When Building Becomes Abundant

There is a reason this way of working feels possible now in a way it did not ten years ago, and avoiding the subject because everybody is tired of hearing about AI would be dishonest.

AI changed the economics of creation. It lets a person with deep operational knowledge cross distances that once required a small multidisciplinary team. Analysis, prototyping, copy, interface work, code, testing support and documentation can all move faster. The important word there is support.

For all of that extra capability, AI has not suddenly handed us an unlimited supply of good judgement, and if anything I think it has made the judgement we already have more valuable.

When implementation was expensive, bad ideas often died because they could not get resources. That was frustrating, but it was also a crude filter. Now a bad idea can have a login screen by lunchtime.

The capability has been democratised, which is genuinely exciting, but the judgement needed to use it well has not magically arrived in the same box.

That is why domain expertise matters more, not less. A restaurant manager can tell you when a forecast is nonsense in a way a language model cannot. A mechanic can tell you which interruption actually ruins the day. A finance person can tell you which number will survive a board meeting. AI can help turn that understanding into something tangible at astonishing speed, but it cannot replace the years that made the understanding valuable.

There is another trap. Because we can build quickly, we can start thinking quickly. We hear a problem, ask an AI for an answer and mistake velocity for insight. We should do the opposite. Use AI to compress execution time, not thinking time.

That is the bit I keep coming back to, because the ability to build quickly is not permission to understand lazily.

AI is the how, not the headline. The interesting bit for me is that people who genuinely understand the work can now express some of that understanding in software without waiting for the old economics, a project team and three diary invitations to grant them permission.

I have lived through enough technology waves to be slightly suspicious whenever somebody tells me this one changes everything. Usually it changes some things, creates several new acronyms and leaves a surprising amount of ordinary work exactly where it was. That is one reason I have resisted putting AI at the front of this book. I do not want the thinking tied to whichever model happens to be fashionable when somebody picks it up.

What does feel different this time is the distance between understanding a problem and expressing a possible answer. I can sit with somebody who knows their operation, talk through the work, and have something tangible in front of them while the conversation is still fresh. Ten years ago that translation required more people, more hand-offs and more money. Every hand-off created delay and, more importantly, another opportunity for the original understanding to become diluted.

That is where I think experienced operators have an extraordinary opportunity. The person who has spent twenty years learning why Friday behaves differently from Tuesday has accumulated knowledge that was always valuable but difficult to turn into software. AI lowers that translation barrier. It does not give them the twenty years. It gives the twenty years a new way to speak.

The reverse is also true, and it is worth saying plainly. Someone with almost no understanding of an operation can now create something that looks extremely convincing. The buttons work. The copy sounds confident. The charts move. That polished appearance can disguise a complete absence of judgement underneath. We are going to need better noses for that.

So I am excited by AI, enormously so. I use it constantly. But I am much more interested in what it does to the value of experience. When execution becomes cheaper, knowing what deserves to be executed becomes more valuable. That is a far more human conclusion than the technology headlines tend to suggest.