Two Things We Call Strategy

One is a direction you can test. The other is a generalization you can execute. We use the same word for both — and that's why so many strategies die in the room where they're born.

A colleague and I keep having the same argument. He wants to write a high-level strategy for a client before anything has been built. I keep telling him there's no such thing — that a high-level strategy is what you get after you abstract a thousand small decisions, not before.

We were both right. We were also talking about two different things and calling them by the same name.

That's the actual problem, and it isn't ours alone. "Strategy" is one word doing two jobs. Until you pull them apart, every strategy conversation is two people nodding at the same noun while meaning opposite things — which is exactly why the deck everyone approves is the deck nobody executes.

The two things

Direction is a bet. It says: here is where we should go, and here is why we think advantage lives there. Its defining property is that it's testable. You can write it down before you've done anything, because all it really claims is "this is worth finding out." It's cheap, it's provisional, and the honest version of it expects to be wrong in places.

Generalization is a pattern. It's what you get when a body of real decisions has been made, has compounded, and you can finally see the shape they formed. Its defining property is that it's executable — it encodes know-how, the feel for how things actually move when you push on them. But you can only have it after the work. You can't write it down in advance, because the information it's made of doesn't exist yet.

We've got it backwards

Here's the part nobody says out loud. The high-level statement an executive points to and calls "the strategy" is the direction — the most hypothetical, least executable thing in the building. The granular work they wave away as "just execution" is where the executable strategy actually gets manufactured.

We treat the testable thing as if it were the proven thing, and the proven thing as if it were beneath strategy. It's the wrong way round, and it's expensive in two specific ways.

Mistake the direction for a generalization and you commit hard to an untested bet. You resource it, you announce it, you tie people's targets to it. Then reality disagrees — and because you sold a hypothesis as a certainty, the correction reads as failure instead of the information you were supposed to be buying.

Refuse to set a direction until you have a finished generalization and you get paralysis. Or worse: you borrow someone else's. Best practice. A maturity curve. The reference architecture. You present a pattern abstracted from other people's decisions as if it were grounded in yours. That's the strategy deck that feels theoretical because it is — a pattern with no decisions of your own underneath it. Everyone nods. Nobody moves.

They're not rivals — they're two ends of one process

The verb between them is execution. The direction tells you where to put your hands. The small decisions you make there compound — but only into a coherent generalization if the direction is living inside each decision as a constraint, deciding which moves reinforce the pattern and which quietly leak away from it. And the compounding feeds back: what you learn rewrites the direction. The strategy you end up able to execute is never the direction you started with.

So a direction without execution stays a guess. And a generalization you didn't earn through your own decisions isn't yours — it's a quote.

What this means when a board asks for "a high-level strategy"

This is why the expected answer is the wrong one.

Take a board asking for a high-level AI strategy. The reflex deliverable — the maturity model, the twenty-use-case grid, the governance box — is a generalization with no execution behind it. Theoretical by construction. And a board can smell it, even when they can't name why.

What you can honestly hand over up front is a direction: a testable bet — say, that AI's near-term value here is margin expansion, not new revenue — plus two or three concrete probes that begin converting that direction into a generalization in front of them. The probes aren't the strategy. They're how you make the executable strategy visible before it fully exists. They're the first small decisions, chosen so the board can feel what the pattern is going to become — instead of being asked to approve a pattern that was never theirs.

Stop asking "what's our strategy"

It's the wrong question, because the word can't tell you which of the two things you're holding.

Ask instead: is this a direction, or a generalization?

If it's a direction, the only sane next question is how cheaply can we test it. If it's a generalization, the question is what did we actually do to earn it — and if the answer is "nothing, we read it in a report," it isn't yours and it won't execute.

The argument my colleague and I kept having was never about strategy. It was about which strategy. Once you can name which one is on the table, the argument mostly disappears — and the work can start.