The AI Bill Audit Kit

Why this kit, why now

Two things happened this month that most teams haven’t acted on yet:

1. Frontier prices collapsed. OpenAI cut its flagship API pricing by more than 20% — and that came on top of an 80% cut to its smaller model weeks earlier. Google’s latest workhorse model is running at half-price introductory rates through the end of the year.
2. Open weights caught up. Independent benchmarks now show open-source checkpoints matching closed models on hard reasoning tasks — at roughly one-fifth the cost when self-hosted.

If your AI bill was built on last year’s price sheet, you’re paying a spread that somebody in your P&L should be capturing.

What you’ll do with it

The audit takes one focused afternoon:

1. Map — enter your actual usage into the spend map. Most teams find 2-4 services doing overlapping work.
2. Benchmark — run your real prompts through the benchmark suite to see what your tasks actually cost on today’s models.
3. Decide — the migration decision tree sorts what to switch, what to keep, and what to renegotiate.
4. Ask — the vendor negotiation script gives you the exact language to take to your current providers before you switch anything.
5. Report — the one-page summary is written for a board or leadership meeting, not a Slack thread.

The honest fine print

The audit only pays if you act on it — expect a 20-60% reduction on modeled spend for teams still on legacy pricing. Teams already optimized will find less, and the kit says so up front.