Share with your CEO
Most companies are buying AI speed and calling it strategy. Jeff DeGraff, a University of Michigan Ross School of Business professor, argues that’s a losing position: once every competitor can summarize faster and draft cheaper, the efficiency gains cancel out. His three-level AI leadership framework distinguishes between optimizing existing processes, using AI to stress-test assumptions, and genuinely co-creating with AI to produce outcomes neither human nor machine could reach alone. The differentiator, he claims, isn’t intelligence access. It’s judgment about which questions deserve asking.
What this means for your business
The uncomfortable truth for most executive teams is that nearly all current AI investment sits at Level One, the optimization layer, where the competitive half-life is already shrinking. If your AI roadmap is a list of efficiency wins, you’re not building a moat. You’re renting one. The question worth pressure-testing in your next strategy session isn’t “what can we automate” but “what would a well-resourced adversary see in our assumptions that we can’t.” That’s the Level Two posture, and most leadership teams haven’t built the habit.
DeGraff’s framework is genuinely useful, though it arrives from an academic innovation consultancy context that tilts toward the aspirational. His George Martin analogy is vivid but slippery: Martin’s creative instinct was irreproducible, which is precisely the opposite of a leadership framework you can install. The more defensible version of his argument is structural rather than romantic. Organizations that only use AI to answer existing questions will systematically under-invest in the harder capability, using AI to generate better questions. That gap is real and measurable in how AI initiatives get scoped and funded.
The leading indicator to watch is where your organization’s AI prompts originate. If they consistently come from execution layers asking for faster outputs, Level One is your ceiling by default. The leadership teams that will pull ahead are the ones treating AI as a sparring partner for strategy, not a faster typist. I’d revise this view if organizations using AI primarily for optimization demonstrably outperform those experimenting with adversarial simulation, but no evidence currently points that direction.
Based on reporting from Three levels of AI leadership every CEO needs, originally published 2026-07-26 09:35:00.
