The world’s largest analyst firm just made a receipt-free recommendation. I want to talk about the receipts.
Gartner published a piece this month with a diagnosis I agree with — which is not something I say often, so let me say it plainly before I say anything else.
The argument is that traditional IT operating models have become a liability in the age of AI. Not because the technology is lacking, but because the way organizations actually work can’t absorb the pace of change. The headline finding: 71 percent of CEOs say their operating models aren’t fit for the age of AI, while only 24 percent of CIOs believe their models can adapt to what the business needs.
That argument has a documented history going back nearly three decades. The determining variable in whether a technology initiative succeeds was never the technology. It was whether the organization was ready to let its actual operating reality drive the change. So when Gartner writes that the operating model — not the tool — is the constraint, there is nothing to argue with. They’ve arrived at the doorstep of something the record has been documenting since 1998.
Here is where I stop nodding.
The prescription
Having correctly identified that the operating model is the problem, Gartner’s recommended solution is to move “from IT control to orchestration” — to redesign the technology operating model around a new archetype, new governance, new decision rights.
It may even be the right direction. But it is a recommendation about the future, and I have one question that the article never answers:
Where are the receipts?
Not the survey. The survey tells you 71 percent of CEOs are uncomfortable — that’s a thermometer, and a good one. It measures the discomfort. It does not prove that the prescribed cure produces the outcome. For that you would need something the article does not contain and, I’d argue, the firm does not have: a contemporaneous record showing that organizations which made this move actually succeeded because of it — and that comparable organizations which didn’t, failed — tracked over enough years to rule out coincidence.
That record is the difference between a trend that is true today and a constant that has been proven over time. And it is the entire difference between an opinion about what to do and evidence that doing it works.
What a record actually lets you do
Here is the difference, shown through two things the archive can do that a survey cannot.
The first is a controlled comparison across two decades. In a 2008 comment thread on the blog — dated, public, still there — two companies are described adopting the same operating idea: an adaptive, networked approach to coordinating work beyond their own four walls. Cisco sustained it. By 2026 it sits among the most admired supply chains in the world, credited for exactly the approach it committed to eighteen years earlier. Boeing had the same idea, the same capability, and abandoned the behavioral discipline required to run it — and stumbled. Same blueprint. Opposite outcomes. (I walked through that case in detail in a recent post: Cisco Thrived. Boeing Didn’t.) The variable that separated them is visible because the record was written down at the time, not reconstructed afterward to fit the conclusion.
The second is the archive’s ability to surface the same determining variable operating in two cases that have nothing else in common. Virginia’s eVA state e-procurement program, documented in the record in 2007. Logitech, the consumer-electronics company that repatriated its procurement in 2020, written up at the time. Different decades, different sectors, different technologies — and the same variable deciding the outcome in both. (That pairing is its own recent post: In 2020, Logitech Repatriated Procurement Into a Pandemic.) That isn’t a trend. A trend is what is happening now. This is a constant: the same thing deciding the result across cases that share nothing else. Identifying that constant is a method, and it is the thing that turns “here is what I believe” into “here is what has held every time I have tried to break it.”
Neither of those is available to a firm working from this quarter’s survey data. Not because they aren’t capable — but because a contemporaneous record cannot be acquired after the fact. You either wrote it down at the time, across decades, in public where it can be checked against you, or you didn’t. The archive did. It is, candidly, the one thing a small, highly specialized, independent practice can hold that the largest analyst firm in the world cannot — because it was never a question of resources or headcount. It was a question of having been there, on the record, in public, the whole time.
The circularity problem
There is one more issue, and it is the one my whole method exists to avoid.
“Legacy models are failing, therefore adopt the new model” is an argument that cannot be wrong. If the new model succeeds, it is vindicated. If it fails, the answer will be that it wasn’t truly orchestration, wasn’t implemented correctly, didn’t go far enough. A claim defined by its own success can never be falsified — which means it can never really be proven either. It can only be asserted, and re-asserted, with each failure explained away.
The reason I can identify the determining variable before the outcome — and not just narrate it afterward — is that the variable is observable independently of whether the project succeeds. You can check whether an organization is willing to confront its actual operating reality before you know how the story ends. That is what makes the claim falsifiable. And a claim you can try to break, and which keeps surviving the attempt, is worth more than a recommendation that is structurally incapable of being wrong.
The receipts company
There is an irony I can’t let pass. Gartner’s entire enterprise is built on evidence — the Magic Quadrant, the Hype Cycle, the benchmark, the rating. Show me the data is the brand.
And yet on the single most important claim in this article — that moving from control to orchestration will produce the outcome — there is no longitudinal proof offered. There is a snapshot of how uncomfortable executives feel, and a prescription. The receipts company, on this one, made a receipt-free recommendation.
I’m not saying they’re wrong about the direction. I’m saying you should ask the same question of their recommendation that you’d ask of mine, or anyone’s: where is the evidence that this actually produces the result, tracked over enough time and enough cases that it can’t be coincidence — and is it written somewhere it could have proven the opposite?
The archive can answer that. It has been published openly since 2007, against work reaching back to 1998 — nearly three decades of contemporaneous observation, gathered in one place rather than created there, through an unwavering commitment to get it right rather than be right. The record was built to be disconfirmed — published in the open, dated, where a broken pattern would show plainly. That it hasn’t broken is not a matter of cleverness. It is because the variable underneath is real, and the record was and is honest enough to surface an error if it weren’t. The receipts are public. Check them.
Move to orchestration if the evidence supports it. But ask for the evidence. In an age where everyone is selling a model for the age of AI, the only thing worth trusting is the one with receipts.
In a follow-up, I’ll take this one step further — past whether to trust a method and into what AI actually does to an organization once it arrives. Because AI isn’t the model that fixes you. It’s the pressure that reveals whether you were ever ready. More on that next.
Truth Is Believing. Accuracy Is Knowing.
-30-
Related
Gartner Says Move From IT Control to Orchestration. Where Are the Receipts?
Posted on June 28, 2026
0
The world’s largest analyst firm just made a receipt-free recommendation. I want to talk about the receipts.
Gartner published a piece this month with a diagnosis I agree with — which is not something I say often, so let me say it plainly before I say anything else.
The argument is that traditional IT operating models have become a liability in the age of AI. Not because the technology is lacking, but because the way organizations actually work can’t absorb the pace of change. The headline finding: 71 percent of CEOs say their operating models aren’t fit for the age of AI, while only 24 percent of CIOs believe their models can adapt to what the business needs.
That argument has a documented history going back nearly three decades. The determining variable in whether a technology initiative succeeds was never the technology. It was whether the organization was ready to let its actual operating reality drive the change. So when Gartner writes that the operating model — not the tool — is the constraint, there is nothing to argue with. They’ve arrived at the doorstep of something the record has been documenting since 1998.
Here is where I stop nodding.
The prescription
Having correctly identified that the operating model is the problem, Gartner’s recommended solution is to move “from IT control to orchestration” — to redesign the technology operating model around a new archetype, new governance, new decision rights.
It may even be the right direction. But it is a recommendation about the future, and I have one question that the article never answers:
Where are the receipts?
Not the survey. The survey tells you 71 percent of CEOs are uncomfortable — that’s a thermometer, and a good one. It measures the discomfort. It does not prove that the prescribed cure produces the outcome. For that you would need something the article does not contain and, I’d argue, the firm does not have: a contemporaneous record showing that organizations which made this move actually succeeded because of it — and that comparable organizations which didn’t, failed — tracked over enough years to rule out coincidence.
That record is the difference between a trend that is true today and a constant that has been proven over time. And it is the entire difference between an opinion about what to do and evidence that doing it works.
What a record actually lets you do
Here is the difference, shown through two things the archive can do that a survey cannot.
The first is a controlled comparison across two decades. In a 2008 comment thread on the blog — dated, public, still there — two companies are described adopting the same operating idea: an adaptive, networked approach to coordinating work beyond their own four walls. Cisco sustained it. By 2026 it sits among the most admired supply chains in the world, credited for exactly the approach it committed to eighteen years earlier. Boeing had the same idea, the same capability, and abandoned the behavioral discipline required to run it — and stumbled. Same blueprint. Opposite outcomes. (I walked through that case in detail in a recent post: Cisco Thrived. Boeing Didn’t.) The variable that separated them is visible because the record was written down at the time, not reconstructed afterward to fit the conclusion.
The second is the archive’s ability to surface the same determining variable operating in two cases that have nothing else in common. Virginia’s eVA state e-procurement program, documented in the record in 2007. Logitech, the consumer-electronics company that repatriated its procurement in 2020, written up at the time. Different decades, different sectors, different technologies — and the same variable deciding the outcome in both. (That pairing is its own recent post: In 2020, Logitech Repatriated Procurement Into a Pandemic.) That isn’t a trend. A trend is what is happening now. This is a constant: the same thing deciding the result across cases that share nothing else. Identifying that constant is a method, and it is the thing that turns “here is what I believe” into “here is what has held every time I have tried to break it.”
Neither of those is available to a firm working from this quarter’s survey data. Not because they aren’t capable — but because a contemporaneous record cannot be acquired after the fact. You either wrote it down at the time, across decades, in public where it can be checked against you, or you didn’t. The archive did. It is, candidly, the one thing a small, highly specialized, independent practice can hold that the largest analyst firm in the world cannot — because it was never a question of resources or headcount. It was a question of having been there, on the record, in public, the whole time.
The circularity problem
There is one more issue, and it is the one my whole method exists to avoid.
“Legacy models are failing, therefore adopt the new model” is an argument that cannot be wrong. If the new model succeeds, it is vindicated. If it fails, the answer will be that it wasn’t truly orchestration, wasn’t implemented correctly, didn’t go far enough. A claim defined by its own success can never be falsified — which means it can never really be proven either. It can only be asserted, and re-asserted, with each failure explained away.
The reason I can identify the determining variable before the outcome — and not just narrate it afterward — is that the variable is observable independently of whether the project succeeds. You can check whether an organization is willing to confront its actual operating reality before you know how the story ends. That is what makes the claim falsifiable. And a claim you can try to break, and which keeps surviving the attempt, is worth more than a recommendation that is structurally incapable of being wrong.
The receipts company
There is an irony I can’t let pass. Gartner’s entire enterprise is built on evidence — the Magic Quadrant, the Hype Cycle, the benchmark, the rating. Show me the data is the brand.
And yet on the single most important claim in this article — that moving from control to orchestration will produce the outcome — there is no longitudinal proof offered. There is a snapshot of how uncomfortable executives feel, and a prescription. The receipts company, on this one, made a receipt-free recommendation.
I’m not saying they’re wrong about the direction. I’m saying you should ask the same question of their recommendation that you’d ask of mine, or anyone’s: where is the evidence that this actually produces the result, tracked over enough time and enough cases that it can’t be coincidence — and is it written somewhere it could have proven the opposite?
The archive can answer that. It has been published openly since 2007, against work reaching back to 1998 — nearly three decades of contemporaneous observation, gathered in one place rather than created there, through an unwavering commitment to get it right rather than be right. The record was built to be disconfirmed — published in the open, dated, where a broken pattern would show plainly. That it hasn’t broken is not a matter of cleverness. It is because the variable underneath is real, and the record was and is honest enough to surface an error if it weren’t. The receipts are public. Check them.
Move to orchestration if the evidence supports it. But ask for the evidence. In an age where everyone is selling a model for the age of AI, the only thing worth trusting is the one with receipts.
In a follow-up, I’ll take this one step further — past whether to trust a method and into what AI actually does to an organization once it arrives. Because AI isn’t the model that fixes you. It’s the pressure that reveals whether you were ever ready. More on that next.
Truth Is Believing. Accuracy Is Knowing.
-30-
Share this:
Like this:
Related