The danger is not that AI will make bad strategic decisions. It is that AI could make bad strategic decisions appear extraordinarily intelligent, evidence-based and convincing. At unprecedented speed and scale. A recent Harvard Business Review article, “AI Is Revolutionizing Strategic Decision-Making”, explains how AI fundamentally changes strategy. It can generate thousands of alternatives, build continuously updated models of markets and competitors, and use AI agents to challenge competing proposals. The author calls this an unbounding of rationality. He also stresses that humans must still frame questions, evaluate evidence and remain accountable.
The capabilities are impressive. But read through the lens of STOP PREDICTING and some of my previous blogs about strategy and AI, and a much more disturbing possibility emerges.
1. AI is moving from answering questions to defining reality
I explored this in my September blog, “The real AI risk isn’t the answer. It’s the frame.” Traditionally, leadership teams framed strategic problems and used technology to analyse alternatives. AI increasingly participates in defining the problem, identifying relevant information, generating choices and evaluating their attractiveness. That represents an enormous transfer of cognitive influence. Whoever defines the problem also determines which possibilities are considered and which disappear from view. AI can generate thousands of options within a particular interpretation of reality without questioning whether that interpretation is useful. You can have extraordinary analytical sophistication and still fundamentally misunderstand the world. The danger is that organisations will confuse the number of alternatives examined with the quality of their strategic thinking.
2. We risk automating yesterday’s thinking
This connects directly to my blog about digital transformation and curiosity. Corporate AI systems increasingly draw on proprietary data, existing processes, established performance indicators and institutional knowledge. That information is valuable, but it also reflects how the organisation has understood the world until now. Imagine a company using AI to evaluate 10,000 strategies for improving its existing business model. The analysis could be brilliant. But what happens if the entire business model is becoming irrelevant? The system may optimise a business that should be reinvented. AI can identify emerging signals and challenge established thinking, but only if organisations are willing to explore possibilities beyond their existing business logic. Digital transformation without intellectual curiosity is the industrialisation of yesterday’s thinking. AI could turn that into the industrialisation of yesterday’s strategy.
3. The sea squirt problem becomes digital
Remember my blog about the sea squirt that settles on a rock and effectively abandons its brain? I used it to explain how successful organisations gradually dismantle their capacity to sense change. AI could accelerate that process. Why invest in uncomfortable conversations, external perspectives, dissenting employees and speculative thinking when a machine can deliver sophisticated strategic analysis in seconds? Over time, leadership teams could become increasingly dependent on AI-generated interpretations of reality. The technology becomes more capable while the organisation’s independent thinking becomes weaker. The paradox is extraordinary. We could build the most intelligent strategic systems in history while simultaneously making the people using them less strategically intelligent.
4. Synthetic debate is not necessarily genuine disagreement
The HBR article suggests using AI agents to debate, challenge and stress-test strategic options. That is genuinely promising. But there is a fundamental limitation. Ten AI agents debating a strategy are not necessarily ten independent perspectives. They may share models, training patterns, data sources and intellectual blind spots. They can produce something that looks like rigorous disagreement while reinforcing the same conclusions. This connects to my recurring argument about strategic diversity, mental sandboxing and challenging conventional thinking. A convincing simulation of disagreement is not equivalent to genuine dissent. You can automate the devil’s advocate without ever inviting a genuine heretic into the room.
5. Speed becomes dangerous when direction is uncertain
One of the central lessons of STOP PREDICTING is that I was wrong about roughly 80% of my attempts to anticipate what was coming. The real value came from recognising errors, learning from them and developing the capacity to change direction. AI changes the economics of strategic decision-making. Organisations can analyse more information, evaluate more alternatives and act faster. But speed amplifies both good and bad decisions. An organisation that is strategically blind but operationally brilliant can now move towards the wrong destination with extraordinary efficiency. This makes leadership agility and future fitness more important than ever. Adaptation velocity matters more than decision velocity. The crucial capability is not how quickly an organisation can reach a decision. It is how quickly it can recognise that reality has changed and adjust its course.
6. Certainty becomes the ultimate strategic vulnerability
Being wrong should be a lot more respectable than it is right now. There is a growing danger of confusing increasingly sophisticated representations of the future with genuine knowledge of what will happen. AI-generated strategies can be accompanied by detailed scenarios, financial forecasts, competitor simulations and compelling arguments. The result can feel like certainty. But an elegant model is not necessarily an accurate representation of reality. And because AI can produce convincing analyses continuously, organisations risk becoming addicted to an endless supply of apparent certainty. The future remains uncertain. Our tools are simply becoming much better at concealing that uncertainty.
The future fitness response
The HBR article is not advocating blind faith in AI. It explicitly recognises the need for human judgement and accountability. The danger lies in organisations adopting extraordinary analytical capabilities without developing the human capabilities needed to question and challenge them. I would introduce five disciplines into every AI-supported strategic decision:
- Expose the frame. Make the underlying logic, boundaries and exclusions explicit.
- Protect independent thinking. Require human perspectives before revealing the AI’s preferred strategy.
- Invite genuinely different perspectives. Use outsiders, unconventional thinkers and dissenting employees, not simply more AI agents.
- Define disconfirming signals. Identify what would demonstrate that the strategy is failing.
- Maintain the capacity to reverse course. Turn strategic decisions into testable commitments rather than declarations of certainty.
AI is becoming part of the infrastructure through which organisations perceive, interpret and respond to reality. That creates extraordinary opportunities. It also creates the possibility of institutionalising strategic blindness on a scale we have never experienced. Perhaps the most dangerous organisation of the future will not be the one that ignores AI. It will be the one that uses AI brilliantly, trusts its strategic intelligence completely and slowly loses the ability to recognise when it is wrong. That is a far more unsettling prospect than AI replacing strategic consultants.
The central lesson from STOP PREDICTING is that being wrong isn’t dangerous. Staying wrong is. AI could make staying wrong faster, more convincing and much harder to recognise. We are giving organisations superhuman analytical capabilities without ensuring they retain the very human capacity to doubt themselves.
