Why Faster Technology Makes Human Judgement More Important
AI has changed my working life far more than I expected.
When I went all-in on my executive coaching business in 2024, I was always going to make it work. AI did not create the ambition or make the decision for me, but it has made building the business materially easier. I estimate it saves me around 40% of the time I might otherwise spend on administration, research, organisation and content support. In practical terms, I believe that has probably saved me employing at least one additional full-time person.
For a small business, that is significant. I can explore an unfamiliar subject, find research, compare different perspectives, structure my thinking and get from a question to something useful considerably faster than I could only a few years ago.
Yet there is an important contradiction sitting underneath all that additional speed. AI can remove one bottleneck from a business while exposing another one altogether.
The human.
The Human Bottleneck Is Not New
I learned about becoming a bottleneck long before generative AI existed.
Early in my management career, in the late 1990s and early 2000s, I was ambitious, energetic and desperate to be involved in everything. I wanted to lead from the front, work alongside the team, take on additional projects, solve problems and keep moving. I was working six and a half days a week and, at the time, I genuinely loved much of it.
I thought I was being a fantastic employee and a committed manager. To a large extent, I probably was. But after a couple of years I began to recognise another consequence of that behaviour: I was getting in the way.
There were capable people around me with ambitions of their own, yet I was taking too much responsibility. I was not always giving them the space to make decisions, make mistakes and develop. Because I wanted to be involved in everything, too much activity naturally started to involve me.
I had become the bottleneck.
That experience has stayed with me because the underlying leadership problem has not really changed. What has changed is the speed at which technology can now magnify it.
Scepticism Can Become a Bottleneck Too
Today I see another form of bottleneck among senior professionals: scepticism.
I work with doctors, lawyers, CEOs, founders and other senior leaders, many of whom have spent years developing deep expertise. It would be too simplistic to say that higher intelligence automatically creates greater scepticism, and I have not seen evidence that would justify making that claim. But I do think there is an interesting question around whether expertise can sometimes make highly capable people more cautious about trusting technology they do not fully understand.
Some of that caution is entirely justified. Doctors and lawyers, for example, operate in environments where confidentiality, professional accountability and data security are not optional considerations. In healthcare, The BMJ has highlighted the difficulty of placing clinicians “in the loop” as the ultimate safeguard for AI systems when those clinicians may already be working at the limits of their available time and attention.
There is therefore nothing wrong with scepticism. I remain sceptical myself despite being an enthusiastic daily user of AI.
The difficulty comes when sensible scepticism becomes permanent avoidance. A leader who refuses to explore tools that could broaden their knowledge, accelerate research or remove low-value work may eventually become just as much a bottleneck as the leader who insists that every traditional decision must still come through them.
By 2026, simply refusing to engage is becoming increasingly difficult to defend.
The Opposite Extreme Is Just as Dangerous
The other group worries me just as much…. if not more, and that is the group who trust AI too readily.
AI can produce something polished, articulate and completely wrong with remarkable confidence. I experience that regularly enough that I have learned not to equate the confidence of the answer with the reliability of the answer.
That means my relationship with AI is deliberately argumentative. I use it to challenge me, but I challenge it back. I ask it to find weaknesses in an argument, then I question its conclusions. I use it to accelerate research, but I still want to understand the evidence underneath the summary.
Research from Microsoft involving 319 knowledge workers makes this particularly interesting. The researchers found that greater confidence in generative AI was associated with less critical-thinking effort, while greater confidence in the individual’s own expertise was associated with more critical thinking. The study also suggested that AI changes the nature of critical thinking towards verification, integration and oversight of the output.
That feels very close to what I experience in practice. AI has not reduced my responsibility to think. In some circumstances, it has increased it.
I explored this earlier in The AI Burden No One Talks About, where I argued that AI can create an additional psychological and cognitive burden for senior people even while reducing administrative workload. The problem is not simply learning how to use AI. It is learning how much to trust it, when to challenge it and what should never be delegated.
More Information Can Produce Less Clarity
One of AI’s greatest strengths is its ability to give us access to more information extraordinarily quickly. That is also one of the areas where I think another human bottleneck is beginning to emerge.
A leader can ask for five options, then another five. They can ask for an opposing argument, additional research, another scenario, a risk analysis and a completely different interpretation. What once required several hours of research can now appear on the screen in minutes.
Initially, that feels incredibly liberating. But there comes a point at which more information stops improving the decision.
It can start muddying the water.
Someone who previously combined experience, instinct and available evidence to reach a judgement can find themselves confronted with almost unlimited optionality. The temptation is then to search for one more answer before deciding.
I see that as an extension of something I have written about before in The Decision Fatigue Nobody Talks About at Senior Level. Senior leaders were already processing a relentless flow of decisions, interruptions, people issues and unnecessary escalations before generative AI increased the volume of information available to them.
Microsoft’s research into what it calls the “infinite workday” gives some indication of the environment into which AI has arrived. Among the highest-volume Microsoft 365 users studied, employees were receiving an average of 275 meetings, emails or chat interruptions across a day; effectively one interruption every two minutes during core working hours.
AI may help us deal with that workload. But if every hour it saves is immediately filled with more analysis, more communication and more possibilities, we have not necessarily created more space to think.
We may simply have accelerated the treadmill.
Faster Is Not Automatically Better
This is why I think businesses need to become more sophisticated in how they measure AI productivity.
If a person creates a report in 20 minutes instead of two hours, that is a productivity improvement. But if the report is weaker, insufficiently checked or contains a confident factual error, speed has not created value.
The Financial Times recently argued that judgement will become more important as AI assumes more routine work. Andrew Likierman, professor of management practice at London Business School, makes the case that judgement combines knowledge, experience and personal qualities, and that it can be deliberately developed rather than treating it as some innate leadership gift. Is he right?
Harvard Business Review has made a related point from an organisational perspective. As companies attempt to use increasingly capable AI agents, one of the constraints becomes their ability to make previously implicit decision-making processes explicit. Technology can execute a process, but organisations first need to understand how their best people actually arrive at good decisions.
That distinction matters.
AI can accelerate execution. It cannot automatically tell us what deserves to be executed.
This is closely connected to my earlier article, When AI Makes Everyone Faster, Clarity Becomes the Advantage. If every competitor eventually has access to similar levels of technological speed, speed itself becomes less distinctive. Knowing where to direct that speed becomes considerably more important.
Human Oversight Still Has Limits
Medicine provides a particularly powerful example of the problem.
The BMJ has questioned whether simply keeping a clinician involved is enough to guarantee safe AI use. Clinicians can be influenced by incorrect algorithmic recommendations, while fluent and plausible AI outputs may actually create additional cognitive work because somebody still needs to verify them.
That principle travels well beyond medicine.
A lawyer remains accountable for legal advice. A CEO remains accountable for the strategy. A board still owns the decision it approves. A leader cannot reasonably defend a poor judgement by saying that the AI recommended it.
Accountability cannot be outsourced.
Neither, in my view, can the final judgement.
That does not mean human beings will always be right. Clearly we are not. It means that somebody has to take responsibility for balancing the evidence, the circumstances, the people involved and the consequences of acting.
Adoption Is About More Than Buying the Technology
None of this is an argument for moving more slowly on AI adoption.
Quite the opposite.
McKinsey’s 2025 global survey found widespread organisational use of AI, but considerably fewer businesses had successfully moved from experimentation towards enterprise-wide scaling. Importantly, the organisations generating the strongest value were more likely to be redesigning workflows and defining where human validation was required rather than simply adding AI tools to existing ways of working.
That is where I think many organisations still have work to do. Giving somebody an AI licence is not transformation. Nor is asking them to perform exactly the same processes they performed five years ago, only slightly faster.
The real opportunity is to ask what work AI should remove, what work it should improve and what work should remain resolutely human.
AI Has Exposed Leadership
When I ask myself whether AI has changed what good leadership looks like, my answer is increasingly no.
It has exposed it.
Good leaders have always needed judgement, curiosity, self-awareness and the ability to distinguish important information from noise. They have always needed the confidence to act without perfect information and the humility to challenge their own assumptions. AI simply places those qualities under a brighter light.
It is also one reason I believe the need for genuine thinking space is increasing. Some of the senior people I work with already operate with almost no mental white space. We occasionally talk as though finding ten minutes while walking the dog constitutes meaningful reflection. Compare that with the amount of uninterrupted space leaders would once have experienced simply because information could not reach them every second of the day.
Coaching cannot and should not compete with AI on information retrieval. AI is extraordinarily good at retrieving information and it can ask very good questions.
But effective coaching is not merely asking a clever question. It is recognising when to ask it, noticing hesitation, seeing what is not being said, hearing a change in tone and understanding when challenge is required and when silence will achieve more.
For now, those things remain profoundly human.
AI has already made me substantially more productive. I expect it to become considerably more capable from here, and businesses that fail to engage with it risk being left behind.
But the faster the technology becomes, the more important the person directing it becomes too.
The human bottleneck has always existed. AI is simply making it much harder to ignore.
References and Further Reading
- Lee, H-P. et al. (2025), The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers, Microsoft Research / CHI 2025.
- Likierman, A. (2026), Leaders’ Judgment Matters More Than Ever in the Age of AI, Financial Times.
- Stave, J., Kurt, R. & Winsor, J. (2026), Teach Your AI How You Make Decisions, Harvard Business Review.
- Toro-Tobon, D. et al. (2026), Clinician in the Loop: A Flawed Solution for AI Oversight, The BMJ.
- McKinsey & Company (2025), The State of AI in 2025: Agents, Innovation, and Transformation.
- Microsoft Work Trend Index (2025), Breaking Down the Infinite Workday.
I’m Laurence Loxam – I’ve pushed limits in business, on mountains, and at the finish line.
Now I help elite professionals do the same, pushing past the point most people stop. I coach CEOs, doctors, lawyers and founders who have achieved success but still feel there is another level available to them.
Together, we unlock clarity, sharpen confidence and lead with conviction.
Ready for your next leadership breakthrough? Let’s connect.
Leave a Reply