Truth Is Believing. Accuracy Is Knowing. Outcome Is Proof.™
McKinsey has just documented, in luxury retail, an operating principle this archive has been recording since 2004. The technology keeps changing. The principle does not.
In a report published in May 2026, McKinsey set out how artificial intelligence is reshaping luxury retail. One line carried the whole argument: in an agent-mediated world, “the first interpretation of a consumer’s desire is occurring outside the brand’s universe.“
It is an important observation. What struck me was not that it was new — it was that I had read its structure before, many times, in my own archive, across technologies that had nothing to do with luxury retail, and nothing to do with AI.
This is not a story about luxury. It is not a story about Amazon. It is not, in the end, even a story about AI. It is a story about where decisions begin — and about a principle that has stayed fixed while the technology beneath it has turned over four times.
Strip the setting from McKinsey’s sentence and it is exactly the shape of a sourcing need, a requisition, or a service request whose first interpretation now happens outside the enterprise’s systems — inside a personal AI, before anyone opens a platform. The domain is luxury retail. The structure is the one I have been writing about in procurement since 2004.
The principle was written down before the technology existed
In the fall of 2004, in a paper titled Acres of Diamonds, I set down a position I have never since had cause to revise: that a true centralization of procurement objectives requires a decentralized architecture built on the real-world operating attributes of all transactional stakeholders. I called it the cornerstone of agent-based modeling. In 2004, “agent-based” had nothing to do with today’s AI agents. It described something more durable — that organizations make better decisions by aligning with the real operating behavior of distributed participants than by imposing control from the center.
That principle has not moved since. Only the technology has.
2007 — Cisco demonstrated it
Three years later, writing about Cisco’s adaptive supply chain, I did something worth noticing now: I reached back and quoted that same 2004 passage, by name. Cisco wasn’t winning on common systems; it was winning on common objectives — synchronized processes, shared visibility, a single source of truth across distributed participants. The technology was becoming the enabling layer; the operating model was the determining one. I did not need a new idea to describe Cisco in 2007 — the 2004 one fit exactly, which is the first time the record shows the principle outliving the moment that produced it.
2014 — Amazon moved the door
In September 2014, I asked whether Amazon was the new Oracle — whether the arrival of the B2C world marked the end of the ERP era. The real question underneath was not which software would win, but that people were beginning to arrive at purchasing decisions through experiences that started outside the enterprise’s systems. The front door had begun to move. In 2023 I returned to that piece, because by then procurement was living what it had only described.
2026 — Personal AI institutionalizes it
More recently, I wrote that Personal AI is becoming the new Shadow Spreadsheet: employees increasingly begin their work inside a personal AI before they open an enterprise application; buyers frame a sourcing question inside ChatGPT, Gemini, or Copilot before they ever open a procurement platform. The enterprise no longer controls the first interpretation. That is the shift — and it is the same shift, in new clothes.
And now McKinsey, in luxury
Set their finding — that desire is now first interpreted somewhere the brand does not control — beside the record:
- 2004: decision effectiveness comes from decentralized architectures aligned with real operating behavior.
- 2007: Cisco demonstrates adaptive synchronization across distributed participants.
- 2014: Amazon changes where purchasing decisions begin.
- 2023: procurement lives the same migration.
- 2026: Personal AI becomes the first operating layer for the individual.
- 2026: McKinsey, independently, documents the same upstream migration in luxury retail.
Different industries. Different technologies. One governing principle. McKinsey reached it from luxury; I reached it from procurement, twenty-two years earlier — and neither of us invented it. We are both describing something that was already true.
This is what I mean by Invariant Physics™
For years I described these observations with different words — adaptive architecture, agent-based modeling, Metaprise™, synchronization, stakeholder operating attributes, implementation readiness. I now describe them together as Invariant Physics™, because the archive keeps demonstrating the same thing: the technologies change; the operating principle does not. ERP, e-commerce, cloud, mobile, agentic AI, personal AI — each is one more manifestation of a single structural reality. The place where intent is first interpreted increasingly determines every decision that follows.
Which is why optimizing where a transaction finishes — the requisition, the checkout, the contract — is increasingly solving yesterday’s problem. Advantage is moving to whoever understands, and governs, where interpretation begins. In investment terms, that means underwriting not only a company’s systems of record, but its systems of first interpretation. That is not a forecast about AI. It is a description of a principle that AI has merely made impossible to ignore.
Why a well-thought-out program tanks
I am not claiming I predicted any of this. I am claiming something more durable, and more useful: the archive documents a governing operating principle across twenty-two years of technological change.
And the principle does not only describe where decisions begin. It also determines whether the programs meant to serve those decisions succeed or fail — because a program that misreads where the operating reality actually lives will fail no matter how well it is designed. The record contains an early, dated example of exactly that.
In 2004 I researched North Carolina’s At Your Service procurement platform. What impressed me had nothing to do with the platform. It was a Memorandum of Understanding the State signed with its universities, built on what I called collaborative autonomy: institutions kept their own purchasing, their bids were cross-referenced against central contracts, and best value won — from either source. It was the right instrument, a design built to earn cooperation rather than force it.
The program still tanked. Not because the governance was wrong — because when the time came to make it work, they looked at the wrong layer: at the technology’s architecture and integration rather than at how the agents actually operated in the real world. They tried to understand compliance as an interface problem, when compliance was a point-of-capture problem — a question of operating reality, not integration.
That is the exact mirror of what worked at Canada’s Department of National Defence. There, the parameter that reset the entire program wasn’t found in any architecture; it was found at the point of capture, in a plain question — what time of day was an order placed, and what did that do to cost and delivery? That single question exposed a behavioral driver hiding in plain sight: orders were being timed to how people were measured, not to when the goods were actually needed — and no architecture diagram would ever have surfaced it. Technology was the last piece introduced, not the first, and only after the framework was already in place. Same governance ambition in both cases. Opposite place to look. Opposite result. The variable was never the platform, and never the MOU. It was where each program went to understand how the work actually happened.
Which is the mistake every AI-first response is now poised to repeat. People are already defaulting to personal AI, and they are doing it for a point-of-capture reason: it improves how they actually perform, in the moment of the work. Try to win them back at the interface — with governance rules, compliance mandates, blocking, integration diagrams — and you are making North Carolina’s error at scale, treating a capture-reality problem as an interface problem. You will get compliance on paper and people routing around you, into the AI you don’t govern. Understand the capture reality instead — how the AI and the governance model affect their performance where the work is actually done — and you get DND’s result: the framework earns the cooperation, and compliance follows from alignment rather than standing in for it.
AI did not begin this principle. AI is simply the latest technology through which it became unmistakable — and the latest test of whether organizations will read the shift at the interface, or at the point of capture where it actually lives.
One Additional Note: Where the Language of Technology Merges with the Language of Procurement
Years ago I read an article by Paula Hodgins of HPE Canada, titled “To Succeed in the Digital Age, You Need to Work at the Edge,” and it has stayed with me ever since. Her subject was edge computing — the case for processing data at the point of capture, on the device where the event actually happens, rather than pushing everything back to the cloud to be interpreted somewhere else. I have never forgotten it, because it names, in the language of infrastructure, the exact principle this archive has been describing in the language of procurement. The technologists call it the edge. I have been calling it the point of capture. They are the same place, reached from opposite directions — which is what you would expect if the principle underneath is invariant rather than domain-specific.
You can debate whether a model is right. You cannot debate the dated record it was built on.
This analysis draws on the Procurement Insights archive — an independent record, carrying zero vendor sponsorships, that I have published openly since 2007 and that consolidates documented client work, lectures, and writing reaching back to 1998. Every claim in it is held to the Provenance Ledger™: a verify-before-publish discipline that traces each assertion to a primary source and reconciles the record forward rather than editing it in place. Invariant Physics™ is the constant it keeps testing — that however far the technology advances, the operating logic must be in place first. Getting it right, rather than being right.
-30-
Related
McKinsey Documents in Luxury What This Archive Has Recorded Since 2004: It’s Time to Meet AI at the Edge
Posted on July 20, 2026
0
Truth Is Believing. Accuracy Is Knowing. Outcome Is Proof.™
McKinsey has just documented, in luxury retail, an operating principle this archive has been recording since 2004. The technology keeps changing. The principle does not.
In a report published in May 2026, McKinsey set out how artificial intelligence is reshaping luxury retail. One line carried the whole argument: in an agent-mediated world, “the first interpretation of a consumer’s desire is occurring outside the brand’s universe.“
It is an important observation. What struck me was not that it was new — it was that I had read its structure before, many times, in my own archive, across technologies that had nothing to do with luxury retail, and nothing to do with AI.
This is not a story about luxury. It is not a story about Amazon. It is not, in the end, even a story about AI. It is a story about where decisions begin — and about a principle that has stayed fixed while the technology beneath it has turned over four times.
Strip the setting from McKinsey’s sentence and it is exactly the shape of a sourcing need, a requisition, or a service request whose first interpretation now happens outside the enterprise’s systems — inside a personal AI, before anyone opens a platform. The domain is luxury retail. The structure is the one I have been writing about in procurement since 2004.
The principle was written down before the technology existed
In the fall of 2004, in a paper titled Acres of Diamonds, I set down a position I have never since had cause to revise: that a true centralization of procurement objectives requires a decentralized architecture built on the real-world operating attributes of all transactional stakeholders. I called it the cornerstone of agent-based modeling. In 2004, “agent-based” had nothing to do with today’s AI agents. It described something more durable — that organizations make better decisions by aligning with the real operating behavior of distributed participants than by imposing control from the center.
That principle has not moved since. Only the technology has.
2007 — Cisco demonstrated it
Three years later, writing about Cisco’s adaptive supply chain, I did something worth noticing now: I reached back and quoted that same 2004 passage, by name. Cisco wasn’t winning on common systems; it was winning on common objectives — synchronized processes, shared visibility, a single source of truth across distributed participants. The technology was becoming the enabling layer; the operating model was the determining one. I did not need a new idea to describe Cisco in 2007 — the 2004 one fit exactly, which is the first time the record shows the principle outliving the moment that produced it.
2014 — Amazon moved the door
In September 2014, I asked whether Amazon was the new Oracle — whether the arrival of the B2C world marked the end of the ERP era. The real question underneath was not which software would win, but that people were beginning to arrive at purchasing decisions through experiences that started outside the enterprise’s systems. The front door had begun to move. In 2023 I returned to that piece, because by then procurement was living what it had only described.
2026 — Personal AI institutionalizes it
More recently, I wrote that Personal AI is becoming the new Shadow Spreadsheet: employees increasingly begin their work inside a personal AI before they open an enterprise application; buyers frame a sourcing question inside ChatGPT, Gemini, or Copilot before they ever open a procurement platform. The enterprise no longer controls the first interpretation. That is the shift — and it is the same shift, in new clothes.
And now McKinsey, in luxury
Set their finding — that desire is now first interpreted somewhere the brand does not control — beside the record:
Different industries. Different technologies. One governing principle. McKinsey reached it from luxury; I reached it from procurement, twenty-two years earlier — and neither of us invented it. We are both describing something that was already true.
This is what I mean by Invariant Physics™
For years I described these observations with different words — adaptive architecture, agent-based modeling, Metaprise™, synchronization, stakeholder operating attributes, implementation readiness. I now describe them together as Invariant Physics™, because the archive keeps demonstrating the same thing: the technologies change; the operating principle does not. ERP, e-commerce, cloud, mobile, agentic AI, personal AI — each is one more manifestation of a single structural reality. The place where intent is first interpreted increasingly determines every decision that follows.
Which is why optimizing where a transaction finishes — the requisition, the checkout, the contract — is increasingly solving yesterday’s problem. Advantage is moving to whoever understands, and governs, where interpretation begins. In investment terms, that means underwriting not only a company’s systems of record, but its systems of first interpretation. That is not a forecast about AI. It is a description of a principle that AI has merely made impossible to ignore.
Why a well-thought-out program tanks
I am not claiming I predicted any of this. I am claiming something more durable, and more useful: the archive documents a governing operating principle across twenty-two years of technological change.
And the principle does not only describe where decisions begin. It also determines whether the programs meant to serve those decisions succeed or fail — because a program that misreads where the operating reality actually lives will fail no matter how well it is designed. The record contains an early, dated example of exactly that.
In 2004 I researched North Carolina’s At Your Service procurement platform. What impressed me had nothing to do with the platform. It was a Memorandum of Understanding the State signed with its universities, built on what I called collaborative autonomy: institutions kept their own purchasing, their bids were cross-referenced against central contracts, and best value won — from either source. It was the right instrument, a design built to earn cooperation rather than force it.
The program still tanked. Not because the governance was wrong — because when the time came to make it work, they looked at the wrong layer: at the technology’s architecture and integration rather than at how the agents actually operated in the real world. They tried to understand compliance as an interface problem, when compliance was a point-of-capture problem — a question of operating reality, not integration.
That is the exact mirror of what worked at Canada’s Department of National Defence. There, the parameter that reset the entire program wasn’t found in any architecture; it was found at the point of capture, in a plain question — what time of day was an order placed, and what did that do to cost and delivery? That single question exposed a behavioral driver hiding in plain sight: orders were being timed to how people were measured, not to when the goods were actually needed — and no architecture diagram would ever have surfaced it. Technology was the last piece introduced, not the first, and only after the framework was already in place. Same governance ambition in both cases. Opposite place to look. Opposite result. The variable was never the platform, and never the MOU. It was where each program went to understand how the work actually happened.
Which is the mistake every AI-first response is now poised to repeat. People are already defaulting to personal AI, and they are doing it for a point-of-capture reason: it improves how they actually perform, in the moment of the work. Try to win them back at the interface — with governance rules, compliance mandates, blocking, integration diagrams — and you are making North Carolina’s error at scale, treating a capture-reality problem as an interface problem. You will get compliance on paper and people routing around you, into the AI you don’t govern. Understand the capture reality instead — how the AI and the governance model affect their performance where the work is actually done — and you get DND’s result: the framework earns the cooperation, and compliance follows from alignment rather than standing in for it.
AI did not begin this principle. AI is simply the latest technology through which it became unmistakable — and the latest test of whether organizations will read the shift at the interface, or at the point of capture where it actually lives.
One Additional Note: Where the Language of Technology Merges with the Language of Procurement
Years ago I read an article by Paula Hodgins of HPE Canada, titled “To Succeed in the Digital Age, You Need to Work at the Edge,” and it has stayed with me ever since. Her subject was edge computing — the case for processing data at the point of capture, on the device where the event actually happens, rather than pushing everything back to the cloud to be interpreted somewhere else. I have never forgotten it, because it names, in the language of infrastructure, the exact principle this archive has been describing in the language of procurement. The technologists call it the edge. I have been calling it the point of capture. They are the same place, reached from opposite directions — which is what you would expect if the principle underneath is invariant rather than domain-specific.
You can debate whether a model is right. You cannot debate the dated record it was built on.
This analysis draws on the Procurement Insights archive — an independent record, carrying zero vendor sponsorships, that I have published openly since 2007 and that consolidates documented client work, lectures, and writing reaching back to 1998. Every claim in it is held to the Provenance Ledger™: a verify-before-publish discipline that traces each assertion to a primary source and reconciles the record forward rather than editing it in place. Invariant Physics™ is the constant it keeps testing — that however far the technology advances, the operating logic must be in place first. Getting it right, rather than being right.
-30-
Share this:
Like this:
Related