Why Paying for AI Still Feels Like a Puzzle

Mark, who founded Atelier, spent part of Wednesday morning trying to give two AI companies his money. It took minutes to find where to pay at all. One service showed a balance due but no way to click through and pay it. The other insisted no card was on file — the same card it charges every month — and then displayed a raw payment-system error meant for engineers, not customers. Neither service's plan page said plainly what the plan actually included.

This is a lesson, not a pile-on. Paying should be the easiest thing a company lets you do, because it is the one moment the customer is already on your side. When a checkout hides its payment page, contradicts its own records, or shows an error code, it is not a software bug; it is a broken promise.

What paying should feel like: one click from "you owe" to "paid." Plain words about what a plan includes before you choose it. No error codes in front of a customer. Those are the three promises this paper's own pages will keep, according to Mark's pitch this morning.

The pattern is familiar outside AI: cable bills, airline fees, gym memberships. The AI industry, which asks us to trust it with everything from writing to code, should be better at the smallest transaction. A company that cannot accept payment cleanly is telling you something about how it handles the rest of its relationship with you.

The next time you hit a "pay" button and it works, notice that you didn't notice. That silence is the whole job. The question is why so many companies, including ones that sell intelligence, still make noise.

Sources: the brain (Atelier’s own record)