A claim is only worth anything if it could have been wrong. The model I built made one that could have been — through every technology era for more than two decades. It wasn’t.
In 2005, I stood in front of a room of PTDA members in Calgary and said something that was easy to dismiss at the time: the initiatives failing all around us were not failing because of the technology. They were failing because no one had understood the process beneath the technology first.
I had been developing the models behind that claim since 1998. They said this: technology is the final piece, not the first. Outcomes are decided by the alignment of operating conditions — how human and non-human agents fit the work that already exists — not by the capability of whatever tool is purchased to address them. The technology layer was never the determinant. The condition beneath it was.
I did not arrive at that by watching the market. I built it from the work — beginning with a 1998 defence-sector engagement that forced the question, through the Metaprise™ agent-field model, Strand Commonality™, and the body of method that followed.
And it was a falsifiable claim. It could have been wrong. If the next generation of technology had delivered the outcomes its vendors promised — if better tools had, in fact, produced better results — the model would have been refuted, and rightly. The claim put itself at risk against the future.
So consider what the future did with it.
More than two decades of chances to be wrong
Since that 2005 keynote — and since the 1998 work behind it — the enterprise has been handed one “answer” after another, each arriving with the same promise: this is the technology that finally delivers.
Enterprise resource planning matured and consolidated. Software-as-a-service moved the stack to the cloud. Analytics and big data promised decisions grounded in evidence. Robotic process automation promised to remove the manual work. Digital transformation became the budget line every board approved. And now agentic AI arrives as the most capable layer of all — reasoning, not just executing.
Each of these was a genuine advance in capability. And each was a fresh opportunity for the claim to fail. If technology were the determinant, somewhere across more than two decades and six waves of it, the failure rates should have fallen. More capability should have bought better outcomes.
They didn’t fall. Across ERP, transformation, procurement technology, and now AI, independent research has reported failure rates above 50% the entire time — and recent work, including widely cited MIT research, reports that the large majority of enterprise GenAI pilots return no measurable impact to the bottom line. The most capable technology in forty years, arriving into the highest failure rates yet.
The technology changed completely. The determinant did not move.
Why the determinant didn’t move
The model explained this before the waves arrived, and the explanation has not needed revision.
Failures of this kind do not live inside the technology. They live in the seams between agents — the points where a buyer, a supplier, a courier, an approver, a system, and now an AI agent must align, and where each can be individually correct while the relationship between them is wrong. No tool fixes a seam, because the tool was never the broken part. Automating across a misaligned seam does not repair it. It accelerates it.
This is why each new wave produced the same result. ERP automated the existing conditions faster. Analytics measured them more precisely. RPA executed them more cheaply. AI now reasons over them more fluently. None of them diagnosed the condition first — and a capability applied to an undiagnosed condition does not change the outcome. It reaches the wrong outcome sooner.
The proof of the principle predates even its public articulation. In 1998, the defence-sector engagement that started it all moved on-time delivery from 51% to 97.3% in three months — sustained for seven years — with no new technology introduced at all. The lever was the alignment of order timing against fulfillment capacity: a seam, diagnosed and corrected. The technology of 1998 could not have produced that result, and it did not need to. The condition did. That engagement is where the models began.
A record you can check
The claim is not asserted from memory. It is on the record, dated, at each step.
In 2005, at a keynote in Calgary, the substrate argument was stated plainly to a room of PTDA members — that failed initiatives failed for reasons beneath the technology, and that you cannot build a solution and then ask people to fit it. The host introduced the work, on the record, as the study of enterprise outcomes “under an agent-based model versus an equation-based model.” That is a third-party timestamp on the framing, spoken aloud in 2005.
In 2007, in writing, the same logic appeared in print — clustering, the global value chain, and value-chain governance described as the coordination of agents across a field, with the agent-based model and its architecture named directly.
In 2008, a published white paper carried the mechanism into a case study: an organization with the capability to lead a new market, undone not by any deficiency in the technology but by the internal conditions that governed whether it would act.
In 2010, a lecture in London put it most plainly of all: organizations that do not understand how they actually operate will fail at any intervention — technological or otherwise.
Each of these is a dated artifact, written before the outcome was known. Hindsight is everywhere. A record of having said it first, at the time, is rare — and it is the difference between a description of an era and a model of the mechanism underneath every era.
The pattern is being rediscovered
The strongest confirmation is not mine to claim. It arrives, unprompted, from the people building the current wave.
This week, Nakshatra Gupta — a go-to-market lead at an AI-native supply chain software company — reached out. He had not read the archive and was not prompted by me. He wrote to describe, in his own words, the same pattern the models named two decades ago.
When I wrote to him about sequence across the technology eras, he replied:
“You picked up on exactly the thread I care most about. Capability is becoming commoditized; what separates organizations now is how they govern autonomous decisions, where they place accountability, and whether the operating model can actually absorb what the technology makes possible. The ‘order matters more than the vendor’ point comes straight from watching teams buy the right tools in the wrong sequence and wonder why nothing changed. Your framing of sequence across ERP, procurement tech, and digital transformation is precisely the kind of historical pattern I’m trying to learn from. Those eras rhyme more than most people in AI seem to assume, and the failure modes look familiar.”
He arrived at it independently, from inside the AI era, looking at the present. The models arrived at it in 1998, looking at a defence-sector fulfillment problem. Two routes, two decades apart, to the same place. That is not an endorsement — it is something more useful. It is convergence, and convergence on a claim that was made first, on the record, is the closest thing there is to proof that the claim was about the mechanism and not the moment.
What it means that the models held
A principle that explains one era might be a good description of that era. A principle that correctly predicts the outcome of six successive technology waves, across more than two decades, while the technology turns over completely beneath it, is something else. It is not a description of a moment. It is a model of the underlying mechanism — and the mechanism did not change, because human and organizational conditions do not turn over on the technology’s schedule.
This is the test that separates durable knowledge from topical commentary. Most of what is written about technology dates with the technology, because it is about the technology. When the wave passes, the writing passes with it. A model that was right about ERP and right about AI — for the same reason, stated the same way, two decades apart — was never about the wave. It was about the thing underneath every wave.
That is the entire claim, and it is checkable. Anyone who has worked through the last two decades can test it against their own record: every wave that arrived as the answer, every initiative that stalled for reasons that had nothing to do with the tool’s capability, every “we bought the platform and nothing changed.” The model said, before the waves arrived, that this is what would happen. It is what happened.
The technology aged. The principles did not. I built them to describe the part that doesn’t.
Technology changes capability. Substrate determines survivability. The substrate isn’t more technology — it is how humans and AI agents align with the technology that already exists. The Hansen models that say so were proven in 1998, stated on the record from 2005 onward, and confirmed by every technology wave since.
Truth Is Believing. Accuracy Is Knowing.
-30-
Related
The Models Were Right Across Every Technology Era. The Technology Kept Changing. The Determinant Didn’t.
Posted on June 10, 2026
0
A claim is only worth anything if it could have been wrong. The model I built made one that could have been — through every technology era for more than two decades. It wasn’t.
In 2005, I stood in front of a room of PTDA members in Calgary and said something that was easy to dismiss at the time: the initiatives failing all around us were not failing because of the technology. They were failing because no one had understood the process beneath the technology first.
I had been developing the models behind that claim since 1998. They said this: technology is the final piece, not the first. Outcomes are decided by the alignment of operating conditions — how human and non-human agents fit the work that already exists — not by the capability of whatever tool is purchased to address them. The technology layer was never the determinant. The condition beneath it was.
I did not arrive at that by watching the market. I built it from the work — beginning with a 1998 defence-sector engagement that forced the question, through the Metaprise™ agent-field model, Strand Commonality™, and the body of method that followed.
And it was a falsifiable claim. It could have been wrong. If the next generation of technology had delivered the outcomes its vendors promised — if better tools had, in fact, produced better results — the model would have been refuted, and rightly. The claim put itself at risk against the future.
So consider what the future did with it.
More than two decades of chances to be wrong
Since that 2005 keynote — and since the 1998 work behind it — the enterprise has been handed one “answer” after another, each arriving with the same promise: this is the technology that finally delivers.
Enterprise resource planning matured and consolidated. Software-as-a-service moved the stack to the cloud. Analytics and big data promised decisions grounded in evidence. Robotic process automation promised to remove the manual work. Digital transformation became the budget line every board approved. And now agentic AI arrives as the most capable layer of all — reasoning, not just executing.
Each of these was a genuine advance in capability. And each was a fresh opportunity for the claim to fail. If technology were the determinant, somewhere across more than two decades and six waves of it, the failure rates should have fallen. More capability should have bought better outcomes.
They didn’t fall. Across ERP, transformation, procurement technology, and now AI, independent research has reported failure rates above 50% the entire time — and recent work, including widely cited MIT research, reports that the large majority of enterprise GenAI pilots return no measurable impact to the bottom line. The most capable technology in forty years, arriving into the highest failure rates yet.
The technology changed completely. The determinant did not move.
Why the determinant didn’t move
The model explained this before the waves arrived, and the explanation has not needed revision.
Failures of this kind do not live inside the technology. They live in the seams between agents — the points where a buyer, a supplier, a courier, an approver, a system, and now an AI agent must align, and where each can be individually correct while the relationship between them is wrong. No tool fixes a seam, because the tool was never the broken part. Automating across a misaligned seam does not repair it. It accelerates it.
This is why each new wave produced the same result. ERP automated the existing conditions faster. Analytics measured them more precisely. RPA executed them more cheaply. AI now reasons over them more fluently. None of them diagnosed the condition first — and a capability applied to an undiagnosed condition does not change the outcome. It reaches the wrong outcome sooner.
The proof of the principle predates even its public articulation. In 1998, the defence-sector engagement that started it all moved on-time delivery from 51% to 97.3% in three months — sustained for seven years — with no new technology introduced at all. The lever was the alignment of order timing against fulfillment capacity: a seam, diagnosed and corrected. The technology of 1998 could not have produced that result, and it did not need to. The condition did. That engagement is where the models began.
A record you can check
The claim is not asserted from memory. It is on the record, dated, at each step.
In 2005, at a keynote in Calgary, the substrate argument was stated plainly to a room of PTDA members — that failed initiatives failed for reasons beneath the technology, and that you cannot build a solution and then ask people to fit it. The host introduced the work, on the record, as the study of enterprise outcomes “under an agent-based model versus an equation-based model.” That is a third-party timestamp on the framing, spoken aloud in 2005.
In 2007, in writing, the same logic appeared in print — clustering, the global value chain, and value-chain governance described as the coordination of agents across a field, with the agent-based model and its architecture named directly.
In 2008, a published white paper carried the mechanism into a case study: an organization with the capability to lead a new market, undone not by any deficiency in the technology but by the internal conditions that governed whether it would act.
In 2010, a lecture in London put it most plainly of all: organizations that do not understand how they actually operate will fail at any intervention — technological or otherwise.
Each of these is a dated artifact, written before the outcome was known. Hindsight is everywhere. A record of having said it first, at the time, is rare — and it is the difference between a description of an era and a model of the mechanism underneath every era.
The pattern is being rediscovered
The strongest confirmation is not mine to claim. It arrives, unprompted, from the people building the current wave.
This week, Nakshatra Gupta — a go-to-market lead at an AI-native supply chain software company — reached out. He had not read the archive and was not prompted by me. He wrote to describe, in his own words, the same pattern the models named two decades ago.
When I wrote to him about sequence across the technology eras, he replied:
“You picked up on exactly the thread I care most about. Capability is becoming commoditized; what separates organizations now is how they govern autonomous decisions, where they place accountability, and whether the operating model can actually absorb what the technology makes possible. The ‘order matters more than the vendor’ point comes straight from watching teams buy the right tools in the wrong sequence and wonder why nothing changed. Your framing of sequence across ERP, procurement tech, and digital transformation is precisely the kind of historical pattern I’m trying to learn from. Those eras rhyme more than most people in AI seem to assume, and the failure modes look familiar.”
He arrived at it independently, from inside the AI era, looking at the present. The models arrived at it in 1998, looking at a defence-sector fulfillment problem. Two routes, two decades apart, to the same place. That is not an endorsement — it is something more useful. It is convergence, and convergence on a claim that was made first, on the record, is the closest thing there is to proof that the claim was about the mechanism and not the moment.
What it means that the models held
A principle that explains one era might be a good description of that era. A principle that correctly predicts the outcome of six successive technology waves, across more than two decades, while the technology turns over completely beneath it, is something else. It is not a description of a moment. It is a model of the underlying mechanism — and the mechanism did not change, because human and organizational conditions do not turn over on the technology’s schedule.
This is the test that separates durable knowledge from topical commentary. Most of what is written about technology dates with the technology, because it is about the technology. When the wave passes, the writing passes with it. A model that was right about ERP and right about AI — for the same reason, stated the same way, two decades apart — was never about the wave. It was about the thing underneath every wave.
That is the entire claim, and it is checkable. Anyone who has worked through the last two decades can test it against their own record: every wave that arrived as the answer, every initiative that stalled for reasons that had nothing to do with the tool’s capability, every “we bought the platform and nothing changed.” The model said, before the waves arrived, that this is what would happen. It is what happened.
The technology aged. The principles did not. I built them to describe the part that doesn’t.
Technology changes capability. Substrate determines survivability. The substrate isn’t more technology — it is how humans and AI agents align with the technology that already exists. The Hansen models that say so were proven in 1998, stated on the record from 2005 onward, and confirmed by every technology wave since.
Truth Is Believing. Accuracy Is Knowing.
-30-
Share this:
Like this:
Related