Who Approved This

Something strange has happened in boardrooms over the past two years. The people with the least understanding of artificial intelligence have become its most aggressive champions.

That is not a criticism of curiosity. It is a description of a structural problem. When boards push for faster AI implementation precisely because they do not understand it well enough to know what ready looks like, the organisation is no longer making strategic decisions. It is reacting to anxiety.

A BCG survey published this month captured it plainly. According to BCG, 61% of CEOs say their boards are rushing AI transformation, and more than half say hype is actively distorting boardroom judgment. The telling detail is this. Board members with the lowest confidence in their own AI knowledge are the most likely to believe the company is moving too slowly. Ignorance is accelerating urgency, not slowing it. In any other function, that dynamic would be called a governance failure.

This is what might usefully be called AI psychosis. Not the technical kind, where a system hallucinates. The organisational kind, where the pressure to appear ahead on AI has decoupled decision-making from judgment. Leaders are approving initiatives they cannot interrogate, setting targets they cannot validate, and calling it transformation.

The consequences are already visible. McDonald's spent three years deploying AI across more than 100 drive-thrus before pulling the plug in June 2024, after a run of viral incidents including an AI that kept adding Chicken McNuggets to a customer's order until it reached 260. The decision to scale had moved faster than any evidence that the thing was ready to scale.

That is an easy example to dismiss as a consumer novelty gone wrong. The harder version sits inside organisations, in the gap between how executives perceive AI performance and how implementation teams experience it. A 2025 study found that 75% of C-suite leaders felt their AI rollout had been successful. Only 45% of the people doing the actual work agreed. That is not a minor discrepancy. That is a company operating on two separate versions of reality at the same time.

The same research found that roughly 90% of data practitioners said their leadership held unrealistic expectations for what GenAI initiatives could technically deliver or commercially produce. Nine in ten. These are not a handful of sceptics grumbling at the back of the room. They are the people closest to the work, consistently reporting that the targets set from above bear little relationship to what is actually happening.

At board level, McKinsey found that two-thirds of directors globally report limited to no knowledge or experience with AI. That knowledge gap is not producing caution. It is producing pressure. A separate Dataiku study of 900 CEOs worldwide found that 62% say their boards are actively pushing for measurable AI outcomes, even as many of those same CEOs admit they do not fully trust or govern the systems they are being held accountable for. According to BCG, CEOs estimate that 35% of their performance evaluation is now tied to delivering AI ROI, nearly 10 points higher than boards realise. That is not strategy. That is a performance target attached to something nobody has clearly defined.

The problem with AI psychosis is that it rarely produces visible disasters. What it produces is an organisation in motion without direction. Pilots that never scale and governance documents nobody opens sit alongside training programmes that teach the tool while ignoring the workflow, and board updates that measure activity rather than impact. An organisation performing AI rather than using it.

This is the version that does not make the news but quietly touches every line of the P&L it was supposed to be improving.

What to do instead

Separate urgency from clarity. Boards wanting AI progress faster is not inherently wrong. The problem starts when urgency substitutes for definition. Before approving the next initiative, get clear on what it would need to deliver in business terms to be considered working. If that conversation has not happened, the funding decision is premature.

Close the perception gap. If leadership believes AI is performing well and the implementation team does not, one group is wrong. A structured audit of both perspectives is not a sign of weakness. It is the only reliable way to find out which version of reality the business is operating in.

Govern the decisions, not just the tools. Most AI governance frameworks focus on compliance and ethics. Both matter. But the more immediate question for most organisations is simpler. Who is authorised to approve AI initiatives, and against what criteria? Ambiguity there is not neutral. It is the condition under which the psychosis spreads.

Build board fluency before board pressure. The BCG data suggests boards with lower AI knowledge push harder for faster implementation. Investing in genuine board-level AI literacy, not a vendor presentation and a roundtable, is a prerequisite for rational oversight.

The companies that will look back on this period clearly are not the ones that moved fastest. They are the ones that noticed when pressure started doing the work that judgment should have been doing and stopped long enough to tell the difference.

𝛀

If you found this Independent A.I. Think Tank release useful, consider subscribing to D!sruptive newsletter. Built for business leaders who need to cut through the noise and understand what AI means for their organisation.

If you need the help discussed in this article, get in touch to see how we can work with your business 🇺🇸 +1 512 518 4196, 🇦🇪+971 585 629 975 or hello@byDisrupt.com