AI adoption as an operating-model programme
Per-use-case value verified under the Markham Verification Standard at month 10. Full audit trail available under NDA.
The situation
A consumer goods group arrived with what most organisations had in 2025. Fourteen AI pilots, a platform subscription, a steering committee, and no verified operating result. The pilots reported to IT. Ordering, allocation and customer service — the decisions the models were meant to change — reported somewhere else entirely, and nothing in the governance of the programme connected the model to the decision. Eighteen months of activity had produced demonstrations.
The group wanted to know whether to double the AI budget or halve it. Our answer was that the budget was not the variable that mattered. Ownership was.
What the diagnostic found
The portfolio held 42 candidate use cases, fourteen pilots and twenty-eight proposals. Each one had to name the operating decision it would change and the person who owned that decision, which sounds like a low bar until you apply it, and then it is quietly brutal. Of those, 31 could not name the decision at all, and they were killed on paper before any further spend.
The eleven survivors clustered where the operating value actually sat, which was demand forecasting feeding the weekly ordering decision and service automation resolving contact classes that had never needed a queue in the first place.
How Markham helped
Ownership moved first. AI became an operating-committee agenda item with a value ledger, and every use case got a named business owner, meaning the person whose decision it changed rather than the person who had built it. Each carried a value case and a month-0 baseline. Platform and data work were costed inside the use cases that needed them and never as a programme of their own.
The eleven use cases were then drawn in three phases under the standard operating model, three or four at a time, and released into production only when the previous set had passed verification. HolisticAutomation built the pipelines and integrations. We held the architecture and the oversight. At month 10, all eleven were in production with verified per-case value.
Impact in detail
Every use case carries its own baseline and attribution line; month-10 readings verified under MKM-F-003.
What we took from it
Killing 31 use cases was the highest-value work in the programme, and it is the work clients resist hardest, because every pilot already has a sponsor who has presented it internally. We ran the test anyway.
The reporting line did what no model upgrade could: when the ordering forum owned the forecast, the forecast started changing orders.
Pricing platform work inside use cases kept the infrastructure honest. It was built exactly as large as the verified value required, and no larger.