Some time ago, an experienced engineer told me a story that has stayed with me.
He was involved in a motorway bridge project. The bridge was being prefabricated and would eventually be transported into position and connected to the motorway.
As part of the engineering work, calculations had been undertaken to determine the load-bearing capacity of the bridge and its connections. Essentially, how much weight could the structure safely support?
Modern engineering software had been used to perform the calculations.
But when this engineer reviewed the results, something didn't look right.
He couldn't immediately identify the problem. It was more of an instinct, developed through years of working with structures, calculations and the underlying principles of engineering.
Rather than simply accepting the numbers, he went back and questioned them.
He discovered a serious error in the load-bearing calculations, one that had significant implications for the safety of the structure.
What struck me wasn't simply that an experienced engineer had caught a mistake. It was how he caught it.
He had spent years developing an intimate understanding of the mathematics and engineering principles behind the calculations. He knew what a reasonable answer should look like, even before he could explain precisely why the numbers were wrong.
The software could produce the calculations. But it took human judgement to recognise that something wasn't right.
That experience has become increasingly relevant as I think about the way we're embracing artificial intelligence.
When information is no longer the advantage
For much of our working lives, knowledge has been a source of professional advantage.
People built expertise through study, experience, research and years of learning their craft. Knowing something others didn't, or being able to access and interpret information that wasn't readily available, carried considerable value.
AI is changing that equation.
Today, we can ask AI to analyse a market, interpret financial data, summarise complex research or compare strategic options. Tasks that once required significant time and specialist resources can often be completed in minutes.
And that's a remarkable development.
It opens possibilities for individuals and smaller organisations that previously couldn't afford the expertise or resources needed to undertake that work.
But it also raises an interesting question.
If everyone has access to increasingly sophisticated information and analysis, where does the real advantage lie?
I suspect it's shifting away from simply possessing knowledge and towards something more difficult to replicate.
Judgement.
Knowing something and understanding it are not the same
One of the things I find fascinating about AI is how convincingly it can present information.
Ask a complex question and you'll often receive a well-structured response, complete with explanations, comparisons and recommendations.
Everything appears logical and considered.
It's easy to feel that the thinking has already been done.
But a convincing answer isn't necessarily the right one.
Information can be accurate yet incomplete. An analysis can be technically sound but based on questionable assumptions. A recommendation can make perfect sense on paper while being entirely unsuitable in practice.
Consider a business contemplating a major organisational change.
AI can help analyse operating costs, identify inefficiencies, compare structures and model potential outcomes.
All useful information.
But knowing whether the organisation is ready for change, understanding the relationships and personalities involved, recognising resistance that hasn't been openly expressed, or deciding whether the timing is right requires something more.
It requires context, experience, curiosity and an understanding of people.
And perhaps most importantly, it requires knowing which questions haven't yet been asked.
That's an important distinction.
The value of judgement isn't simply choosing between the options presented to us. Sometimes it's recognising that the options themselves need reconsidering.
The danger of trusting the answer
There's another aspect of AI that I think deserves our attention.
The more capable these tools become, the easier it is to accept their outputs without sufficiently questioning them.
Not necessarily because we're careless, but because the answers arrive so quickly and with such apparent confidence.
Why spend an hour working through something when a credible explanation can appear in seconds?
Research by Lee et al. (2025), published in collaboration with Microsoft Research and Carnegie Mellon University, examined how knowledge workers used generative AI and its relationship with critical thinking.
One finding was particularly interesting. People who expressed greater confidence in AI's ability to perform a task reported engaging in less critical thinking effort.
That doesn't establish that AI is making people less capable of thinking. But it does raise a concern about how readily we might defer to technology.
After all, critical thinking isn't just something we know how to do. It's something we practise.
We strengthen it by challenging assumptions, considering alternatives, working through uncertainty and occasionally discovering that our original thinking was wrong.
If we increasingly allow technology to do those things for us, I wonder whether we risk becoming less inclined to do them ourselves.
The real danger may not be that AI sometimes produces incorrect answers.
It may be that we become so accustomed to receiving convincing answers that we stop looking closely enough to recognise when something is wrong.
Experience, curiosity and the confidence to question
I don't believe good judgement comes simply from having spent many years in a profession.
Experience certainly matters, but we've all encountered people with considerable experience who remain firmly attached to outdated thinking.
Equally, someone relatively new to a field can sometimes ask a question that challenges years of accepted practice.
Good judgement, I suspect, develops through a combination of knowledge, experience, curiosity and a willingness to question our own assumptions.
It's about recognising patterns, understanding context and being comfortable with the possibility that the obvious answer may not be the right one.
And that is where I see an enormous opportunity for AI.
Rather than simply asking it to give us answers, we can use it to challenge our thinking.
What assumptions am I making? What alternative explanations should I consider? Where might my reasoning be weak? What important information could I be overlooking?
Those are very different questions from simply asking AI what we should do.
One approach risks outsourcing our thinking. The other has the potential to strengthen it.
For leaders in particular, that distinction matters.
The ability to make sound decisions has never depended solely on having the most information. It also involves knowing what matters, recognising uncertainty, weighing competing perspectives and taking responsibility for the consequences.
AI can make that process considerably better.
But only if we remain actively involved in the thinking.
Knowing when something doesn't look right
I keep coming back to the engineer and his motorway bridge.
He didn't question the calculations because he distrusted technology. He questioned them because he understood what the numbers meant.
His experience had given him the ability to recognise something that the software hadn't flagged.
There is an important lesson in that.
As information and analysis become increasingly accessible, our value may lie less in our ability to produce answers and more in our ability to interpret, challenge and apply them.
Knowing when to trust the analysis and when to look again.
Understanding what the information tells us, but also recognising what it doesn't.
And having the confidence to ask a question even when everything appears to make sense.
Perhaps that's one of the more interesting challenges AI presents.
Not simply how much more we can know, or how quickly we can obtain an answer.
But whether we're becoming better at understanding what those answers actually mean.
Because when information gets cheap, judgement becomes more valuable.
References
Lee, H.-P., Sarkar, A., Tankelevitch, L., Drosos, I., Rintel, S., Banks, R., & Wilson, N. (2025). The impact of generative AI on critical thinking: Self-reported reductions in cognitive effort and confidence effects from a survey of knowledge workers. Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems, Article 1121, 1–22. https://doi.org/10.1145/3706598.3713778