For months I rationed my AI agent.

Not because it wasn't capable — because every extra attempt cost tokens, and tokens cost money. "Try one more version" was a spending decision. So I made fewer attempts, and got exactly the results that fewer attempts produce.

Then my token situation changed (details in my first comment — keeping this post about the question). The rationing stopped. What's running now, every day, without me watching a meter:

  • A morning brief assembled overnight from sources I picked
  • Watchers that flag changes I used to check by hand
  • A weekly report I used to spend Sunday evening writing
  • Draft-fix-redraft loops on posts like this one, where I only judge the final version

None of these are impressive individually. Together they replaced maybe 6–8 hours a week of my own checking, assembling, and first-drafting.

So here's my actual question for this community:

If tokens cost you nothing — unlimited attempts, unlimited drafts, unlimited background runs — what is the first thing you would automate?

Not the sensible answer. The one you've been putting off because it felt too expensive to try.

I'll read every answer. The best ones, I'll write up in a follow-up: what it would actually take to set each one up.