The transition from programme to operations is where AI capability stalls

The final session of an AI programme typically produces high energy. Prototypes are functioning. Governance frameworks are documented. The team has a shared vocabulary and a set of commitments. There is a genuine sense that something has been built.

Thirty days later, the operating model is already beginning to dissolve. The governance framework sits in a shared drive. The prototypes are in maintenance mode while people wait for a decision about scaling. The weekly rhythm designed in the programme has been pushed twice to accommodate other priorities. The momentum that was real at the end of the programme is already historical.

This is the most common failure mode in AI transformation, and it is also the most preventable. The stall happens not because the work was wrong but because the transition from programme structure to operational structure was not managed. The programme provided scaffolding. When the scaffolding was removed, the structure that was supposed to stand independently had not yet developed the load-bearing capacity to do so.

The 90 days after programme completion are the period in which the operating model either institutionalises or reverts. Month one is the most critical. What happens in month one determines what is possible in months two and three.

Month one is a governance month, not an implementation month

The instinct in month one is to deploy. The programme built prototypes. The team has tools and frameworks. The natural next step feels like scaling what was built.

This instinct is premature. Month one is a governance month. Its purpose is to activate and stress-test the structures that will carry AI capability forward without programme support. Implementation can accelerate in months two and three, once those structures are confirmed to be functioning.

Month one must contain five specific elements:

  • The first governance review session, held within the first two weeks. The governance owner convenes the review. Agenda items: which prototypes from the programme are ready for operational deployment, which decisions are blocked and by what, and what governance questions have arisen since programme completion.
  • Confirmation that the operating rhythm is active. Every cadence scheduled during the programme should have run or have a confirmed date within the month. If a session has been postponed, the reason should be documented and a new date set within the month.
  • A 30-day review of the governance framework. What decisions has the framework been asked to make since programme completion? Did it produce clear answers? Where did it produce ambiguity or require escalation?
  • A communication to the broader organisation. What is the governance owner's remit? What is the process for raising AI-related decisions or requests? Who is the point of contact for AI governance? This communication prevents the governance framework from being invisible to the people who need to use it.
  • A documented month-two plan. What will be deployed, evaluated or decided in month two? The plan should be specific enough to be held accountable and flexible enough to accommodate the governance review findings from month one.

Month one is complete when all five elements are confirmed, not when all five are initiated. Initiating the governance review and then letting it drift is not completion. Scheduling the operating rhythm and then postponing it twice is not activation. The measure is whether the structures are functioning, not whether they exist on paper.

Draft your 90-day plan before the final programme session

Organisations that move from programme to sustained capability design their 90-day plan during the programme, not after it. This is not because the plan is needed urgently. It is because the programme is the only window where the plan can be designed with full accountability and facilitation. After the programme ends, the team disperses, competing priorities resurface and the structural conditions that made planning productive are removed.

The 90-day plans drafted during the programme specify:

  • The five month-one elements listed above, with named owners and specific dates for each.
  • The two or three operational deployments targeted for months two and three, with success criteria drawn from the baselines established during the programme.
  • The governance decisions that are anticipated in the first 90 days: tools to be evaluated, workflows to be reviewed, policies to be updated.
  • The escalation path for decisions that exceed the governance owner's authority.
  • The 90-day review: who convenes it, what it assesses and what outcomes it is designed to produce.

The plan does not need to be detailed beyond 90 days. Planning further in advance creates false precision about a landscape that will change as deployments proceed and the governance framework develops. The value of the 90-day plan is in the specificity of month one and the directional commitment of months two and three.

AI capability that compounds does so because there is a functioning operating model to compound within. The 90-day plan is the bridge between the programme that built the model and the operations that will run it. Organisations that cross that bridge in month one sustain what the programme built. Those that defer the plan until after the programme ends find the crossing becomes significantly harder.