How I Use AI to Run My Businesses

I've written about how I run seven businesses, and AI is a big part of that. Here's how I got into it and what I've learned.

How did I get started?

It began before AI. During the pandemic there wasn't much going on, so I looked for a new skill and got into Bubble, a no-code platform for building apps. I used it for five or six years and built the Pattaya Rent a Car system and a dozen or so others. Along the way I picked up the basics of how software works: front ends, back ends, repositories, and how systems evolve.

Then I saw a post online about how easy it had become to build a website with AI, using Google AI Studio. Google seemed a sensible company to learn with, since they were already a big player, so I started there. I quickly hit its limits, switched to Claude, and I've never really looked back.

When did it click?

When I rebuilt the Pattaya Rent a Car booking system. We'd always tracked our bookings timeline in Google Sheets. The new system gives the team a live view of every car and booking, but it does much more than that, because everything is connected, from the website the customer sees to the back end. A lot of manual work the staff used to do simply isn't needed any more.

It was faster, cheaper and could do far more than the old version.

What's the catch?

The last 20%. AI gets the first 80% of a job done very quickly. But the last 20%, refining it, testing it and making sure it really works before staff or customers use it, normally takes three times longer than the first 80%.

Has AI ever gone wrong for me?

Yes. Early on, an API key, the password that lets software use an AI service, was left exposed. Someone got hold of it and ran through just short of ฿30,000 of my budget in minutes. Fair play to Google, they refunded all of it.

Now I never paste API keys into an AI chat. I add them myself or through the terminal, and I set spending caps, so if a key ever leaks, the damage is limited.

What am I building now?

No big projects at the moment. Instead, I look for problems in each business and see whether AI can solve them.

A recent example: some of our property agents needed a refresher on our procedures. So we added a training section to the internal portal where agents chat with an AI that plays the customer. At the end, they get graded on their answers.

My advice if you want to build your own systems

Make your knowledge portable. AI companies keep releasing better models, and you want to be able to switch to the best one. So keep your systems, workflows and instructions outside any one company's platform.

I keep everything in a GitHub repository. That way I can use Claude, Gemini or ChatGPT, and switch between them without much trouble.

If you've never used AI at all, start with the simpler advice in how I run seven businesses.