Years ago we argued that IT strategy exists only to serve the corporate strategy. AI changed the technology. It did not change the rule.
A few years ago we wrote that a company’s IT strategy has exactly one job: to serve the corporate strategy. We used Microsoft as the example (IT Strategy vs. Corporate Strategy: Microsoft).
Nothing about that has changed. The technology changed. The rule did not.
Today the same mistake is being made again — only faster, and with a much bigger budget. Swap two letters. IT becomes AI.
We see it with our own clients. A team walks us through “our AI strategy” — a slide of use cases, a copilot here, an agent there, a proud number for how many employees now use the tool. Everyone nods. It feels like progress.
Then we ask the one question we always ask: how does this serve the corporate strategy?
And the room goes quiet.
The answer is not complicated. An AI strategy is a functional strategy: it exists to serve the corporate strategy — and in the long run, that means widening and growing the spread between the return on capital and its cost. It never sets the direction itself.
Same river, new tributary
In the Microsoft piece we described it as a river. The corporate strategy is set upstream. IT strategy flows from it, downstream. You cannot design the IT strategy before you know what the company is for, because the IT strategy exists only to support a direction that has not been set yet.
The head of IT who says “my IT strategy is right, the corporate strategy is wrong” has made a logic error. He built a flawless answer to a question nobody had asked yet.
AI is just the newest tributary. Same river. Same rule.
AI strategy is a functional strategy — like IT, like operations, like corporate finance. It supports the corporate strategy. It never sets it.
A hammer is a wonderful tool. But nobody has ever written a “hammer strategy.” You decide to build a house, and the hammer follows. AI is the most powerful hammer we have ever built — and half the market is busy writing hammer strategies, then wondering why the house looks nothing like a home.
What is different this time — and it is not good
IT strategy was boring enough that nobody ever rolled it out in a panic. AI is exciting enough that everybody does.
The fear of being left behind — of being the executive who “didn’t do AI” — is doing the deciding. And fear is a powerful motivator. It is also a terrible strategist. What most companies call an “AI strategy” is not strategy at all. It is a shopping list of capabilities, bought because a competitor was seen holding the same bag.
“If you build it, they will come” makes for a wonderful movie. It makes for a terrible corporate strategy. That worked for Kevin Costner and a cornfield. It will not work for your balance sheet. Value does not show up because you built an AI. It shows up when the AI serves the corporate strategy — and not a day before.
This is the oldest error in our field, wearing a new outfit: confusing “strategic” with “strategy.” Adopting AI may well be strategically important. That does not make your list of AI projects a strategy, any more than installing a new payroll system is a strategy.
The only scoreboard that matters
So how do you tell whether an AI initiative is serving the corporate strategy — or just spending money that photographs well?
You measure it against the one thing corporate strategy is ultimately accountable for: value creation. And value creation, stripped of the noise, is two numbers.
The spread between the return on invested capital and the cost of that capital — ROIC minus WACC.
And growth.
Tim Koller made this point plainly when we spoke (our conversation is here): a company creates value when it earns more on a dollar of capital than that dollar costs, and then grows while the spread holds. That is the whole game. Everything else is commentary.
Here is the part the AI enthusiasts skip.
Growth is only good when the spread is positive.
If you earn less on capital than it costs, growth does not rescue you — it accelerates the loss. Every extra dollar poured into a negative-spread business, AI-powered or not, destroys value more efficiently than the last one. So an AI program that lifts revenue while dragging ROIC below the cost of capital is not a strategy. It is an expensive way to shrink the company, with a beautiful dashboard bolted on top.
Three questions that put AI back in its place
Hold any AI initiative up against the spread and ask:
Does it widen the spread? Does it earn more per dollar of capital, or free capital up — better returns, not just more activity?
Does it lengthen the spread? Does it build a moat that keeps rivals from copying you, so the advantage lasts longer?
Or does it just add growth? More revenue, more usage, more motion — with no effect on the spread at all.
Only the first two create value. The third is motion mistaken for progress. And most AI roadmaps we see are almost entirely the third.
The second question — lengthening the spread — is really a question about risk and durability. We wrote about that side of strategy in Understanding Risk in Strategy.
The edges of this rule
A rule with no edges is a slogan, so here are the edges of this one.
First, “long term” is load-bearing. There are AI investments worth making that depress ROIC today — because they widen or lengthen the spread later, or buy you an option you would regret not holding. The test is never “does this lift returns this year.” It is “does this credibly improve the spread over the life of the advantage.” Investing for that is discipline. Spending because everyone else is, is not.
Second, the river usually runs one way — but not always. AI is a general-purpose technology, and once in a while a technology is powerful enough to change what the corporate strategy should be. So feed what AI makes newly possible back upstream, into the corporate-strategy refresh, and let corporate strategy decide. What you never do is let the AI team set the company’s direction from downstream.
What this means for leaders
Three changes, and they are changes of discipline, not of software.
Refuse to approve an “AI strategy” as a standalone. Every AI initiative must name the corporate-strategy objective it serves and the spread it is meant to move. No named objective, not funded.
Kill the initiatives that only add activity. If it does not widen or lengthen the spread, it is a hobby, however impressive the demo.
Feed AI’s possibilities up, not down. Route what AI makes possible into the corporate-strategy refresh, then let corporate strategy choose. Never let the tributary redraw the river from below.
The real point
None of this is really about AI. It is about the discipline it has always been about: knowing what the company is for, and making every function — IT, finance, operations, and now AI — serve that, and nothing else.
AI did not repeal corporate strategy. It just made it easier, and far more expensive, to forget it.
AI strategy is not a strategy. It is the servant of one.
Treat it as the servant, and it will help you widen the spread and grow. Treat it as the master, and it will grow you — all the way to the ground.
P.S. If an AI initiative cannot name the corporate-strategy objective it serves and the spread it moves, you do not have an AI strategy. You have an AI shopping list. Thank the team, and send it back.
Leave a Reply