You promised the board an AI strategy by next quarter. The deck is half-built, the slide template is open, and you already sense it is not going to change much. The board will sign it off, the meeting will move on, and the question of how AI actually works in the business will sit exactly where it was before you walked into the room.
That instinct is worth trusting. The document satisfies the ask. The ask is the wrong shape.
What does your board mean when they ask for an AI strategy?
When a board asks for an AI strategy, they are picturing a document. A vision statement, a shortlist of use cases, a roadmap with milestones. They want evidence that leadership has thought about it and has a plan. This is a reasonable request, and it is also a category error. What they are describing is a plan to plan, not a plan to operate.
The document lives in a different register from the thing the business actually needs. A strategy deck answers the question “what do we intend?” An operating model answers the question “how does this actually work?” Boards are comfortable with strategy documents because they resemble everything else they approve, a direction, an investment, a set of milestones. They are less likely to ask for the second thing, and they do not always know they need it.
This matters because a business can have an excellent AI strategy and still get nothing useful from AI. Use cases are chosen, priorities are ranked, the timeline is signed off. And then nothing changes in how the business runs day to day, because the document never addressed how anything would change.
Why does a strategy document leave the real work undone?
A document can satisfy the board without changing anything in the business. That is the structural problem. The board signs it off, the meeting moves on, and the question of how AI gets built, governed and run day to day stays unanswered. Approving a direction is not the same as creating a capability, and the gap between the two is where many AI efforts stall.
The problem is specific. Boards are trained to think in terms of strategy, so they want to know whether leadership has a point of view, a plan, and a reasonable basis for investing. A well-structured document answers all three of those things. It does not answer the operational questions about who decides which AI tools are used, how outputs get checked, what the process looks like when a model generates something wrong, and how the team learns and adapts as the tools evolve.
BCG research across more than 1,500 executives found that organisations are projected to double their AI spending as a share of revenue in 2026, yet a third of businesses still do not treat AI as a top-three priority. The plan has been approved in many of those businesses. Building the thing the plan describes is a different conversation entirely.
What does a real AI operating model actually cover?
Prolego’s framework for AI operating models identifies five dimensions a genuine AI operating model has to address, and all five need to mature in parallel rather than in sequence. The dimensions are capabilities, infrastructure, talent, governance and tools. A business that focuses on one while neglecting the others will find gaps showing up later, often at the worst possible moment.
Capabilities are the AI systems that solve actual business problems. Infrastructure is the computing and data environment that makes those systems reliable at scale. Talent is the people who can run, maintain and improve what gets built. Governance is the oversight that keeps AI use legal, safe and aligned with how the business wants to operate. Tools are the third-party resources, platforms and applications that make it possible to move at a reasonable pace.
The reason this list matters to you as the founder is that a strategy document typically only touches the first dimension, the use cases you are planning to pursue. The other four stay invisible until something goes wrong. You do not want to discover that infrastructure was never designed for scale three months into a rollout, or that governance was never defined when a team member uses a tool in a way that exposes customer data. Together, the five dimensions make the case for treating AI as a multi-year operational programme rather than a series of tool purchases.
How do you reframe the board ask without sounding like you are dodging it?
The reframe works when you shift from technology language to organisational change language. Boards fund outcomes. Prolego’s research on AI operating models is specific on this point. Executives who secure large AI budgets frame the case around how the business will operate differently, with the platforms and tools treated as supporting evidence. The subject of the conversation is the company, not the software.
The practical version is straightforward. You give the board the document they asked for, clean and well-structured, covering vision, use cases and a first-year sequencing. Then, in the same meeting, you introduce what that document does not yet address, which is the operating model work that will take the next two to three years and needs its own budget, its own governance and its own milestones.
The framing that tends to land is this. The board has approved the direction. Now it needs to approve the delivery mechanism. Boards understand capital projects. They know that approving a building plan and actually building the building are two different decisions. AI follows the same logic.
NACD data shows that 62% of board directors now set aside dedicated agenda time for AI in regular meetings. The board is already expecting this to be a multi-meeting conversation. Use that expectation in your favour. The deck satisfies the first meeting. The operating model structure gives you the frame for all the ones that follow.
What else connects to this decision?
The strategy-versus-operating-model distinction touches several questions that will come up as you build this out. Board liability is the most immediate. As director accountability for AI decisions increases, a well-documented operating model becomes the evidence that appropriate oversight existed. Morgan Lewis’s analysis of AI in deal activity confirms that private equity acquirers now assess AI operating maturity as a standard part of due diligence.
There is also the question of the person you have put in charge of this work. If you have handed the AI mandate to someone inside the business, they will hit this distinction before you do. They will be asked to deliver the strategy document, and they will understand clearly that the document is not the delivery work. The operating model question is the one they need your backing to answer.
Spencer Stuart’s research on CEO behaviour in AI-mature organisations found that the founders making the clearest progress are those who stay actively involved in the strategic framing. The strategy-versus-model distinction is exactly where your involvement is load-bearing. Your delegate can run the delivery. The reframe with the board is yours to own. Once you have made that distinction clearly in the boardroom, it tends to stick, and the board starts expecting operating-model thinking from you, which is a far easier expectation to meet than the one you started with.
