AI is a restaurant business
When you ask ChatGPT or Claude a question, a lot happens behind the counter. Answering costs the company electricity, compute and the chips it had to buy. It works like a restaurant: you pay for the meal, and behind the scenes someone pays the cook and buys the ingredients for every plate.
So more customers does not mean free money for an AI company. More customers means more cost. It may be the only restaurant that loses money as it gets busier, and the way out for these companies is more usage and more subscriptions, to pay back the investors who funded them.
What that means for you
Expect the plans, the limits and the prices to keep moving. The model you build on today will not be the same product in a year, and you will not get a vote when it changes.
A monthly plan and a second brain is not a system
A lot of people are closing clients by setting them up on one AI company and calling it done: pay for the plan, here is your second brain, it handles the rest. It works until the provider goes down, changes its limits or raises its prices. A business's back office should not depend on one company it has no say over. We made a second video on exactly this.
What to build instead
Start from how the business actually runs, its operations first, and design the system around that. Treat the model as one part you can replace. Keep the data, the rules and the prompts in the business's own system, so moving to another model is a change of settings and not a rebuild.
When these companies change, you will have to change too, and how much you are willing to change decides what you offer. That is how an engineer would approach it, and it is how we build or rescue an app.