When a company asks for an AI strategy before it can clearly state its actual strategy, the AI is not the problem that needs solving.

I get some version of this request often now: help us figure out our AI strategy. It is almost always asked by a team that has not yet answered a more basic question, which is what the business is actually trying to win at. AI strategy, in those conversations, is functioning as a stand-in for a decision nobody wants to make yet.

The tell

Here is a fast diagnostic. Take the “AI strategy” document and try renaming it to just “our strategy,” with AI appearing as one line item inside it instead of the headline. If the document falls apart when you do that, it was never a strategy. It was a technology initiative wearing a strategy’s clothes.

A real strategy says what the business is going to do, and just as importantly, what it is going to stop doing or refuse to do. It names a bet and accepts the tradeoff that comes with it. Most “AI strategy” documents I have reviewed do neither. They list capabilities: automate this, personalize that, deploy a chatbot here. A capability list is not a strategy. It is a shopping list, and shopping lists do not require you to choose anything.

Why this happens

Committing to a real strategy means admitting what you are not going to do, and that decision has a name attached to it if it goes wrong. Committing to “explore AI opportunities” carries almost no personal risk, sounds forward-looking in a board meeting, and can occupy a team indefinitely without ever forcing a hard prioritization call. That is precisely why it is such an appealing thing to fund instead of a real decision.

What the useful version looks like

The engagements where AI work actually produces something real always start with the boring question first: what is the business trying to become, who exactly is it trying to win, and what will it say no to along the way. Once that is answered, the AI questions get much smaller and much more concrete. Which specific workflow benefits from automation. Where a human needs to stay in the loop and why. What happens when the model is wrong, and who is accountable when it is.

None of that requires hype. It requires sequencing the questions correctly, which is most of what strategy work actually is.

If your team has an “AI strategy” that could not survive being renamed to “our strategy,” that gap is usually fixable in a few working sessions. Here’s how I approach that kind of work, or reach out directly if you want to talk through where yours currently stands.