Every January, strategy documents are finalised. Every December, most of them are quietly buried. That gap between what a business decides and what it actually does is not new. What is new is that the gap now has a name, a budget line, and a competitor on the other side of it who is moving.
For a decade the execution gap was a management problem. A strategy was sound; the operating model never changed to deliver it. The deck said “become data-driven” and the team went back to the same spreadsheets on Monday. Consultants were paid handsomely to widen the gap: brilliant analysis, no working system, and an invoice.
AI has not closed that gap. It has raised the stakes on it. Two firms in the same market now write the same sentence into their plans, “we will use AI to move faster.” One turns it into deployed systems inside a quarter. The other turns it into a workshop, a tool subscription nobody adopts, and a slide for the board. Twelve months later they are not in the same race.
Strategy about AI is not the same as AI in production
Most of what is sold as AI strategy is a document about AI. It surveys the landscape, ranks use cases on a two-by-two, and recommends a roadmap. It is useful for exactly as long as it takes to read. Nothing in the business works differently the day after it lands.
The firms pulling ahead skipped the document. They picked one workflow that visibly costs them hours or money, deployed a system that changes how that work is done, measured the result, and moved to the next one. The strategy is real, but it lives in what shipped, not in what was written.
The firms winning with AI are not writing strategies about it. They are quietly compounding on systems already in production.
The gap compounds monthly
This is the part that does not show up in a planning cycle. A deployed system does not sit still. It gets audited, corrected, and extended every month. The competitor who started in March is not twelve weeks ahead by June; they are twelve weeks of compounding ahead, and the distance is widening while the slower firm is still scoping.
The 67 per cent figure is old news; well-formulated strategies have always failed in execution more often than not. The new number that matters is the second one. Ninety days is now a realistic window from a standing start to measurable impact, if the work is deployment and not deliberation. Every quarter a business spends deliberating is a quarter it hands to whoever deployed.
What to do about it
Stop commissioning strategy about AI. Commission a diagnosis of where the business actually loses hours, money and quality, then deploy against the top one or two findings inside the same cycle. Insist on a system in production, not a recommendation. Measure the cost line that fell or the revenue line that rose; if you cannot point to one, you ran an experiment, not a project.
The execution gap was always the real problem. AI just made it visible, expensive, and fast. The firms that treat it as an engineering task rather than a strategy exercise are the ones you will be reading about, or competing against, by year end.