Richard Rumelt's strategy framework separates clarity from noise. Learn how to diagnose whether your AI initiative is a coherent competitive move or expensive theater.
By OperatorRadar Editorial
In *Good Strategy/Bad Strategy* (2011), Richard Rumelt defines good strategy as having three elements: a diagnosis of the challenge, a guiding policy that addresses it, and coherent actions that execute the policy. Bad strategy, by contrast, is fluff—vague aspirations, disconnected initiatives, and failure to acknowledge real constraints. Rumelt argues that most strategic plans fail because they mistake goals for strategy. A goal is 'increase revenue 40%.' Strategy is *how* you'll do it given your actual position and constraints.
Rumelt's framework emerged from decades of strategy consulting and academic research. He observed that executives often conflate ambition with strategy, creating plans that sound impressive but lack diagnostic rigor. His work challenged the 1990s-2000s trend of 'big hairy audacious goals' (BHAG) without corresponding operational clarity. The book became a standard reference for distinguishing strategic thinking from strategic theater.
Rumelt's core insight: strategy is not a wish list. It requires (1) honest diagnosis of your competitive position and the specific problem you're solving, (2) a clear policy or principle that constrains your choices, and (3) coordinated actions that reinforce each other. Bad strategy sounds like: 'We will be the AI leader in our industry.' Good strategy sounds like: 'Our customers struggle with manual data reconciliation, which costs them 15% of operational margin. We will build AI-assisted reconciliation that reduces their manual work by 70%, priced at 20% of their current savings, and distribute it through their existing ERP vendors.' The second diagnosis identifies a real problem, the policy (20% of savings) constrains pricing and feature scope, and actions (ERP partnerships) are coherent.
AI has amplified both good and bad strategy. The technology is powerful enough that mediocre strategy can produce short-term results—a chatbot or predictive model can generate value even without coherent positioning. This creates false confidence. Simultaneously, AI's rapid evolution tempts executives to adopt 'follow-the-hype' strategies: 'We must do generative AI' without diagnosing what problem it solves for their business. The capital intensity of AI (compute, talent, data infrastructure) makes bad AI strategy more expensive than bad strategy in other domains. A poorly diagnosed AI initiative can burn $2M+ annually with minimal competitive advantage.
Rumelt's framework is more relevant now, not less. The three elements still separate viable AI strategies from noise: (1) Diagnosis: Can you articulate the specific customer problem, internal bottleneck, or competitive gap that AI addresses? If you can't, it's not strategy. (2) Guiding policy: What principle constrains your AI scope and resource allocation? Examples: 'We will only deploy AI where it reduces cost by >30%' or 'We will only build AI features that improve customer retention.' (3) Coherent action: Do your AI investments reinforce each other? If you're building a proprietary LLM, a third-party API integration, and a no-code AI tool simultaneously without connection, that's bad strategy. Rumelt's insistence on diagnosis over aspiration is the antidote to AI theater.
Rumelt's framework assumes relatively stable competitive environments and long planning horizons. AI's pace of change compresses both. A guiding policy set in Q1 2024 may be obsolete by Q4 2024 if a new model or capability emerges. Additionally, Rumelt emphasizes 'focus'—doing fewer things better. In AI, some organizations benefit from portfolio approaches: testing multiple models, APIs, and use cases in parallel to learn quickly. This isn't bad strategy; it's adaptive strategy under uncertainty. The framework also assumes you have clear data on customer problems and competitive positioning. Many organizations entering AI lack this data, requiring diagnosis-through-experimentation rather than diagnosis-then-execution. Rumelt's linear model (diagnose → policy → action) may need to become iterative (diagnose → test → refine → scale).
Use Rumelt's framework as a diagnostic tool for your AI roadmap. For each AI initiative, answer: (1) What specific problem does this solve? (Be concrete: 'Reduces customer support ticket resolution time from 48 hours to 4 hours' not 'improves efficiency.') (2) Why can't we solve this without AI? (If we can, do that instead.) (3) What resource constraint does this policy impose? (Budget, team size, data requirements, timeline.) (4) How does this initiative connect to others? (Does it share infrastructure, talent, or customer value with other AI projects?) If you can't answer these clearly, the initiative is likely bad strategy. Red flags: 'We need AI to stay competitive,' 'Our board expects an AI strategy,' 'We're exploring AI opportunities.' Green flags: 'We're reducing manual work in X by Y% using AI model Z, which costs $W and serves N customers.' Operators should demand this specificity before funding.
Take one AI initiative your organization is considering or currently running. Write answers to these questions: (1) What is the specific, measurable problem this AI solves? (2) What would happen if we didn't pursue this—would we lose customers, margin, or capability? (3) What is the one constraint (budget, timeline, team size, data) that will determine success or failure? (4) How does this AI project reinforce or conflict with other technology investments? If your answers are vague, you may have bad strategy. If they're specific and connected, you likely have good strategy. Share your answers with your leadership team and see if they agree on the diagnosis.
Richard Rumelt
Good Strategy/Bad Strategy: The Difference and Why It Matters
“Rumelt defines good strategy as having three elements: diagnosis of the challenge, guiding policy, and coherent actions. Bad strategy mistakes goals for strategy and lacks diagnostic rigor.”
Richard Rumelt
Good Strategy/Bad Strategy (2011)
“Strategy is not a wish list or a collection of goals. It requires honest diagnosis of competitive position, a clear policy that constrains choices, and coordinated actions that reinforce each other.”
OperatorRadar Editorial
AI Strategy Framework Application
“AI's capital intensity and rapid evolution make bad AI strategy more expensive than bad strategy in other domains. A poorly diagnosed AI initiative can burn $2M+ annually with minimal competitive advantage.”
Your AI strategy may be theater. Let's diagnose it. OperatorRadar's strategy review service applies Rumelt's framework to your AI roadmap, identifying which initiatives are coherent competitive moves and which are expensive distractions. Book a 30-minute diagnostic call.
Request implementationNeed a custom implementation?
Have Ekofi Lyrae design and implement AI agents and automations.