Phil Fersht is right that the services pyramid is collapsing into something better. But the synchronous structure he is describing is one I was mapping as the Metaprise model nearly thirty years ago — and it carried a warning the Doer Model still needs: put AI at the center of a misaligned system and you do not get a doer model. You get a pyramid that fails faster.
Phil Fersht has written the clearest articulation I have seen of where professional services is heading, and I want to start by saying that plainly, because what follows is an extension of his argument, not a quarrel with it. His June 12 piece, “Stop Mourning the Pyramid and Build the Doer Model Instead,” makes the case that the old services pyramid — a thin layer of senior architects, a band of coordinating managers, and a wide base of low-margin juniors doing the grunt work — is breaking under AI. The firms still defending it, he argues, are not being cautious but falling behind. In its place he sees a “Doer Model”: AI owns the execution layer, juniors move straight to interpretation and judgment, mid-level managers stop being relay stations and become player-coaches, and senior leaders finally get information that has not been filtered through four layers of organizational politics. He is right on the structure, right on the sequencing, and right on the discipline it demands. I would not change a word of his diagnosis.
What I want to add is the thing my own work has been circling for twenty-seven years, because Phil has described — precisely and independently — a structure I drew a very long time ago, and the version I drew came with a condition attached that his framing implies but does not yet name.
Look at what the Doer Model actually is underneath the labor economics. The pyramid is not really a headcount shape. It is an information architecture. Work enters at the bottom, gets processed and passed up through layers, each one aggregating, translating, and filtering, until a partner at the top makes a decision on information that is already stale and already distorted. That is a sequential architecture — a relay, hand-off after hand-off, value and fidelity leaking at every seam. The Doer Model replaces it with something else: AI in the center, the human roles clustered around it, information moving to where it is needed in something much closer to real time. That is a synchronous architecture — a hub, not a chain.
And the distinction matters more than it first appears, because the pyramid and the old sequential pipeline share one hidden assumption: that if each box does its job, the whole will add up. It rarely does. The failure does not live inside any box — not in the juniors, not in the managers, not in the partners. It lives in the seams between them, where incentives, timing, and decision rights meet and quietly contradict one another. That is the part the org chart cannot show you, and it is why redrawing the boxes — even into a better shape — does not, by itself, touch the thing that was actually failing.
I have drawn that exact transition before — and not recently. In 1999 I set it out in a Metaprise™ brochure: “sequential” enterprise applications, supplier to manufacturer to reseller each handing off down a line, contrasted against “synchronous” applications in which every participant coordinates through a shared hub rather than a chain. Phil’s “relay station” is my sequential hand-off. Phil’s AI-in-the-center is my hub. His pyramid-to-doer transition and my sequential-to-synchronous transition are the same structural move, described from two different directions — his from the labor economics of professional services, mine from the information architecture of the supply chain.
The Metaprise™ model (1999): enterprise applications wired as a sequential chain — supplier to manufacturer to reseller, each handing off to the next — contrasted with a synchronous architecture in which every participant coordinates through a shared hub. The diagram predates the technology that would eventually make the hub buildable.
Here is the part that matters, and I want to be clear about the spirit in which I say it, because it is not a complaint. For the better part of two decades, that synchronous model gained almost no traction. The field organized itself around the sequential pyramid, and the hub sat largely unused — not because it was wrong, but because the technology that would make it buildable did not yet exist. Then, recently, it began showing up on its own. ConvergentIS was among the first I saw mirror the structure; now Phil has arrived at the same shape from the labor side. I am genuinely glad they did. When capable people independently evolve toward a structure you drew before the enabling technology existed, that is not something to be possessive about — it is the strongest evidence you could ask for that the structure was never about the technology. An idea that catches on immediately might just be fashionable. An idea ignored for twenty years and then reached independently, from multiple directions, once the conditions allow it, was structural all along.
Credit: ConvergentIS, used with permission.
Which brings me to the condition the Doer Model still needs, and it is hiding inside Phil’s own best sentence. He writes that under the new model, senior leaders get cleaner information not because they got smarter, but because “the information architecture finally got honest.” That line is exactly right, and it is exactly the place to be careful, because an information architecture does not get honest simply by putting AI in the middle of it.
Here is the trap. If you take a system whose seams are misaligned — whose internal and external agents, human and AI alike, hold incentives that diverge, whose decision rights are unclear, whose functions are optimizing for different things — and you insert AI at the center as the new hub, you do not get honesty. You get synchronized dysfunction. The misalignment that used to travel slowly up through the relay now travels at machine speed through the hub. Orchestration does not resolve an underlying substrate inconsistency; it just moves the inconsistency faster. The hub is necessary. It is not sufficient. What determines whether the Doer Model delivers clean information or merely accelerates the distortion is not the architecture itself — it is whether the conditions beneath the architecture are aligned before the AI is dropped into the center.
Look at the AI node sitting in the middle of Phil’s “Doer” triad — leaders, mid-level player-coaches, and juniors arranged around a central AI. That node is the whole question in one symbol. Put it at the center of an aligned system and it does exactly what he describes: it elevates everyone, because clean information flows to where it is needed. Put the identical node at the center of a misaligned one and it elevates nothing — it just distributes the misalignment faster and more confidently than the old pyramid ever could. The diagram is the same in both cases. What is different is the floor underneath it, and the floor is the one thing the diagram cannot show you.
Phil actually knows this, which is why his piece is better than most. His strongest argument is that the move from Pyramid to Doer is “a sequenced capability build,” and that organizations which restructure headcount before building the new capability “end up with a smaller pyramid, not a Doer Model.” That is a substrate-before-technology argument, and it is correct. I am only naming the mechanism underneath it. The reason sequence matters is that the technology harvests whatever conditions it lands on. Drop it onto an aligned substrate and it produces the clean, fast, honest information architecture Phil describes. Drop it onto one whose conditions still conflict and it produces a faster, more expensive failure, because now the bad information is synchronized.
Let’s revisit the Department of National Defence engagement in the late 1990s. Delivery performance was failing, and every instinct said to add systems and process more orders. The breakthrough was not more technology. It was discovering an interaction nobody believed belonged inside the system at all: a service-department metric, optimized in good faith, was quietly distorting order timing — and order timing governed procurement, which governed the sourcing process, which governed delivery. The number that looked successful was breaking the number that was failing, and no one had seen it because no one treated those functions as a single system. Align that one seam, with no new technology, and delivery performance moved from 51% to 97.3%. Only then was the technology introduced, onto a substrate that could finally support it, and it reduced the collective buyer FTE count from twenty-three to three and sustained the result for seven consecutive years. The technology did not fix the system. Aligning the conditions let the technology work. Reverse that order and you get the generational failure that has repeated from ERP through to AI.
The Doer Model is the technology-and-structure half of that lesson, stated beautifully. The determinant — the conditions beneath — is the other half, and it is the half that decides which firms get the Doer Model and which ones get a synchronous pyramid that quietly fails inside three years, exactly as Phil warns.
So: stop mourning the pyramid. Phil is right. The shape he is pointing at is the right shape, and it is the shape I have believed in for nearly three decades. Just do not mistake the shape for the fix. The pyramid is collapsing into a hub. Whether that hub produces honesty or accelerated dysfunction will not be decided by the AI in the center. It will be decided, as it always has been, by whether anyone aligned the floor before they built on it.
-30-
Truth Is Believing. Accuracy Is Knowing.
Jon Hansen is the creator of Implementation Physics™, a research-based framework developed over nearly three decades to explain why technology initiatives succeed or fail regardless of the technology being deployed. His work spans six technology generations — from ERP through Agentic AI — and includes the Metaprise™ model first articulated in the late 1990s. His research forms the foundation for the Hansen Method™, Hansen Fit Score™ (HFS™), Phase 0™ Readiness Assessment, and the ARA™ RAM 2025™ multimodel verification architecture. He currently serves as a Board Member of the CIPS Americas Chapter.
With acknowledgment to Phil Fersht and HFS Research, whose “Doer Model” framing prompted this piece.
The Doer Model Is the Right Shape. Whether It Holds Depends on the Floor.
Posted on June 15, 2026
0
Phil Fersht is right that the services pyramid is collapsing into something better. But the synchronous structure he is describing is one I was mapping as the Metaprise model nearly thirty years ago — and it carried a warning the Doer Model still needs: put AI at the center of a misaligned system and you do not get a doer model. You get a pyramid that fails faster.
Phil Fersht has written the clearest articulation I have seen of where professional services is heading, and I want to start by saying that plainly, because what follows is an extension of his argument, not a quarrel with it. His June 12 piece, “Stop Mourning the Pyramid and Build the Doer Model Instead,” makes the case that the old services pyramid — a thin layer of senior architects, a band of coordinating managers, and a wide base of low-margin juniors doing the grunt work — is breaking under AI. The firms still defending it, he argues, are not being cautious but falling behind. In its place he sees a “Doer Model”: AI owns the execution layer, juniors move straight to interpretation and judgment, mid-level managers stop being relay stations and become player-coaches, and senior leaders finally get information that has not been filtered through four layers of organizational politics. He is right on the structure, right on the sequencing, and right on the discipline it demands. I would not change a word of his diagnosis.
Credit: Phil Fersht / HFS Research (© 2026 HFS Research). Link back to the source post: “Stop Mourning the Pyramid and Build the Doer Model Instead.”
What I want to add is the thing my own work has been circling for twenty-seven years, because Phil has described — precisely and independently — a structure I drew a very long time ago, and the version I drew came with a condition attached that his framing implies but does not yet name.
Look at what the Doer Model actually is underneath the labor economics. The pyramid is not really a headcount shape. It is an information architecture. Work enters at the bottom, gets processed and passed up through layers, each one aggregating, translating, and filtering, until a partner at the top makes a decision on information that is already stale and already distorted. That is a sequential architecture — a relay, hand-off after hand-off, value and fidelity leaking at every seam. The Doer Model replaces it with something else: AI in the center, the human roles clustered around it, information moving to where it is needed in something much closer to real time. That is a synchronous architecture — a hub, not a chain.
And the distinction matters more than it first appears, because the pyramid and the old sequential pipeline share one hidden assumption: that if each box does its job, the whole will add up. It rarely does. The failure does not live inside any box — not in the juniors, not in the managers, not in the partners. It lives in the seams between them, where incentives, timing, and decision rights meet and quietly contradict one another. That is the part the org chart cannot show you, and it is why redrawing the boxes — even into a better shape — does not, by itself, touch the thing that was actually failing.
I have drawn that exact transition before — and not recently. In 1999 I set it out in a Metaprise™ brochure: “sequential” enterprise applications, supplier to manufacturer to reseller each handing off down a line, contrasted against “synchronous” applications in which every participant coordinates through a shared hub rather than a chain. Phil’s “relay station” is my sequential hand-off. Phil’s AI-in-the-center is my hub. His pyramid-to-doer transition and my sequential-to-synchronous transition are the same structural move, described from two different directions — his from the labor economics of professional services, mine from the information architecture of the supply chain.
The Metaprise™ model (1999): enterprise applications wired as a sequential chain — supplier to manufacturer to reseller, each handing off to the next — contrasted with a synchronous architecture in which every participant coordinates through a shared hub. The diagram predates the technology that would eventually make the hub buildable.
Here is the part that matters, and I want to be clear about the spirit in which I say it, because it is not a complaint. For the better part of two decades, that synchronous model gained almost no traction. The field organized itself around the sequential pyramid, and the hub sat largely unused — not because it was wrong, but because the technology that would make it buildable did not yet exist. Then, recently, it began showing up on its own. ConvergentIS was among the first I saw mirror the structure; now Phil has arrived at the same shape from the labor side. I am genuinely glad they did. When capable people independently evolve toward a structure you drew before the enabling technology existed, that is not something to be possessive about — it is the strongest evidence you could ask for that the structure was never about the technology. An idea that catches on immediately might just be fashionable. An idea ignored for twenty years and then reached independently, from multiple directions, once the conditions allow it, was structural all along.
Credit: ConvergentIS, used with permission.
Which brings me to the condition the Doer Model still needs, and it is hiding inside Phil’s own best sentence. He writes that under the new model, senior leaders get cleaner information not because they got smarter, but because “the information architecture finally got honest.” That line is exactly right, and it is exactly the place to be careful, because an information architecture does not get honest simply by putting AI in the middle of it.
Here is the trap. If you take a system whose seams are misaligned — whose internal and external agents, human and AI alike, hold incentives that diverge, whose decision rights are unclear, whose functions are optimizing for different things — and you insert AI at the center as the new hub, you do not get honesty. You get synchronized dysfunction. The misalignment that used to travel slowly up through the relay now travels at machine speed through the hub. Orchestration does not resolve an underlying substrate inconsistency; it just moves the inconsistency faster. The hub is necessary. It is not sufficient. What determines whether the Doer Model delivers clean information or merely accelerates the distortion is not the architecture itself — it is whether the conditions beneath the architecture are aligned before the AI is dropped into the center.
A Pause Reference Point: Tell me how Agentic AI would have known to ask, “What time of day do orders come in?”
Look at the AI node sitting in the middle of Phil’s “Doer” triad — leaders, mid-level player-coaches, and juniors arranged around a central AI. That node is the whole question in one symbol. Put it at the center of an aligned system and it does exactly what he describes: it elevates everyone, because clean information flows to where it is needed. Put the identical node at the center of a misaligned one and it elevates nothing — it just distributes the misalignment faster and more confidently than the old pyramid ever could. The diagram is the same in both cases. What is different is the floor underneath it, and the floor is the one thing the diagram cannot show you.
Phil actually knows this, which is why his piece is better than most. His strongest argument is that the move from Pyramid to Doer is “a sequenced capability build,” and that organizations which restructure headcount before building the new capability “end up with a smaller pyramid, not a Doer Model.” That is a substrate-before-technology argument, and it is correct. I am only naming the mechanism underneath it. The reason sequence matters is that the technology harvests whatever conditions it lands on. Drop it onto an aligned substrate and it produces the clean, fast, honest information architecture Phil describes. Drop it onto one whose conditions still conflict and it produces a faster, more expensive failure, because now the bad information is synchronized.
Let’s revisit the Department of National Defence engagement in the late 1990s. Delivery performance was failing, and every instinct said to add systems and process more orders. The breakthrough was not more technology. It was discovering an interaction nobody believed belonged inside the system at all: a service-department metric, optimized in good faith, was quietly distorting order timing — and order timing governed procurement, which governed the sourcing process, which governed delivery. The number that looked successful was breaking the number that was failing, and no one had seen it because no one treated those functions as a single system. Align that one seam, with no new technology, and delivery performance moved from 51% to 97.3%. Only then was the technology introduced, onto a substrate that could finally support it, and it reduced the collective buyer FTE count from twenty-three to three and sustained the result for seven consecutive years. The technology did not fix the system. Aligning the conditions let the technology work. Reverse that order and you get the generational failure that has repeated from ERP through to AI.
The Doer Model is the technology-and-structure half of that lesson, stated beautifully. The determinant — the conditions beneath — is the other half, and it is the half that decides which firms get the Doer Model and which ones get a synchronous pyramid that quietly fails inside three years, exactly as Phil warns.
So: stop mourning the pyramid. Phil is right. The shape he is pointing at is the right shape, and it is the shape I have believed in for nearly three decades. Just do not mistake the shape for the fix. The pyramid is collapsing into a hub. Whether that hub produces honesty or accelerated dysfunction will not be decided by the AI in the center. It will be decided, as it always has been, by whether anyone aligned the floor before they built on it.
-30-
Truth Is Believing. Accuracy Is Knowing.
Jon Hansen is the creator of Implementation Physics™, a research-based framework developed over nearly three decades to explain why technology initiatives succeed or fail regardless of the technology being deployed. His work spans six technology generations — from ERP through Agentic AI — and includes the Metaprise™ model first articulated in the late 1990s. His research forms the foundation for the Hansen Method™, Hansen Fit Score™ (HFS™), Phase 0™ Readiness Assessment, and the ARA™ RAM 2025™ multimodel verification architecture. He currently serves as a Board Member of the CIPS Americas Chapter.
With acknowledgment to Phil Fersht and HFS Research, whose “Doer Model” framing prompted this piece.
Share this:
Like this:
Related