Thinking in public

When the instruments measure themselves

An organization can trust every measure it has and still miss the change in its market that matters.

A salesperson hears hesitation that will not enter the churn data for months. An engineer sees a technology cross a threshold that the current product agenda has no place for. A pilot is assessed with the same business case it was meant to challenge. None of these is necessarily a failure of diligence. The instruments can be accurate and still describe the world they were designed to see.

The useful question is where a disconfirming signal would have to come from. If every trusted measure is produced inside the operating model being tested, that model cannot report its own obsolescence. Someone has to be listening to the street. Strategic foresight builds that listening into the work. It gives a signal from a customer, the front line, or the edge of the market a route to someone with the authority to act on it.

The organization has a grain

An organization remembers its earlier choices in the structures of everyday work.

Last year's budget, hire, and customer promise become this year's constraints. Knowledge lives with people who each see only part of the whole. Some practices shift quickly while others have taken years to form. These are properties of the material a leader works with.

Every decision changes that material. Customers and competitors respond as well, so the next decision meets a different organization in a different market. Strategy begins with attention to this response: what did the action make easier, harder, newly visible, or newly possible?

Organizations run on several clocks

A market can move in weeks while an operating model takes years to change.

Strategy requires more than a single planning horizon. A product decision, a habit of coordination, an institutional promise, and the conditions of a place do not move at the same rate. Treating them as though they do creates false urgency in one place and dangerous delay in another.

The work is to notice which clock governs which choice, then bring commitments made at different speeds into a coherent relation. Time is one of the materials through which an organization is made. A team that reads its clocks well can choose when to move, and timing becomes an advantage over competitors that run everything on one schedule.

Everyone got faster, and nothing shipped sooner

An artificial intelligence (AI) tool raises output only when it relieves the step that limits the whole system.

A company gives every engineer a coding assistant, and code gets written much faster. The queue in front of code review grows, testing runs at its old pace, and the release date stays where it was. Marketing drafts a campaign in an afternoon, then waits two weeks for legal approval. Leaders watch activity rise while finished work stays flat, and they begin to doubt the tool.

The tool may be doing what it promised. It sped up a step that was not setting the pace. Every system has one step that limits what the whole can finish, and Eliyahu Goldratt, who developed the Theory of Constraints, held technology to that standard: “Technology can bring benefit if, and only if, it diminishes a limitation.” Current research points the same way. Google’s DevOps Research and Assessment (DORA) program found that AI raised delivery throughput while instability kept rising. McKinsey found that the organizations attributing a meaningful share of earnings to AI were nearly three times as likely as the rest to have redesigned their workflows.

The useful question before the next purchase is where the work waits. A team that can name its constraint can ask which limit a tool should relieve, and which rules the organization built to live with that limit. Those rules outlast the limit unless someone retires them, so a tool can relieve the real constraint and still deliver little while they remain. This account would be wrong if finished output rose while the constraint and its rules stayed as they were.

The Dice Game shows the pattern in twenty-five minutes. Round 4 doubles one station, the way an AI tool might, and only the station where the work waits moves the result. Play it in the Dice Game

Sources: Eliyahu M. Goldratt, Eli Schragenheim and Carol Ptak, Necessary but Not Sufficient (North River Press, 2000). Google DORA, State of AI-assisted Software Development (September 2025), as reported by InfoQ. McKinsey & Company, The State of AI in 2026: On the Road to ROI (August 2026).

05 / Focus

Strategic centering is necessary, and it is not sufficient

A Strategic Center and a strategy cascade are both necessary. Neither identifies the constraint or the crux.

Read the essay