AI bills exploding? Where tokens leak and 5 habits that cut costs
Summary in three lines
- AI costs spike from how tools are used, not price. Sending only what is needed can halve the cost of the same job.
- A news report showed that even a large firm reviewing its AI bill found ordinary staff habits, not engineers, burned the most tokens. Turning PDFs into images and back into text was a prime example.
- Tokens are what AI eats, and both input and output are billed. Paste text, send only the relevant part, set a per-run spending cap, reuse prompt templates and check the bill monthly.
What we learned
- Paste text instead of files and send one section instead of the whole document. Images cost several times more than the same content as text.
- Give every automation a spending cap so it stops and reports instead of quietly burning money overnight.
FAQ
Do I need to worry if I only pay a monthly subscription?
Yes. Instead of a big bill you hit usage limits sooner. Dropping whole files in for small questions wastes your allowance the same way.
Will switching to a cheaper AI fix it?
Not necessarily. With poor habits a cheaper model may need more retries and end up costing more. Fix the habits first, then pick tools.
This page is a summary BELLA AI wrote after reading the original. The full process and screenshots are in the original post.