I have been building websites since 1996, when I put up my first page on Angelfire. In the last few years my teams have expanded what they can build with AI, mainly in web development.
I wrote recently about what AI will change about building companies. This piece is narrower. It is about how we actually build with it, where it falls short and the things I keep away from it entirely.
What AI Cannot Do for Us
AI cannot really do creative or design. It can produce something that looks like a design, and it can produce a lot of it quickly. What it cannot do is decide what a brand should feel like, what a page should make someone feel or which idea is worth building at all.
So we do not start with AI. We start with the right creative: the design, the look, the direction that a person made. That becomes the base. When you feed AI the right creative as its starting point, the work it does on top of it gets much better. When you start with nothing and ask it to invent, you get something generic.
A Team of 1,000 Developers
Once the creative is right, the picture changes. If you keep front-end and back-end developers involved, it is like having a team of 1,000 developers following your command.
The developers are the part people leave out. They know what the code should do, they can read what the AI produces and they can tell when it is wrong. They direct it, check it and connect the pieces. The AI does the volume. Without them, you have a lot of code nobody fully understands. With them, you have speed and someone accountable for it.
What used to take years of building now takes months. That is the real change I see. Projects that would have waited for budget or hiring can start now, and they move in stages instead of all at once.
It Does Not Scale the Business for You
AI does not scale a business. People sometimes talk as if it will, and that has not been my experience.
What it does is help at each stage. It helps you develop the idea and plan how to execute it. It helps maintain what you have built. It helps you grow in stages, adding the next piece when the last one works. The decisions about what to build, who to serve and when to push are still yours.
Where it changes the math is staff. In my view, AI can save about 90% of what a startup would otherwise spend on staff, through the startup, midway and growth phases. That is my own estimate from what I have seen in my companies, not a study. It does not mean you need no one. It means a small, skilled team can do what used to take a much larger one, as long as that team includes the people who can direct and check the work.
What I Never Use It For
There are things I do not do with AI at all.
I do not talk to it about personal relationships or personal situations. I do not use it as a therapist or a counselor. Those conversations belong with people who know me and who can be accountable for what they say.
I do not use it to discuss things that are illogical. It is a logic tool, and I use it that way.
And I do not use it to learn something new. I use it to accelerate things I already understand. The answers are not always accurate, and if I do not already know the subject, I cannot tell when it is wrong. When I know the work, I can see the mistake and correct it in seconds. When I do not, the mistake looks exactly like the right answer.
That is why 30 years of building still matters to me. I can use AI the way I do because I learned the basics first, by hand, long before any of these tools existed.
The Human Part
Put together, the human work is clear. People bring the creative, the judgment and the direction. Developers keep the code honest. Founders decide what gets built and in what order. AI does the rest, fast.
I have spent most of my career building online, from the first wrestling sites to Emblaze One, our agency for web development, e-commerce and digital marketing. AI is the biggest change to that work I have seen. It works best for the people who already know what good work looks like.