Strategy without shipping is just expensive opinion, and the AI era has stripped away the last excuse for confusing the two. For decades the slide-deck class had a defensible arbitrage: analysis was slow and costly, so synthesizing a market into forty pages was genuinely scarce work. That scarcity is gone — a competent operator with AI tooling produces the analysis in an afternoon. What was never abundant, and still isn't, is the willingness to make the strategy falsifiable: to compile it into a system someone builds, a number someone owns, and a deadline someone can miss. If your strategy can't be wrong by a specific date, it isn't a strategy. It's a mood board with margins.
The compile test
Borrow a habit from software: code that doesn't compile isn't judged on its elegance. Run the same check on any strategy document that crosses your desk. Does it name the system that will exist afterward — a pipeline, a product change, an agent workflow, a new motion someone will build? Does it commit to a number — not 'improve win rates' but 'demo-to-close from 22% to 30%'? Does it carry a date after which we all agree it failed? Most strategy decks fail all three checks, and that's not an accident of formatting. Vagueness is the business model: an unfalsifiable strategy can be re-presented next quarter with new stock photography. A compiled strategy has the one property the slide-deck class can't survive — it can visibly not work.
Why the slide-deck class is losing
- Their scarce input became free: market synthesis, competitive analysis and framework application are now largely machine work. The deck's production cost signaled effort; the signal is dead.
- Zero feedback loops: a deck is approved or ignored, never falsified. Operators accumulate corrections from reality; deck-makers accumulate approvals from committees. Guess which one improves.
- The handoff tax: strategy formed without contact with execution arrives at the builders as an untested spec, and the expensive rework gets booked as 'execution failure' — protecting the strategy layer from its own errors.
- AI shrank the distance from idea to artifact so much that 'we recommend piloting X' is now weaker than 'we built a pilot of X; here's what the data says.' When showing beats telling, telling loses its salary.
- The honest caveat: bad shipping exists too. Operators who ship without direction produce fast, well-instrumented waste. The winner isn't shipping instead of strategy — it's strategy that ships.
The new standard: strategy as a shipped artifact
- 1Cap strategy documents at two pages: the bet, the system to build, the number, the date. If it needs forty slides, the thinking isn't done.
- 2Require a walking skeleton within 30 days of any approved strategy — the smallest end-to-end version of the system, live, with real data flowing.
- 3Assign the number to a named owner who was in the room when the strategy was set. Strategy authorship without metric ownership is how opinions stay expensive.
- 4Hold pre-scheduled kill reviews: on the committed date, the strategy is extended, amended or killed against its own number. No re-decking.
- 5Promote the people whose strategies compile and survive contact with reality. What you promote is the only strategy document your company actually reads.