We are still early in the corporate deployment of LLMs, which are sold as a productivity multiplier. What everybody seems to overlook is that adopting AI with the pure focus on efficiency will result in cognitive offloading at the exact organizational layers where independent judgement once created differentiation.

Today’s LLMs optimize for statistically probable outputs, which by definition means average outputs. When you now deploy them across an industry, you get identical strategies, identical risk assessments, identical designs, identical code, identical “inventions” etc.

This will hit enterprises the hardest, because their incentive structures reward short-term results over long-term depth. The larger the enterprise, the more will they default to maximum AI integration. They will – as a result – lay off significant parts of their workforce and end up with a remaining workforce – the remaining “best” – who at that point have shifted from active analytical or quite contrarian work to purely reviewing and accepting AI outputs and decisions passively (let’s call it “blindly”).

The organizations that will survive are not the ones with the best and fastest AI implementation but the ones that maintain – deliberately, expensively, and against every efficiency incentive – the human non-algorithmic, non-AI, non-cloneable “anomaly” of human judgment and intuition.

The companies that will win this decade will – IRONICALLY – not be the ones that automate the most reasoning, but those who strictly defend their strategic intuitive intelligence from AI. The winners are those who have leaders with the courage to allow individuals to hold an uncomfortable, non-data-supported position long enough to be proven right (what used to be called conviction).