Anita interviews Felipe Csaszar, Chair of Strategy at the University of Michigan's Ross School of Business, covering AI's role in strategy, organizational decision-making, and the future of strategizing.

Can AI Do Strategy? Felipe opens by discussing his new paper in Strategy Science, arguing that trying to teach AI to "do strategy" forces us to be precise about what strategy actually is — pushing back against the vague "hand-waving" that often passes for strategic thinking. If you can't specify a concept precisely enough for a machine to implement it, you don't really understand it.

The Role of Humans vs. AI in Strategy Both agree that AI will excel at the analytical side of strategy — market research, competitor analysis, data synthesis — but that purpose and aspiration must remain human. Felipe frames it as: AI can handle content (what to do) and process (how to implement), but the underlying utility function — what an organization truly wants to achieve — must be set by people. He also notes that markets will ultimately determine how much strategic authority gets delegated to AI, as investors will back whichever model performs better.

Organizational Structure and Bounded Rationality Felipe connects his earlier research on organizational structure (including his 2012 SMJ paper on mutual funds) to AI. His core insight: all of strategy and organization theory rests on bounded rationality — humans have limits in attention, knowledge, and consistency. AI has different bounds (vast knowledge, high consistency) but not no bounds, and this will lead to genuinely new organizational structures and decision processes. Silos, for instance, exist partly to concentrate expertise — but AI democratizes access to expertise, potentially reshaping how organizations are designed.

Human Judgment and AI Risk Anita raises the concern about AI treating everyone identically, with no contextual judgment (e.g., a fire chief knowing a firefighter is going through a difficult personal situation). Felipe's view: outsourcing high-stakes decisions to AI is only acceptable when there is strong evidence AI outperforms humans in that specific domain, with proper guardrails and simulated testing first. Just as we've accepted autopilots and anti-lock brakes, we may eventually accept AI in more consequential roles — but only with rigorous validation.

Advice for Companies Today Felipe compares the current AI moment to the early internet (1994): the technology exists, but the killer business models haven't been discovered yet. His advice: run many experiments, involve everyone, and devote significant time to exploration — this is precisely the moment to do so.

Closing Thought When asked what keeps him optimistic, Felipe points to the leverage of his work: strategy educators train future CEOs, meaning even small improvements in how strategy is taught ripple into some of the most consequential decisions in the world.