this was the main reason for the lack of compliance.
In December 2013, I traced the same sequence through maverick spend. Poorly designed ERP procurement platforms limited supplier engagement, so buyers went around them, because experience told them there was a better route to savings. That resistance was understandable. The industry’s response was to introduce change management to enforce compliance. The platform created the problem, and change management was brought in to make people accept it (whether it was actually beneficial or not).
Both receipts point to the same finding: the resistance came from the implementation. People were working around a system that did not fit how the work actually ran.
The equation reads resistance the other way around. In D × V × F > R, resistance is the R, a force standing in the way of the change. The equation’s answer is to push harder: raise the dissatisfaction, sharpen the vision, add more first steps, until they outweigh it. It never asks where the resistance came from. In the operations I traced, it frequently came from the implementation itself, so pushing harder meant pushing people further into a system that did not work for them.
Acres of Diamonds showed the alternative. When organizations realigned their purchasing strategy to how their goods actually behaved, they reported that they could minimize their reliance on change management and eliminate compliance issues.
Some resistance is ordinary reluctance. But some of it is an agent reporting that the design does not match reality. An equation cannot tell the two apart. Treating all of it as resistance to be managed suppresses exactly the evidence needed to correct the implementation.
What an agent-based approach within a Metaprise™ framework does instead
An agent-based approach starts with the agents rather than the future state. In practice, that means:
- Identify the agents. Include everyone whose behavior shapes the outcome, inside and outside the organization: the people doing the work, the departments, the suppliers, the couriers, the customs brokers, the regulators and the technologies themselves — including today’s dispatch AI Agents.
- Trace how each one actually behaves. Learn what each agent is measured on and what it does to succeed, including the workarounds and shadow processes that never appear on the official process map.
- Find where their measures cross. Look for the points where one agent’s success defeats another agent’s outcome, as the technicians’ call-response measure defeated next-day delivery.
- Correct the crossings first. Fix the conflicting conditions in the operation itself, then set the firm fences within which people and systems can act.
- Only then decide what technology belongs, and where. Bring the technology in after the operation has been understood, and keep learning from what happens where the strands cross.
Metaprise™ is the framework that holds this together across the whole extended enterprise, including the suppliers and external participants, rather than one department or one implementation at a time. Phase 0™ is the trace that comes first.
This is not a theory I am proposing for the first time. In the DND work, the agent-based model put technology in a supporting role behind people and processes, and change management was minimized, both within the procurement department and with the extended stakeholders, internal and external. As I wrote in June 2025, across more than 100 posts in this archive one pattern has consistently emerged: when change management is a critical part of an initiative, the initiative’s odds of success drop.
This does not mean behavioral change models have no place. As I wrote in July 2025, ADKAR can support individual adoption within a transformation whose operating basis has been understood first. The problem is sequence. Applied before the operation is traced, change management helps an organization move confidently toward a change it has not validated.
Before your next change program
Before funding another change program for an AI implementation, ask four questions:
- Whose dissatisfaction are we counting? If it belongs to executives looking at ROI, the people who must change may be meeting their own measures perfectly well, as the technicians were.
- Is the vision an operation or a technology? “AI-first” describes a tool, not how the work will run.
- Do the first steps trace the operation, or start the implementation? Training and adoption metrics begin after the frame has been set.
- Is the resistance obstruction, or evidence? Before managing it away, find out what it is telling you.
The industry keeps renaming the equation. The outcome persists because the organization was never an equation. It was always a Metaprise™.
Truth Is Believing. Accuracy Is Knowing. Outcome Is Proof.™
-30-
Related
AI Has Not Created a Change Management Problem. It Has Recycled One.
Posted on October 27, 2026
0
Jon W. Hansen, FCIPS · Procurement Insights | Hansen AI Labs™ · October 2026
Executive summary
In 2026, the industry’s explanation for disappointing AI results is that organizations have not changed enough. McKinsey reports that only 6% of organizations qualify as AI high performers, and that nearly three-quarters of those high performers have redesigned their workflows around AI, compared with about a quarter of everyone else. Gartner reports that 78% of CHROs say workflows and roles must change to get the most out of AI investments. BCG recommends putting 70% of the effort into people and processes, 20% into technology and data, and 10% into algorithms.
None of this is new. The same remedy has accompanied every technology era I have tracked: ERP, e-procurement, suites, cloud, analytics, and now generative and agentic AI. Only the vocabulary changes. It was buy-in, then user adoption, then change readiness, then process reengineering, and now workflow redesign, role redesign and AI literacy. Underneath every version is the same assumption: the technology and the future state have already been chosen correctly, so success depends on moving the organization toward them.
That assumption is why the remedy has not changed the outcome. Change management, as it is practiced, is equation-based. It treats the organization as a single body that can be moved from one state to another if the forces for change exceed the resistance. A real organization is not one body. It is a network of agents (people, departments, suppliers, couriers, customs brokers, regulators and now algorithms), each acting rationally on its own measures. What is needed is an agent-based approach within a Metaprise™ framework: trace how those agents actually behave and where their measures cross, and only then decide what should change and whether technology belongs there.
I have the receipts. This is not a position I arrived at in 2026. It is on the dated public record, beginning in 1998:
What the industry is saying in 2026
The call for change is coming from several directions at once, and each means something slightly different by “change.”
The management consultancies mean redesigning workflows around AI. McKinsey’s 2026 State of AI found that nearly nine in ten organizations use AI in at least one business function, but only 6% qualify as high performers, and only 37% report that AI has contributed positively to earnings. McKinsey’s conclusion is that the challenge is shifting from encouraging individual AI use to redesigning workflows so that individual productivity gains become measurable financial results. BCG frames the same idea as its 10-20-70 approach, which is a recommended allocation of effort rather than a measured result.
Gartner means redesigning roles and jobs. In a December 2025 Gartner survey of 110 CHROs, 78% agreed that workflows and roles will need to change. In procurement specifically, a Gartner survey of 101 chief procurement officers found that only 36% were very confident in their ability to redesign roles and processes around AI, and Gartner observed that many procurement organizations are layering AI tools onto existing jobs.
The change-management practitioners mean individual adoption: awareness, desire, knowledge, ability and reinforcement, the stages of Prosci’s ADKAR model.
These are different activities. But they share a starting point. In each case, the technology has already been selected, and the question is how the organization should change around it.
The same remedy under a new name
Every technology era I have tracked since 1998 has followed the same sequence. The technology is presented as transformative. Implementations multiply. The results fall short. The explanation then moves from the technology to the people, the processes and the culture, and the organization is asked to adapt more thoroughly to the implementation.
The implementation failure rate has not moved with the technology. In my tracking, it has held at around 80% across ERP, e-procurement, suites, cloud, analytics and generative AI. If the remedy had worked, the rate would have changed somewhere along the way. It has been applied in every era, and the outcome has stayed the same.
That is the point of this post. The 2026 conversation about change is not a discovery. It is a recycling of a measure that has been tried, era after era, without changing the outcome of technology implementations.
Why it keeps failing: change management is an equation
The best-known model of organizational change is literally an equation. It developed from David Gleicher’s change formula in the 1960s, was published by Richard Beckhard and Reuben Harris, and was later restated by Kathie Dannemiller in the form most people know today:
D × V × F > R
Dissatisfaction with the current state, multiplied by a vision of the future, multiplied by concrete first steps, must be greater than the resistance to change. Kotter’s eight steps are a sequence to be completed in order. ADKAR is a sequence to be completed in order. Each treats change as something you calculate and then push people through, toward a future state designed in advance.
I named this directly in September 2024: the emphasis on change management resulted from the insistence on an equation-based implementation model. That is why the remedy keeps returning. If you begin with an equation, every shortfall looks like a variable that was not pushed hard enough, so the answer is always more of the same remedy.
To work, the equation has to assume several things:
The last assumption is the one that breaks. An organization is not one body with one level of dissatisfaction, one vision and one level of resistance. It is many agents, each with its own objective, its own incentives, its own information and its own definition of success, and each agent’s behavior changes the conditions the others face. The equation asks whether the combined force for change is greater than the resistance. It never asks whether one agent’s success is defeating another agent’s outcome.
The 1998 receipt
In 1998, the services provider managing the Department of National Defence’s MRO procurement came to me saying it needed to automate. The first question I asked was what time of day the orders came in.
The service technicians were measured on how many calls they responded to. Stopping after each call to submit a part order interrupted the next call, so they held their part orders until about four in the afternoon. The late orders reached suppliers too late to make the customs cutoff, so next-day delivery failed. A part that cost $100 when ordered at nine in the morning could cost $900 when ordered at four. Next-day delivery, and with it the technicians’ ability to close their calls, stood at 51%, against a contract that required 90%.
The service provider had described it as a procurement problem, and its answer was to automate. Nobody had connected the technicians’ call-response measure to next-day delivery, because each was measured separately. Automating the procurement segment would have processed the four o’clock orders faster, and they would still have missed the customs cutoff. The same failure would simply have arrived sooner.
The technicians were not resisting anything. They were succeeding against the measure they had been given. Their local success was producing the operation’s collective failure. The agent-based question, about the time of day the orders came in, found the crossing between their measure and everyone else’s. Once that crossing was corrected, delivery performance moved from 51% to 97.3% in three months and held for seven years. The technology came after the operation had been fixed, not before.
Resistance is often evidence
In May 2007, I wrote about what organizations across Canada and the United States had told me at conferences: in the majority of cases, they had “modified its operations to accommodate the technology that was being implemented,” and they believed this was the main reason for the lack of compliance.
In December 2013, I traced the same sequence through maverick spend. Poorly designed ERP procurement platforms limited supplier engagement, so buyers went around them, because experience told them there was a better route to savings. That resistance was understandable. The industry’s response was to introduce change management to enforce compliance. The platform created the problem, and change management was brought in to make people accept it (whether it was actually beneficial or not).
Both receipts point to the same finding: the resistance came from the implementation. People were working around a system that did not fit how the work actually ran.
The equation reads resistance the other way around. In D × V × F > R, resistance is the R, a force standing in the way of the change. The equation’s answer is to push harder: raise the dissatisfaction, sharpen the vision, add more first steps, until they outweigh it. It never asks where the resistance came from. In the operations I traced, it frequently came from the implementation itself, so pushing harder meant pushing people further into a system that did not work for them.
Acres of Diamonds showed the alternative. When organizations realigned their purchasing strategy to how their goods actually behaved, they reported that they could minimize their reliance on change management and eliminate compliance issues.
Some resistance is ordinary reluctance. But some of it is an agent reporting that the design does not match reality. An equation cannot tell the two apart. Treating all of it as resistance to be managed suppresses exactly the evidence needed to correct the implementation.
What an agent-based approach within a Metaprise™ framework does instead
An agent-based approach starts with the agents rather than the future state. In practice, that means:
Metaprise™ is the framework that holds this together across the whole extended enterprise, including the suppliers and external participants, rather than one department or one implementation at a time. Phase 0™ is the trace that comes first.
This is not a theory I am proposing for the first time. In the DND work, the agent-based model put technology in a supporting role behind people and processes, and change management was minimized, both within the procurement department and with the extended stakeholders, internal and external. As I wrote in June 2025, across more than 100 posts in this archive one pattern has consistently emerged: when change management is a critical part of an initiative, the initiative’s odds of success drop.
This does not mean behavioral change models have no place. As I wrote in July 2025, ADKAR can support individual adoption within a transformation whose operating basis has been understood first. The problem is sequence. Applied before the operation is traced, change management helps an organization move confidently toward a change it has not validated.
Before your next change program
Before funding another change program for an AI implementation, ask four questions:
The industry keeps renaming the equation. The outcome persists because the organization was never an equation. It was always a Metaprise™.
Truth Is Believing. Accuracy Is Knowing. Outcome Is Proof.™
-30-
Share this:
Like this:
Related