Why your AI strategy keeps failing on broken hardware

It is late, the second pilot has stalled, and you catch yourself thinking the sentence you would never say in the boardroom: I am exhausted and we have not even properly started, and some quiet part of me suspects the problem this time is me, not the technology. You have led transformations before. You have weathered ERP migrations, the move to cloud, the whole procession. This one feels different, and you are tired in a way that the work alone does not explain.

Here is the substantive answer, and it is good news once you sit with it. The exhaustion is real, it is widely shared, and it is pointing you at the right thing. An AI leadership mindset is not one more framework to bolt onto the org chart. It is the inner state behind your every decision: your capacity to stay clear, regulated and decisive while the ground under your expertise keeps shifting. Most leaders are trying to install new software on broken hardware. Upgrade the operator first, and the strategy that felt impossible starts to land.

Why do I feel like the bottleneck in my own AI strategy?

Because, increasingly, you might be. Not as a failing. As an accurate reading. The behavioural scientist Jeffrey Sanchez-Burks (a professor of management at the University of Michigan who studies how people work under uncertainty) named this precisely in 2026: competence vertigo, the moment you finally feel capable only to discover the bar has already moved again. He argues the self-doubt it produces is no longer a distortion but often an honest signal of an unstable environment, and that the result is exhaustion and quiet demoralisation. That is the feeling you had at 10pm. It has a name, and a great many leaders carry the same weight.

The scale of it is documented. DDI's Global Leadership Forecast 2025, drawn from 10,796 leaders across more than 50 countries, found that 71% report their stress has risen since they took their current role, up from 63% in 2022, and that 40% of those stressed leaders have considered walking away from leadership altogether. Deloitte's 2024 workplace well-being research, drawn from 1,050 C-suite leaders across the US, UK, Canada and Australia, found around four in ten executives say they always or often feel exhausted or stressed. So when you feel depleted before the real work begins, that is the room you occupy, not a private weakness.

Is my AI strategy actually a technology problem?

Mostly it is not, and the research is unusually clear on this. The bottleneck is no longer the technology. MIT's Project NANDA study of 2025 (a research initiative tracking how AI is actually being deployed inside companies) examined 300 public AI deployments alongside 150 leader interviews and a survey of 350 employees, and found just 5% of integrated pilots were extracting serious value while the vast majority sat stuck with no measurable impact on the P&L. The models work. The deployments do not.

Where companies do see returns, the dividing line is human, not technical. The Wharton School's 2025 AI adoption study, run with the consultancy GBK and built on a survey of more than 800 senior leaders at companies above 1,000 employees and 50 million dollars in revenue, found that 75% now report a positive return on their AI investment. The leaders running those returns name the constraint plainly. As the Wharton professor Stefano Puntoni puts it, adoption is "more of a human capital challenge than a tech challenge," and 43% of the leaders surveyed worry their people's skills are slipping rather than sharpening as the tools arrive. The hard part lives in the people and the way the work is shaped, not in the licence.

Read those findings together and a pattern resolves. The expensive part was never the software licence. It was the human system meant to wield it: the clarity to decide, the steadiness to lead through ambiguity, the willingness to change how the work itself is shaped. That is what is depleted at 10pm, and that is what the whole strategy now waits on.

What does upgrading the hardware actually mean?

It means treating your own state as something you can train and measure, the way an athlete treats recovery. This is the human operating system: the four intelligences you run on (your thinking, emotional, physical and intuitive capacities), and the coherence between your head and your heart that determines how well you decide under load. The category of intervention here is self-regulation, and it rests on hard science, not motivational gloss.

Consider how fragile expert judgement becomes when the operator is depleted. In a study of more than 1,000 parole rulings by eight experienced Israeli judges, favourable decisions ran near 65% at the start of a session, fell across the morning, then sprang back to roughly 65% after a food and rest break. The finding is contested in its magnitude, and worth holding lightly, yet the direction is intuitive to anyone who has chaired back-to-back decisions: a tired operator and a rested operator are different decision-makers. Your AI strategy is a long sequence of judgement calls. The state you make them from matters.

And it is measurable. The neurovisceral integration model proposed by Julian Thayer and colleagues (a framework linking the heart's rhythm to brain function) shows that resting heart rate variability, the small beat-to-beat changes in your pulse, indexes the very prefrontal circuits that govern attention, executive function and emotion regulation. Higher vagally-mediated HRV reflects greater self-regulatory capacity. This part of the work I hold as exploratory rather than settled, yet the through-line is firm: the calm, clear, recoverable leader is running better hardware, and it shows up in the choices they make. The leaders who win the next decade are the ones who upgrade themselves first.

So the sequence, where it is sequential:

1. Name the state. Notice the competence vertigo for what it is, an accurate reading of a moving environment, so it stops masquerading as personal failure.
2. Build the capacity. Train regulation and recovery as deliberately as you train strategy, measured against something real such as resting HRV.
3. Then redesign the work. From a regulated state, reshape the workflows and the ownership, the place where EBIT (earnings before interest and tax, the headline measure of operating profit) actually moves.

This is what I call the move into Phase Three. Phase One, the Age of Effort, was muscle: work hard, get a little more, linear growth. Phase Two, the Age of Scale, was machine: build once, sell to millions, exponential growth. Phase Three, the Age of Acceleration, is mind: output decoupled from human effort almost entirely, the phase AI unlocks, and it begins with the one operator you can actually upgrade today. The tiredness was the signal. Follow it, and the strategy you have been pushing uphill begins to move alongside the operator who runs it.

About the author

Thomas Green

British technology futurist, AI keynote speaker and advisor. Thirty years across enterprise technology and AI strategy, helping leaders navigate the future of work. The futurist who died.