Strategic Management Explorer
AI Will Help Formulate Strategy, But Should Never Finish It
By Michael Olenick
As a former research fellow at INSEAD, I was never entirely comfortable with one piece of the canon: the idea that technology is not a key factor of competition. I had devoted much of my life to the premise that machines can help people think — in business, in law, and in life. The suggestion that machines were not where value comes from did not sit well.
The resolution reframed my thinking. Technology is never a key factor in competition, but it is catalytic to value. Cinematic production equipment is not value; it is what lets Cirque du Soleil stage dramatic spectacles inside a tent. Food science is a chemistry lesson; applied to wine, it enabled a consistent product to be mass-produced for the American palate. Technology never appears on the strategy canvas. It unlocks the entire canvas.
Ron Adner and Peter Zemsky formalized this logic years earlier in their published article in the Strategic Management Journal: supply-side technology creates advantage only through the consumer value it unlocks on the demand side. Catalytic, not competitive.
The Practitioner “So What”
AI is catalytic to strategy, not competitive with it. Like production equipment inside a Cirque du Soleil show, generative AI can make the strategy process faster, more rigorous, and better evidenced, but it never becomes the value itself, and it should never make the final call. That call stays with people who own the outcome.
From Research Project to Working System
Can a computer aid in making strategy?
This question was personal: I’d spent years exploring whether computers could meaningfully aid in the formulation of strategy, at a time when most considered the idea far-fetched. I believed computers could help strategists to whatever extent the AI of that era allowed, and I remain grateful to INSEAD for financing that work when most people considered it bunk. I spent a decade there writing software, cases, and articles.
When generative AI arrived, I blended it into VSTRAT, my then twenty-year-old system, and catalytic stopped being a theory. Peter Zemsky, then Deputy Dean of INSEAD, and I wrote about our early findings in Harvard Business Review, including an experiment in which the AI produced, in about an hour, a Blue Ocean strategy comparable to one an
INSEAD MBA team spent a week building.
Eventually, I left INSEAD to commercialize VSTRAT. Having pioneered this space, I may see its limits more clearly than those newer to it. The machines are genuinely remarkable, trudging through the mud of strategic formulation with the dexterity of McKinsey’s best. Felipe Csaszar and colleagues have shown large language models can generate and evaluate strategies at a level
comparable to entrepreneurs and investors. The Kim in that citation is INSEAD’s Hyunjin Kim, who worked on VSTRAT from its earliest days, so the finding is anything but abstract to me. What the machines cannot do is make the final commitment. Only people can.
I asked why the machine cannot simply produce a finished strategy deck. It can, I answer. The deck would be beautiful, compelling, and almost certainly wrong. Setting aside that you would not want it to even if it could, the truth is it can’t.
Only a strategist knows their people, processes, suppliers, customers and even competitors. More to the point, only a person knows what a person cares about. The model may find a real way to raise willingness to pay or lower cost. But if the move does not sync with what executives care about, the output becomes a beautiful paperweight.
The Decision Stays Human
Technology in strategy is what it has always been: catalytic. Generative AI will sit inside the strategy process the way production equipment sits inside a Cirque du Soleil show, invisible in the final product and indispensable to enabling it.
Practical Implications for Strategic Leaders
-
-
-
- Use AI-guided frameworks to do the grinding work of strategy formulation (data-gathering, option generation, value-curve construction), rather than leaving it to whoever has a free week.
- Deploy AI-simulated stakeholders (skeptical customers, cautious CFOs, procurement officers) to pressure-test a strategy before it reaches the boardroom.
- Require that AI-generated outputs be evidence-grounded and citation-backed, not accepted on the model’s authority alone.
- Keep a named human accountable for every strategic commitment; never let “the AI recommended it” substitute for that accountability.
- Always watch for cognitive surrender: build review processes that reward overriding a flawed AI recommendation, not just following one, so the liability asymmetry doesn’t quietly erode judgment.
Strategy formulation is being transformed, and organizations that ignore this will be out analyzed by those that do not. The strategic decision stays where it has always lived: with the people who must own it, execute it, and answer for it.
The rule fits on an index card: AI everywhere in the strategy process, AI nowhere in the strategy decision. That is not a limitation of the technology. It is the definition of strategy.
Michael Olenick, JD, has spent two decades exploring the intersection of expert systems and business strategy. In 2002 he built the first version of an AI-aided strategy system, later bringing it to INSEAD as a research fellow, where he developed it further and co-authored ‘Can GenAI Do Strategy?’ in Harvard Business Review with then-Deputy Dean Peter Zemsky. He is founder of VSTRAT.ai, has authored seven bestselling HBSP teaching cases, and writes on AI and business education for Poets & Quants. Michael lives in Austin, Texas.
