In Twenty years I wrote about why I left the sector I had worked in for two decades. There was a short section in it about AI, and I ended that section by saying I would come back to it in more detail later.

This is that.

The place AI had in it

It played a part. It did not decide.

I described it plainly at the time: a significant portion of the work volume that used to fill my days had shifted, and I could feel it. That was a real observation and it belongs in the picture. But the decision itself was about who I am, how I want to work, and what I need. AI was in the room. It was not holding the pen.

What I want to write about now is the other side of that, because it is the side that is almost never discussed by the people who are worried about it. AI did not only take something away from me. It gave me something I would not have been able to buy, hire or shortcut.

Twenty subjects at once

Starting in a new sector as one person means that everything is new at the same time.
Regulation. Survey practice. Photogrammetry. Thermography. Reporting standards. Inspection frameworks. Invoicing. Software. It is not one subject to master while the rest stays familiar. It is twenty subjects, all of them unfamiliar, and there is no colleague down the corridor who has already solved it.

One example, because it is the clearest.

A drone records its positions in WGS84. Belgian construction and survey work does not. There you work in Lambert 72 for position and in TAW for height. Converting between them is not a matter of subtracting a number. Height in particular does not behave the way you assume it does, because ellipsoidal height and the height a surveyor uses are two genuinely different things, related through a model of the earth's shape rather than through a constant.

Nobody comes to explain that to you. You either understand it or you deliver coordinates that are internally perfect and land in the wrong place. And a client who receives that has no way of knowing, which is precisely the problem.

The same applies to orthophotography and photogrammetry. It is easy to produce something that looks impressive. It is a different matter entirely to understand ground sample distance, overlap and control points well enough to know whether the result is something a professional can actually base a decision on. Looking good and being correct are not the same standard.

Then there is thermography, which is its own discipline entirely. To read a thermal image of a solar installation you first have to understand how a solar panel actually behaves – electrically, and under load, and in the sun – because otherwise you are looking at coloured patches and guessing. And you have to know your aircraft and its sensors through and through, the same way I once had to know my cameras through and through. Not the marketing specifications. The real behaviour, including where it stops being reliable.

That is a handful of subjects out of roughly forty fields of expertise.

What AI was, and why it is nothing on its own

It was not a machine that did the work for me. That is the misunderstanding.

What it gave me was the ability to keep questioning a subject until I genuinely understood it. To ask the same thing in five different ways, at three in the morning, without anyone becoming impatient with me. To be corrected without embarrassment. To go from a vague sense that something did not add up to knowing exactly why.

But here is the part I want to be very clear about, because it gets left out of most of these stories.

A powerful tool is worth nothing if you bring nothing to it.

You have to be technically grounded. You have to be genuinely interested. And above all you have to be able to judge whether the answer you are given is correct, because you will regularly be given answers that are not. That last point is decisive. Without the ability to check, you are not learning – you are collecting confident-sounding text.

Nobody goes from zero to a hundred, not with AI either. It remains a stepwise learning process, driven by your own motivation, your own focus, your own willpower, and a goal you are actually working towards. Take those away and the tool gives you nothing at all.

It worked for me for a specific reason: twenty years of technical work behind me, and a new field where a result is either measurably right or measurably wrong. I could verify. That is the whole difference.

Nothing done only once

The single most useful decision I made this year was about method rather than technology.

Every new request that landed on my desk for the first time, I treated as structural rather than as a one-off. Carry it out. Understand why it works the way it works. Write it down. And where it made sense, automate it. So that the second time is faster than the first, and the tenth requires almost nothing.

That is how the client platform came about, where deliverables live as something a client can return to rather than as a download link that expires. That is how quotations became consistent instead of being reinvented every time. That is how intake and estimation stopped being guesswork.

None of that was planned as a product. It was the accumulation of refusing to solve the same thing twice.

I always joke that I am lazy. It is not laziness. It is simply a refusal to do the same thing more than once.

False productivity

And this is where I have to be honest about something that took me a while to see.

Working this way has a failure mode, and I walked straight into it.

Tools get finished. They are visible, measurable and satisfying. You can look at what you built at the end of the day and see progress. Entering a new market is the opposite: slow, uncertain, mostly invisible, and impossible to tick off a list. So without noticing, you drift towards the thing that rewards you.

You end up in a kind of sealed-off productivity. Genuinely working hard, genuinely producing output, and that output may not be relevant – or at least not relevant yet. It feels like momentum because it is momentum. It is simply pointed somewhere other than where you needed to go.

The point was to start a business. Not to publish an endless series of tools and procedures, however good they are. The tools exist to help with that, to speed it up, to automate parts of it. They are not the objective. Somewhere along the way in a long process, that distinction gets blurry.

The correction is not to build less. It is to step back regularly and ask one question about whatever you are working on: does this bring a client closer, or does it only clear an item off my own list?

I still have to ask myself that. Regularly.

What it is actually about

Which brings me to the thing that matters most in all of this.

Being able to apply technology is not where the work ends. It is where it starts. And in my case I do want to be able to apply it thoroughly, all the way through, because half-understood technology produces confident mistakes.

But a client does not buy a drone flight. Sometimes the drone is simply the fastest, safest or cheapest way to solve the problem in front of him, and sometimes it is not the right instrument at all. The focus is the problem and its solution, not the aircraft.

That is a different business than the one I would have built if I had let the technology lead.

What it cost

I am not going to present this year as effortless.

I worked seven days a week for a year. Eight hours on the light days, frequently fourteen or fifteen. There were nights I got up at two in the morning with an idea and went to bed at midnight the following day with that idea finished, or close to it.

I am aware of how that reads next to what I wrote about two burn-outs. The difference, and it is the whole difference, is that this was mine. Nobody was waiting on it, nobody set the deadline, and I was not proving anything to anyone. That does not make it sustainable indefinitely, and I know that too.

The impact was immense, and it was not carried by me alone. Working that intensely for a year leaves marks on your private life, your social life and your family life. I am not going to dress that up or pretend it came for free. It is how I work, and the people closest to me absorbed a great deal of it. Fortunately there is room for that in my relationships. That is not something everybody has, and I do not treat it as self-evident.

There is also something worth noting in it. If AI made everything so much faster, why the fifteen-hour days?

Because AI accelerated the learning, not the work. The bottleneck moved. It used to sit at what I did not know. Now it sits at how much I can process, decide on and carry out. That is a better problem to have. It is still a limit.

What I have to show for it

What I got back for that is not a finished state, and I want to be careful not to claim one.

But I do stand in this subject matter with confidence now. I know what I am talking about, I know where the limits of my knowledge are, and I know how to close a gap when I find one. That is a different thing from knowing everything, and I have no interest in confusing the two. The curve was steep and it is not over. It should not be over. Continuing to learn and to evolve is not a burden in this work – it is the work.

That is exactly why I enjoy being in a technological sector. Nothing stands still long enough to become comfortable.

The moment I notice the difference most clearly is in conversation. Sitting down with peers or clients who have been in this business for more than twenty years, and going into the substance for two and a half or three hours without running out of depth. Not managing to keep up, but genuinely holding my end of it, in both directions.

Every time that happens I think the same thing. This worked.

Threat and opportunity

Both are true at once, and I am not interested in resolving that into a tidy conclusion.

Something was taken from the work I used to do. I felt it, I named it, and I am not going to pretend otherwise for the sake of sounding optimistic.

But as a one-person business, the opportunity won, and it was not close. The learning curve I got through this year, and the position I am standing in now, I could not have reached on my own. Or rather: I could have, but I estimate it would have taken five years instead of one.

I did not get those five years handed to me. I spent a year working the way I described above to earn them. The tool made it possible. It did not do it.

That distinction is the entire story.

And now the work

One more thing, because it is where all of this was heading.

A year of learning and building means a year of sitting behind a computer. Which had an effect I did not anticipate: it made the days I was actually out doing the job stand out far more sharply. Every one of them reminded me why I chose this in the first place. Being outside. Flying. Solving a problem. Producing a result somebody can use, and a client who is satisfied with it. That is what makes me happy, and a year at a desk only made that clearer.

That is also the point I have now reached. What needed to be built has been built. What could sensibly be automated has been automated. I am more than ready to shift the weight from developing tools to carrying out the work – and if I am honest, I am also simply finished with that phase.

AI will continue to support my processes and my learning. That does not change. But the focus moves now, towards doing the work itself in a structured, well-founded and knowledge-based way.

Which was the objective from the beginning.

And I am proud of where AirScout stands today. A year ago it was a decision. Now it is a business that works