In boardrooms from Silicon Valley to Switzerland, I hear the same refrain. Executives present their AI road maps with breathless excitement, showing charts where labor costs go down and speed goes up. They describe a future where AI agents act as “minions for managers” — tireless, low-cost replacements for the junior staff, the writers, the coders and the analysts.
To a director responsible for governance and efficiency, this looks like a windfall. To a theoretical neuroscientist, it looks like a lobotomy.
Far too many companies are walking into this trap. They are using AI to automatecognition rather than augmentit. While this approach yields a “sugar rush” of productivity in the near term, the data suggests it will lower the collective intelligence of the organization — and the life outcomes of its people — in the long term.
The Minion Mistake
The current corporate strategy for AI is dangerously simple: identifying “well-posed” problems — tasks with clear instructions and right answers — and handing them to an algorithm to solve. We let the AI write the marketing copy, debug the code or screen the resumes.
This creates a “zombie autopilot.” The organization hums along, replicating its past successes with increasing speed. But an organization run by zombies has a fatal flaw: It cannot solve “ill-posed” problems.
Ill-posed problems are the messy, chaotic challenges where the question isn’t clear and the data doesn’t exist yet. A market crash, a reputational crisis, a paradigm shift in consumer behavior — these are problems that require human judgment, resilience and “intellectual trespassing.”
When you use AI to replace the “doers” — the junior associates and apprentices — you aren’t just cutting costs. You are cutting the ladder of career and even cognitive development. By preventing your own workforce from building the neural pathways required to become the experts of tomorrow, you are eating your own seed corn simply to boost this quarter’s margins.
The Science of Atrophy
My research focuses on hybrid intelligence — benchmarks for the collective intelligence of human-AI collaboration rather than for either alone. I’ve found that the smartest system on the planet isn’t a trillion parameter LLM. It is a high-functioning human team augmented by AI.
Fascinatingly, the performance of these superintelligence hybrid systems depends overwhelmingly on the human capital of the people involved, not the AI benchmarks. If humans are passive — if they treat AI as a “minion” that does the thinking for them — not only is hybrid intelligence reduced to that of the AI alone, their own cognitive abilities atrophy. They lose the ability to spot hallucinations, challenge assumptions and innovate.
Already, in the data, we can see that excessive cognitive automation creates a “deskilling” effect. When humans stop wrestling with the difficult, boring, foundational work, they lose the deep, model-based understanding of their craft. They become supervisors of a black box they no longer understand.
The Fiduciary Duty of the Board
As directors, you are the guardians of the company’s longevity. You must recognize the “efficiency trap” as an existential risk.
If your management team presents an AI strategy focused solely on head-count reduction and task automation, they are prioritizing short-term extraction over long-term value creation. They are building a brittle organization that will shatter the moment the world presents a problem AI hasn’t seen before.
What should a board do? You need to change the metrics of success.
Demand augmentation, not automation. Ask your CEO, “How is this AI making our people better, not just faster?” If the answer is, “It saves them time,” push harder. Time to do what? If the time saved is just filled with more busy work, you have failed. The goal should be to use AI to lift employees out of rote tasks so they can tackle harder, higher-value, ill-posed problems.
Audit the “learning pipeline.” If AI replaces the entry-level work, how do your junior employees learn? You must demand a plan for “robot-proofing” your talent. How are you simulating the “hard knocks” experience that builds resilience and judgment?
Measure “return on future.” Stop measuring AI success by the cost of labor saved. Measure it by the increase in your organization’s adaptive capacity. Are your teams solving problems today that they couldn’t solve six months ago? That is the only metric that matters in an exponential age.
The Choice
We are standing at a bifurcation point. We can use AI to build a “Jiffy Lube economy,” where humans are reduced to button-pushers for machines that hold all the knowledge. Or we can build a robot-proof economy where AI serves as a scaffold that allows humans to reach new heights of creativity and agency.
The former is cheaper next quarter. The latter is the only way your company survives the next decade.
It is time to stop treating AI as a cost-cutting minion and start treating it as a tool for human expansion. The board must ensure the company isn’t just becoming more efficient at being mediocre.