Consensus-based decision-making is too slow for the AI era. That is the argument in a recent Harvard Business Review article, and it deserves attention from any leadership team trying to move faster without losing control.
Why consensus breaks down
Consensus works when the environment is stable, information is widely shared and the cost of delay is low. None of those conditions holds today. Markets shift quickly, AI generates new signals at scale and competitors make moves in days rather than quarters.
When every decision requires broad agreement, organisations optimise for comfort rather than outcomes. Dissent is smoothed away, risk is diluted and accountability is unclear. The result is not better decisions; it is later decisions.
Speed and signal detection
The alternative is not authoritarianism. It is a decision architecture that separates signal detection from decision authority. Frontline teams and AI systems can detect signals — changes in customer behaviour, supply-chain disruption, competitive threats — and escalate them quickly. Leaders can then make decisions within clear authority boundaries.
This requires two things. First, explicit rules about who decides what. Second, a culture that rewards raising signals and acting on them, rather than punishing people for surfacing bad news.
Most organisations already have the data they need to decide faster. What they lack is the permission structure. Junior employees see problems first but do not escalate. Middle managers sit on information because they fear blame. Executives make decisions too late because they are waiting for perfect consensus.
AI changes the signal landscape
AI makes the case for redesign more urgent. AI systems can process far more data than traditional reporting, but only if the organisation is structured to act on the output. A signal that nobody owns is just noise. A recommendation that must pass through five committees before anything happens is worthless.
Companies that benefit from AI will be those that pair intelligent systems with clear decision rights. The technology exposes what is possible; the operating model determines whether anything changes.
Practical steps
A useful starting point is to map the ten to fifteen decisions that matter most in your business and ask: who owns each, what input is required, and how long does it currently take? In most organisations, at least a third of these decisions are slower than they need to be because the process is designed for consensus rather than speed.
Redesigning decision rights is less glamorous than buying an AI platform. It is often more valuable.