I Audited Every Dollar of My AI Spending (And What Each Subscription Actually Earns)

I put every AI subscription on one spreadsheet and, for each, answered a question no renewal page had ever asked me: what does this actually earn per month, in hours saved or output produced? The audit took an afternoon and rearranged my stack more than any product launch has.

The headline numbers: twelve subscriptions totaling $214 a month. Seven earned their keep with a clean multiple (time saved valued at even a conservative hourly rate). Three were borderline, kept on probations. Two were pure nostalgia, paying for tools I described to friends more often than I opened.

The calculation that mattered most was per-use cost. The $30 coding assistant amortized across 60 work hours was fifty cents an hour and obviously worth it. The $20 image tool amortized across two actual uses was ten dollars an image and obviously not. The same dollar figure can be either the best deal in the stack or the worst, and only the usage log knows which.

The audit's surprises were behavioral. The tool I praised most was mid-pack on value; the tool I never mentioned to anyone quietly earned the highest hourly rate in the stack. Subjective enthusiasm and earned value were barely correlated, which means my renewal decisions had been running on vibes for a year.

The system that came out of it: a one-line monthly log per tool (uses and outcomes), and a quarterly audit with the earn-rate column. Two subscriptions were downgraded, one was cancelled mid-audit, and the savings funded the one tool that had been earning the most but sitting on a cramped free tier. The stack got cheaper and better in the same afternoon, which is what audits are for.