Over the past couple of years, I've watched countless conversations about AI. Some were exciting. Some were intimidating. Some were incredibly practical.
But after a while, I realised I didn't want to build my understanding from other people's conclusions. I wanted to take my time, build products with AI, make mistakes and form my own opinion.
This essay is simply a reflection on one of those observations.
Looking back, I realise I've been incredibly fortunate throughout my career. I've had managers, team leads and mentors who invested a lot of time in helping me grow. More than anything, they challenged how I thought. They encouraged me to question my assumptions, structure my thinking and understand problems before trying to solve them.
One of those mentors had a particularly lasting impact on me after I transitioned into Product Management.
I wasn't naturally good at structuring my thoughts. Like many people early in their careers, I was eager to jump into solving problems before fully understanding them. He noticed that. Instead of simply correcting me, he patiently helped me slow down.
Before we talked about features, we talked about users. Before discussing solutions, we tried to understand why the problem existed in the first place. Before making decisions, we explored constraints, trade-offs and previous decisions.
One day he handed me a copy of The Pyramid Principle. At the time, I thought it was just another business book. Only years later did I realise why he had chosen it. It wasn't really teaching me how to communicate. It was teaching me how to organise my thinking.
The better I understood a problem, the easier it became to explain it.
I still find myself coming back to that lesson.
When AI became part of my daily work, I approached it like many people probably did. I learned prompting. I experimented with different tools. I tried different techniques. Some worked better than others.
But after a while, I found myself asking a different question.
Is this really the skill I'm trying to develop?
The more I worked with AI, the more I noticed a pattern. Whenever I wasn't happy with the output, I rarely blamed the model. More often, I realised I hadn't given it enough to work with. Not because I had written a poor prompt, but because I hadn't yet understood the problem well enough to explain it.
I experienced this while working on a personal project exploring how AI could help people make better decisions in mountaineering.
At first, I asked AI questions that seemed reasonable. What features should a mountaineering product have? How do other products approach this problem? What should I build?
The answers weren't wrong, but they felt familiar. Route planning. Weather forecasts. GPS tracking. Equipment checklists. Features that almost every outdoor app already offers.
Then I realised I was asking AI to solve a problem that I hadn't fully defined myself.
Instead of asking AI to research the market, I started helping it understand the domain. I shared documentation from alpine organisations, accident reports, competitor products, my product vision and the type of decisions I wanted to help climbers make.
The conversation changed.
Instead of discussing features, we started exploring why climbers continue despite warning signs, how accident reports reveal recurring patterns in decision-making, where human judgement tends to fail, and how a product could help someone recognise risk before committing to a climb rather than simply helping them navigate once they were already on the mountain.
That's when something clicked.
I realised I wasn't giving AI better prompts. I was giving it better understanding.
Looking back, I realised I'd seen this pattern before.
My mentors never expected me to have the right answers from the start. Before they challenged my solutions, they helped me understand the world I was stepping into. They gave me context. They asked better questions. They encouraged me to slow down before moving forward.
Working with AI has brought me back to those lessons.
It has reminded me that understanding a problem is often harder than solving it. And maybe that's why I've become less interested in finding the perfect prompt and more interested in understanding a problem well enough that I can explain it clearly, whether to another person or to an AI.
I'm still exploring this idea, and I'm sure my thinking will continue to evolve. But it's one observation I keep coming back to.
AI has undoubtedly changed the way I work.
It has changed how I research, how I prototype, how I write and even how I learn. It has made me faster in ways I couldn't have imagined a few years ago.
But beneath all of that, I've found myself relying on the same lessons my mentors taught me years ago.
Understanding before solving.
Asking better questions.
Taking the time to understand the problem before searching for an answer.
Maybe that's what surprised me most.
Technology changes the way we work.
But it doesn't change the value of understanding.