I watched from the corner of the boardroom as an executive struggled to respond in real time to a potential disaster.
His company was preparing to relaunch the construction of the world's tallest and most environmentally sustainable skyscraper. The incomplete structure already stood 831 metres high. Now, the team had received credible reports that a large section of its glass exterior had fallen from the 60th floor and plunged to the ground.
We did not yet know whether anyone had been injured or killed, but it was a possibility. When the CEO bustled into the room for an emergency briefing, something extraordinary happened. The leader in charge of the comms team told her there was nothing to worry about. Everything was under control. In fact, nothing had happened.
The disaster scenario was not real. It was part of a business simulation I'd written then brought to life with a talented team of actors. It was designed to expose how executives influence others and make decisions under pressure.
I'd watched hundreds of executives confront this same situation. I had seen them investigate, hesitate, consult, take control, panic and occasionally make matters worse. Denying the incident entirely was a new one. It remains the only time I have stopped the simulation to examine a participant's reasoning before allowing events to continue.
Simulations reveal something that conventional promotion processes frequently miss. Making a snap decision quickly under pressure can look like leadership. Exercising wise judgement is something else entirely.
Who gets invited to step on to the leadership elevator in an organisation has long been an imprecise science. Research shows firms regularly make costly mistakes in promotion because they don't sufficiently distinguish between the skills that bring a person to prominence, as a great salesman, accountant or engineer for example, and the very different skills required to be a value-adding manager. This perennial and well-researched problem has become known as the Peter Principle: firms promote competent workers until they become incompetent managers.
A broken system means there are too many leaders who are unsuited to that position. In a 2026 study by the personality assessment firm Hogan, nearly 10,000 employees across 25 countries were asked a powerful question: What qualities do you value in a leader? Remarkably, it found absolutely no overlap between the top five competencies global execs display and the top five competencies their teams want. None.
Global execs exhibit abilities such as inspiring, competing and presenting to others, taking initiative and driving innovation. Yet their global followers wanted something different: for leaders to display skills such as effective communication and decision-making, accountability, integrity and actually being effective in the role of leading other people.
Isn't it interesting to study those different definitions of what it means to lead? For me, it shows that our current crop of execs learned to be good at building their personal brand to gain entry to that leadership elevator. It's not surprising their followers don't see the value in that self-centred rather political approach. Even less surprising that they desperately need their leaders to send signals it's safe to trust them.
Part of the problem is every organisational leadership pipeline contains a built-in time lag. Organisations observe the behaviours associated with success, turn them into competency frameworks, promotion criteria and development programmes, and then reproduce them. But, by the time those systems are embedded, the work may already have changed. Like generals preparing to fight the last war, organisations become highly efficient at producing leaders for the operating model that created their past success.
The speed at which AI is now changing work is making these traditional indicators of potential even less reliable. Research shows that using AI can substantially increase the apparent performance of individuals. This is particularly true for less-experienced or lower-performing people. A study by the US National Bureau for Economic Research that focussed on human customer-service agents found an average productivity improvement of 14%, including a 34% improvement for novice and low-skilled workers. The impact of using AI on experienced and highly skilled workers was minimal.
For decades, we've judged who should lead by who produces the most, works fastest or appears to possess the most specialist knowledge. AI has now flattened that playing field. This means the proxy to judge leadership potential, which didn't work very well anyway, is finally gone. Good riddance.
But, hold on, we now need a new way to pick our leaders. And my colleagues in HR and L&D have a challenge in supporting that. The Deloitte 2025 Global Human Capital Trends survey found that learning and development was the talent process most in need of reinvention due to AI-related disruption in work.
Here's how we can improve the situation.
Firstly, update our version of what good leadership looks like. To avoid fighting the last war we need to shorten the gap between identifying the winning leadership attributes and selecting for them. The emerging leadership capabilities are:
Curiosity and intellectual humility to be prepared to learn and take others with you
AI fluency, but only sufficient to understand its possibilities and limitations (leaders do not need to be coders)
Knowing when to rely on, challenge or override AI
The ability to redesign roles, decisions and workflows
Collaboration across human, functional and machine boundaries
Ethical ownership when responsibility might otherwise disappear into the system
The ability to generate trust around you – this one has always been important but is now, in such a fast-changing environment, even more vital
Secondly, help people learn on the job to widen the pipeline of potential leaders who can genuinely take responsibility in organisations that need to be flatter and more empowering.
Finally, provide meaningful opportunities in development programmes to test judgement – just like the challenge in the business simulation at the start of this article. Let's stop just teaching ideas and allow people to rehearse them.
What seems clear is good leaders now understand that their value has shifted from personally supplying the answers towards good judgement: deciding which questions actually matter, configuring how to address them, critically evaluating the result and accepting responsibility for the consequences.
What's not good enough is to simply add "AI literacy" to existing leadership frameworks. Sending yesterday's leaders on an AI course does not create an AI-era leadership pipeline.
This article was originally published on Forbes.com