“You are listening to a machine. Do the world a favor and don’t act like one.”
– WarGames, 1983
In the early 2000s, after watching senior executives struggle with drawing strategic frameworks, I posited that a computer system could help, theorizing computers could aid the strategic process. INSEAD’s Chan Kim and Renée Mauborgne became aware of the work – by then documented in places like Wayback Machine – built on it, and soon brought the then-obscure idea to Fontainebleau as a research project.
Years later, while at INSEAD research, Peter Zemsky and I tested the system with both MBAs and executives, our findings detailed in an HBR article, Can GenAI Do Strategy?
The field has blossomed. Felipe Csaszar and others, especially Hyunjin Kim (who joined the project early), delved into the question. At the same time, I left academia to commercialize the system, now called VSTRAT.
Given the subjectivity of the subject, whether GenAI can do strategy remains an open question, though it appears to be trending in that direction. However, much like Pandora, there remains the question of whether it was a good idea to open that box. It’s not “Can GenAI do strategy,” but “Should GenAI do strategy?” No matter how strong the intelligence, is it a good idea to defer one of our most subjective decisions to a very bright robot? What about the mass unemployment many predict? Will we surrender our deepest cognitive work to machines? Should we?
WHAT HAPPENS WHEN THE KIDS OUTPERFORM THE PROS
As an experiment, Andrew Reeves, director of the Center for Transformative Technology at the University of Missouri, invited me to a daylong seminar. We would try the strategy system with a group of undergraduate students, none of whom had any background in strategy. After an introduction they were broken into groups and asked to model a business using my research project.
One group after another eventually presented, outlining ideas, segmentation, pain points along a journey map, personas, and cost reduction or willingness to pay impact. Using the harnessed AI system, these kids, most too young to go to buy a beer, did significantly better than the professionals I routinely hear.
On feedback, a few suggested they could do the same with ChatGPT. That’s unlikely. For the same reason it is theoretically possible for any of us to build our own product catalog, payment processing, and shipping logistics, we use Amazon. That’s because the integration of the information, the harness, the value. Most users do not even register that Amazon is doing all of that work internally. Structured systems that use GenAI at the core work similarly. The model is the underlying capability: the harness is what makes the capability useful.
THE MODEL IS THE ENGINE. THE HARNESS IS THE CAR
We’ve watched models evolve and they’ll no doubt continue to evolve at a breakneck pace. They’ll become more intelligent than the smartest of us. But, like Christopher Langan who repeatedly tested with an IQ of about 200 but ended up working as a bar bouncer, will the machines ever make actionable strategy? We’ve all seen brilliant work from elite consulting firms, work that cost a small fortune in time and money, turn into paperweights on desks. Assuming GenAI produces a brilliant strategy, will it face the same fate? Should it?
We have seen this before. Game theory gave us the ‘ideal’ price decades ago, yet humans still haggle. Why? Because strategy isn’t just a math problem; it’s a social contract. What happens if we automate the ‘brilliance’ but lose the human buy-in? Or worse, automate the very people who are supposed to execute and consume that brilliance? We aren’t just opening Pandora’s box: we’re building a high-speed engine for a world where nobody is left to drive.
Paraphrasing Professor Csaszar, two AI-related things cannot be outsourced. Goal-setting at the beginning, which is what we care about and why, and accountability at the end, which is who answers for the outcome. The AI will not personally go bankrupt or go to jail. The executive will. Evaluation belongs at the front of the process. Risk belongs at the back. What sits between them is evolving, but the bookends are human – or they are nothing.
The machines are brilliant at matching patterns, and yet the consensus pattern has proven problematic in the past, whether the dot-com bust, the real-estate and related financial instrument bust, or even misguided political movements through history. Patterns the machines confidently mimic as correct often aren’t.
If everybody creates a brilliant strategy, does anybody listen? Or, more bluntly, if predictions of AI job losses are correct, who is left to purchase the intelligence?
Can GenAI do strategy? Eventually, most likely. Should it? The jury is out.
Michael Olenick, JD, is a former INSEAD research fellow and founder of VSTRAT.ai. He has spent four decades building AI and expert systems; his research has been used at INSEAD and, before then, cited by congress and media into how to resolve the 2008 era financial crisis. Michael currently lives with his partner, Anastassia, in Austin, Texas.
© Copyright 2026 Poets & Quants. All rights reserved. This article may not be republished, rewritten or otherwise distributed without written permission. To reprint or license this article or any content from Poets & Quants, please submit your request HERE.
