Nobody Asks About the Power Tools

I love YouTube channels of people who are really good at what they do. There is a strange satisfaction in watching a guy pressure-wash a sidewalk for 15 minutes or a lady groom a neglected dog for 45 minutes.

Once in a while, the algorithm offers me long videos of people cooking, trimming horse hooves, or restoring a rusty knife, and I enjoy those too.

But in real life, when I actually need one of those services, I don’t really care. I would rather not know how things were done or get involved in the process. At a few fancy restaurants, they will come and prepare the food beside your table. In a few traditional pizzerias, you can see the prep area and watch them work. But at the end of the day, I just want my pizza to be delicious.

I have paid people to pressure-wash my house before. I said hello when they arrived, checked the beautiful result, and paid when they left. I have no idea how they did it. The result was all I wanted.

Well, the other day I saw someone on social media asking if developers were telling their clients they were using AI to build their systems.

My gut reaction was: Who cares?

I do not mean that dismissively; it is an observation about where the question comes from. The person asking lives in the same world I do: an AI-saturated bubble where “should I admit I use AI?” feels like an existential question.

And most clients do not live in that world at all.

Here is what my bubble looks like.

Friends bring up agents in conversations. My feed is a wall of model releases, prompt tricks, and hot takes. Someone mentions a new workflow on Tuesday, and by Friday three people have tried it. Inside the bubble, AI is the water we swim in, and “built with AI” sounds like a confession.

Then I talk to people outside it, and the picture flips.

Smart people, good at their jobs. To them, AI is the thing that writes emails, the chatbot on the website they had to troubleshoot, or a fad that will pass. They never had a reason to look, and nobody ever showed them.

Inside the bubble, we overestimate how much everyone else knows. Everyone has seen what an agent can do to a codebase, right? They have not.

Outside the bubble, people underestimate what is already possible, because their last data point is a chatbot that hallucinated in 2023.

The bubble distorts in both directions, and that distortion is what produced the thread.

The person asking whether to disclose their AI use is peeking through the membrane and assuming clients see what the feed sees: the slop, the hype, the headlines about jobs evaporating. From inside, “built with AI” sounds like admitting something.

But the client is not in the feed. The client is in their business.

What a client actually buys is results: the report on time, the tool that works, the problem gone.

Nobody who hires a contractor asks whether the house was framed with a nail gun or hand-driven nails. They care whether the house stands. Theodore Levitt made the point decades ago: people do not want a quarter-inch drill; they want a quarter-inch hole.

Hand-driven nails take longer and cost more.

So does refusing to use AI.

That cost shows up somewhere: in the invoice, in the calendar, or in the corners that got cut to fit the deadline. The client pays for the tools either way, whether or not they can see them. The only thing they should never pay for is the ceremony of not using them.

From the outside, “built with AI” is a non-event, the same way “built with an IDE” is.

Nobody asks whether the accountant used a spreadsheet.

Now, the exceptions.

The tools become the client’s business the moment they change what the client owns or carries. If I build something with AI, the client owns what was built, and code-quality standards do not move just because the typing got faster.

If the work handles sensitive data or lives in a regulated space, how it was produced can matter for compliance. If large parts of the deliverable were generated rather than written directly, provenance, licensing, or contractual requirements may matter too.

Those are real conversations, and having them is part of doing professional work. But none of them are “did you use AI?” questions. They are “what am I buying and what does it come with?” questions, the same ones a careful client asks about any dependency or subcontractor.

The tool only matters when it is part of the deliverable.

The membrane will not hold.

People who are oblivious to AI today will eventually use agents, the way everyone eventually used spreadsheets and search engines. The feed will move on to the next shiny thing. And when AI becomes invisible infrastructure, “should I disclose I use AI?” will sound as strange as “should I disclose I use a database?”

Until then, the gap is an opportunity.

People who know what AI actually enables can build in an afternoon what used to be a vendor project. They have a real advantage, precisely because most of the market cannot price that capability yet.

The hard part is still knowing what to build, and I have written about that before.

But for the people who have it, AI is a lever. And the world on the other side of the membrane has no idea what that lever is attached to.

That advantage is on a timer, but it is very real while it lasts.

So use the power tools and sell the outcome. The client wants the house to stand. How you swing the hammer is your business.

If you have a problem that needs solving, I can probably help. I work with teams that want to use AI, automation, and good engineering practices to build things faster and keep them running without adding unnecessary complexity. If that sounds useful, take a look at what I offer and get in touch.