A new practitioner article argues that generative AI should be woven throughout the strategy process but must never make the final strategic decision. The piece, published in the SMS Explorer, pushes back against the growing assumption that AI can take strategy off executives' plates entirely.
"AI Will Help Formulate Strategy, But Should Never Finish It," by Michael Olenick, draws on more than two decades of AI-aided strategy research, including his development of VSTRAT, an AI-aided strategy system he built in 2002 and later advanced as a research fellow at INSEAD. The article cites published findings that large language models can generate and evaluate strategies at levels comparable to entrepreneurs and investors. But Olenick insists that the ability to produce strategic outputs is not the same as the authority to commit to them.
The accountability problem
Olenick's central argument is not that AI lacks analytical power but that it lacks accountability. "Only the people who must execute a strategy and answer for its outcomes can legitimately make the final commitment," he writes.
That distinction matters as AI tools grow more capable and the debate shifts from whether AI can contribute to strategy to whether it should lead it. Olenick argues it should not, and he identifies precisely where the line must be drawn.
The article's core logic draws on Adner and Zemsky's research in the Strategic Management Journal, which established that supply-side technology creates advantage only through the consumer value it unlocks. That makes technology catalytic, not competitive. Applied to generative AI, the rule is clear: AI everywhere in the strategy process, AI nowhere in the strategy decision.
Olenick compares AI to production equipment inside a Cirque du Soleil show: indispensable to enabling the outcome without ever becoming the value itself.
Five practical implications for leaders
For strategic leaders, the article identifies five practical implications:
- Use AI for the grinding work of data gathering and option generation.
- Deploy AI-simulated stakeholders to pressure-test strategies before they reach the boardroom.
- Require citation-backed outputs rather than accepting model authority.
- Maintain named human accountability for every strategic commitment.
- Build review processes that reward overriding flawed AI recommendations to prevent the quiet erosion of judgment.
The distinction between generating and committing to strategy has become a central concern for AI for Executives & Strategy discussions, as organizations race to adopt AI tools without always clarifying who remains responsible for outcomes.
Why this matters for executives and strategy professionals
For executives, the article offers both a permission structure and a warning. Organizations that deploy AI throughout the strategy process will out-analyze those that do not. But the moment a firm lets AI make the final call, it has not accelerated strategy - it has abdicated it. Strategy professionals who want to apply these principles in their own work can find structured guidance in the AI Learning Path for Strategy Managers.
Your membership also unlocks:
