How should a board distinguish an AI strategy from an AI shopping list?

A board should be able to explain an AI proposal as a choice about the business: whom the company will serve, what it will offer or do better, and why that choice is worth pursuing. The software purchases should follow from that explanation. A list of tools, licenses and launches leaves the strategic question open, even when every item has a budget.

The useful test is whether management can connect the proposed capability to a customer or operating need, show what remains uncertain, and explain which finding would change its recommendation. This applies to a new offering as well as an improvement to an existing service. It does not require every experiment to arrive with a proven return.

Describe the business choice before the purchases

Suppose a 250-person market research consultancy serves packaging manufacturers through bespoke research assignments. Its managing director, Sam Taylor, proposes a paid monthly service that alerts existing clients to relevant competitor changes.

Analysts would examine permitted public company announcements and product pages, use AI to prepare internal notes, and check the sources and context before issuing each client brief. A named analyst would remain responsible for what the client receives.

“Buy research software and train the analysts” describes part of the work. The strategic proposition is more demanding: clients have recurring decisions that a monthly monitoring service could support, and the consultancy can serve those decisions well enough to earn repeat business. That proposition could be wrong even if the software produces an accurate sample.

The UK Financial Reporting Council’s governance guidance encourages boards to consider how technology affects strategy and the business model, alongside how the company creates and sustains value. It is non-prescriptive guidance for users of the UK Corporate Governance Code, rather than a rule for every board. Its practical question travels well: what changes for this business if management’s proposal works?

For Sam’s proposal, director Jamie Lee needs to understand the proposed service before discussing the software. Her questions are about the offering:

• Which client decision would the brief inform?
• Why would a client want a monthly service rather than another bespoke assignment?
• What would the consultancy offer beyond a list of public announcements that clients could obtain themselves?

If those answers remain vague, a more detailed purchasing schedule will not resolve them.

Separate evidence for the tool from evidence for the business

An analyst could demonstrate a well-sourced competitor brief. That would be useful evidence about producing a brief under the tested conditions. It would leave other questions unanswered.

Jamie should ask management to distinguish the assumptions and the evidence it has for each:

• Client need. Identify the recurring decision the brief would support. Conversations about a specific recent decision are more informative than general enthusiasm for an AI service. Record what clients currently use and what would make a paid service worth considering.
• Research usefulness. Check whether the proposed sources reveal changes relevant to that decision. An accurate summary can still miss the information a client needs, or arrive too late to help.
• Delivery and repeat business. Examine the analyst work required to verify, interpret and deliver the brief, alongside the proposed terms and evidence of willingness to pay. Fast internal drafting alone does not establish a sustainable service.

These are separate uncertainties. A source-checking exercise cannot answer whether clients will buy. A promising client conversation cannot establish that analysts can deliver the service reliably. Management should label the evidence it actually has, rather than allow the strongest demonstration to stand in for the whole proposal.

The NACD guide to implementing AI governance similarly places strategic relevance and capital allocation alongside risk and competency. It asks directors to connect investment with business strategy and capabilities. This is professional guidance for oversight, not proof that a particular AI investment will succeed.

Keep a credible alternative in the discussion

Sam’s consultancy has two credible directions with different demands on the company:

• Launch recurring monitoring. The subscription service creates an ongoing promise to check, interpret and deliver relevant changes for clients.
• Improve bespoke assignments. Faster, better internal research supports work the consultancy already sells, without adding a monthly service promise.

The board need not turn the discussion into an exhaustive catalogue of possibilities. It should understand why management prefers the proposed direction over the strongest practical alternative, including continuing without a new AI purchase where that remains credible.

In an October 2026 Reddit discussion about strategic planning, a Reddit user described the pressure to accept the first plausible option because it ends the meeting. They reported finding it useful to ask what would have to be true for each option to work and to record what was rejected and why. This was one person’s account of their practice, with no independently verified business outcome. It captures a recognizable obstacle: asking for another explanation can look like delaying agreement.

For Jamie, the useful challenge is specific. If clients value occasional interpretation of a major market change rather than regular monitoring, the subscription proposal may need to change. If checking and interpreting the source material consumes the available analyst time, a fast draft may add little to the offering. Those findings would challenge the business proposition, rather than merely suggest another software configuration.

Management can still argue for a limited experiment. It should say which uncertainty that experiment can reduce and which decision will remain open afterward.

Put the actual dependencies back into view

Describing the proposition without vendor names is a useful clarity test. The real proposal must then account for the dependencies those names represent. Availability, integration, source access, contract terms and the ability to change suppliers can affect which direction is practical.

In an October 2026 LinkedIn article, board president and technology adviser Christopher Ruch described how earlier enterprise-platform choices can constrain later AI options. His account draws on his advisory experience and has a commercial context; it does not establish how frequently this happens. It complicates the idea that management starts with an unrestricted choice of tools.

For the consultancy, Sam should establish whether the proposed source access is permitted, what the service depends on, and what would happen if an important source or supplier became unavailable. A replacement product might preserve internal drafting while failing to preserve the client offering. That difference belongs in the strategic discussion.

Detailed costs and financial assumptions still need their own scrutiny. The guide to building an adoption business case covers cash, capacity, alternatives and sensitivity. Those calculations help assess the proposal; they do not replace explaining why the company should offer this service.

Be clear about what the board is deciding now

A board discussion should distinguish exploring a direction from committing the company to deliver it. For Sam’s proposal, management might be asking for permission to investigate client demand, to test a tightly bounded service, or to fund an ongoing offering. Each asks the board to accept a different level of uncertainty and commitment.

Useful questions include:

• What business choice is management recommending, and why?
• Which material assumptions have evidence, and which remain untested?
• What commitment is being requested now, including the people and obligations it would create?
• What result would cause management to change direction or return with a different recommendation?

There is room for uncertainty. The NACD guidance cautions against prematurely cutting promising AI experimentation simply because early returns are unclear. A reasonable response to missing evidence can be a bounded investigation, with its purpose and limits stated. It need not be either a full launch or rejection.

Once direction is agreed, an AI steering committee can resolve shared priorities and capacity within its authority. The board-level question here is what the company is trying to become or improve, and whether management’s explanation and evidence justify the commitment being requested.

For the next discussion, take one proposal and write the business proposition in ordinary language. Place the strongest supporting evidence beside the assumption it addresses. Leave the unresolved assumptions visible. That gives directors something meaningful to challenge before the purchase list becomes the plan.

Questions about board oversight of AI strategy

Does a successful AI pilot prove that the strategy is sound?

A successful pilot supports the specific capability and conditions it tested. It may show that analysts can produce an accurate competitor brief, for example. It does not by itself establish customer demand, repeat purchases or sustainable delivery. Ask management which strategic assumption the pilot tested and what further evidence is needed before making the next commitment.

Do directors need to choose the AI software?

Directors need enough understanding of the proposed capability and its material dependencies to challenge the business decision. The appropriate purchasing and approval authority depends on the organization. Asking why a supplier is essential, what constraints it creates and whether an alternative preserves the intended offering is more useful here than comparing feature lists without a business proposition.

Does a company need a separate AI strategy?

A company can address AI within its existing business strategy if it makes the relevant choices, responsibilities and uncertainties explicit. A separate document can help organize the work, but its existence does not establish a strategy. For the operational starting point, connect an AI adoption strategy to actual workflows and ownership.

What should management bring back to the next board discussion?

Management should return with evidence that addresses the decision left open. In the consultancy example, a request to investigate recurring demand calls for findings about client decisions and willingness to consider the service, including contrary responses. Another polished sample brief would answer a different question. Explain what the findings change, what remains uncertain and which commitment management now recommends.