Richard P. Rumelt, 2011

“The kernel of a strategy contains three elements: a diagnosis, a guiding policy, and coherent action.”

Richard P. Rumelt presented this kernel in Good Strategy/Bad Strategy to separate strategy from aspiration, budgeting, and motivational language. Diagnosis simplifies a complex situation by identifying the central challenge. A guiding policy chooses an overall way to address it. Coherent actions coordinate resources and behavior so that the policy becomes real. None of the three is sufficient alone. A goal without diagnosis may target the wrong problem; diagnosis without policy produces analysis without direction; a policy without aligned action remains a preference rather than a strategy.

The framework begins with a difficult act of choice. An organization must identify which obstacle matters enough to organize around and which competing concerns will not lead the agenda. A useful diagnosis does more than repeat symptoms; it explains a structure that makes action possible. The guiding policy then sets boundaries and advantage without pretending to specify every task. Coherent actions reinforce one another—product focus, capability investment, pricing, partnerships, and stopping decisions all point toward the same response. Coordination creates leverage that a long list of independent initiatives cannot.

A concise kernel can still be wrong. Leaders may diagnose from a narrow perspective, ignore people who experience the system differently, or select a policy that shifts costs to workers and communities. Coherence is not a moral guarantee; harmful actions can also align. Strategy therefore requires evidence, challenge, and explicit constraints concerning safety, legality, and responsibility. The kernel is best treated as a hypothesis: the diagnosis predicts what drives the difficulty, and the actions test whether the policy can overcome it. Reviews must permit revision when reality answers differently.

The model is valuable for AI programs that otherwise become collections of pilots. Leaders can diagnose a particular customer or operating constraint, choose how AI will help, identify capabilities and controls, and align a limited set of actions around that choice. Projects that do not support the policy can be paused. Measures can test both desired results and unacceptable side effects. Rumelt’s formulation turns strategy from a ceremonial document into an argument expressed through commitments. Its quality becomes visible in what the organization does together—and in what it deliberately decides not to do. A strategy page can be limited to four blocks: the pivotal challenge, the explanation that makes it pivotal, the governing approach, and the mutually reinforcing commitments. A fifth block names what will stop. If leaders cannot complete the page without filling it with aspirations or unrelated projects, the difficulty is diagnostic rather than editorial. The constraint forces unresolved disagreement into view before polished language hides it.