On the Seam: Working Drafts · July 2026
The Eighth R Lives at Layer 8
The seven Rs of cloud migration all operate at or below the application layer. The one that matters now doesn't touch the stack at all.
Every cloud migration conversation eventually arrives at the seven Rs: rehost, relocate, replatform, repurchase, refactor, retain, retire. The framework has earned its ubiquity. Give me any workload and the seven Rs give me a disposition for it — lift it, move it, swap it, rebuild it, keep it, kill it. It is a complete taxonomy of what you can do to an application.
Notice, though, where all seven of them live. Rehosting is an infrastructure decision. Replatforming touches the runtime. Refactoring reaches all the way up into the application itself — Layer 7, the top of the OSI stack, as high as the network model goes. Every R in the framework is a verb you apply to a workload, and workloads top out at Layer 7. The seven Rs are a Layer 1-through-7 framework, and for fifteen years that was the whole game.
There is an old joke in networking that the OSI model has an eighth layer: the human. When the packets are flowing and the app is up and the problem persists, engineers say it's “a Layer 8 issue” — the user, the org chart, the politics. The joke has always been a diagnosis wearing a punchline. The hardest problems were never in the stack. They were in the people and the business sitting on top of it.
Agentic AI is what finally makes the joke load-bearing. Because the eighth R — the one the migration frameworks are now reaching for — does not operate on the workload at all. It operates on the layer above it.
Call it what it is: Reimagine.
Reimagine (the eighth R): to use the migration event as the occasion to ask whether the outcome the workload serves should be produced a different way entirely — by autonomous agents, under new operating assumptions, measured in business results rather than system metrics.
Reimagining is not refactoring with more ambition. Refactoring rebuilds the application to do the same job better; it is still a Layer 7 verb. Reimagining asks whether the job survives. The claims-processing system you were about to replatform assumes humans process claims. An agentic architecture doesn't want that system moved to the cloud — it wants the outcome (claims resolved, correctly, fast) and is largely indifferent to the application that used to produce it. That question cannot be answered by an architect looking at a dependency diagram. It is a Layer 8 question: what is this work for, who answers for it, and what does the organization become if software executes it end to end?
I should note that I am not the first to sense an eighth R coming. AWS has proposed “Reinvention” as an eighth R for public-sector transformation, and their instinct — that agentic AI turns migration into a catalyst rather than a relocation — is the right one. Where I part ways is the framing. Reinvention, as offered, is another item on the migration checklist. My claim is sharper: the eighth R doesn't belong on the checklist, because it isn't a disposition you assign to a workload. The seven Rs answer “what do we do with this application?” The eighth answers “what do we do with this business?” — and pretending those are the same kind of question is exactly how transformation programs end up run like infrastructure projects.
The measurement gives it away, as it usually does. The seven Rs are scored in migration metrics: cutover dates, decommissioned racks, run-rate savings. Necessary numbers, all denominated in the stack. The eighth R is scored at Layer 8: margin, cycle time from order to cash, a revenue line that did not exist before the agents did. If a migration program cannot name the business outcome it will move — not the workloads it will move — it is doing the first seven Rs and borrowing the vocabulary of the eighth.
And if you are not migrating anything? The question does not go away; it just loses its calendar. Migration was never the reason to reimagine — it was only the occasion, the rare moment when an enterprise is already holding every workload up to the light and asking what each one is for. The enterprise that finished its migration three years ago, or never left the data center at all, is sitting on the same question with no forcing function attached, which is worse, because nothing on the project plan will ever surface it. The eighth R is not triggered by a cloud contract. It is triggered by the gap between what your operating model costs and what an agentic one could deliver — and that gap is widening whether or not you have a migration underway.
Which is why this cannot be a technology essay, and why I want to be plain about the motive. You do not break the status quo because agentic AI is impressive. Impressive is not a business case; impressive is how stalled pilots and shadow projects get funded. You break the status quo because it is quietly failing the business — because the outcome costs too much, takes too long, or forecloses a market you should own — and agentic AI happens to be the first technology capable of producing some of those outcomes a different way. The business problem earns the reimagining; the technology merely makes it answerable. If the reimagined answer turns out not to involve an agent at all, the eighth R still did its job.
This is the seam I live on, and the reason I keep writing from it. The seven Rs belong to the architects; the question the eighth R asks belongs to whoever answers for the P&L; and in most enterprises those two groups do not share a vocabulary, a meeting, or a definition of done. So the eighth R keeps getting assigned to the wrong side of the seam — handed to the migration team, who can only answer it with Layer 7 verbs. Someone has to carry it across: capability translated into consequence, architecture translated into accountability. That translation is not overhead on the work. Above Layer 7, it is the work.
So here is the short version, suitable for the whiteboard. The seven Rs tell you what to do with your applications. The eighth R tells you what to do with your business, and it is the only one that operates above Layer 7. If your migration plan has dispositions for every workload and no answer for what the enterprise becomes — or if you have no migration plan and believe that excuses you from the question — you haven't skipped a step. You've skipped a layer.
Arguments offered, not finished.
Dr. Trey Harper writes on trust, legitimacy, and the architecture of the agentic enterprise at treyharper.com and in the LinkedIn newsletter On the Seam: Working Drafts. The views expressed here are entirely my own and do not represent the policy or position of my employer or any customer.