What is a committed minimum for, if not adoption?

It is the floor under the vendor's forecast. In an outcome contract I want five terms in writing: the unit definition, a dispute window, an arbitration path, a committed minimum, and a price ceiling. The minimum protects the seller's revenue plan. It was never evidence that the buyer received value.

What is active utilization?

Whether the thing being paid for is in use, tracked as its own dimension. My migration tracker classified accounts by whether they had migrated and were paying, and had no separate flag for low utilization despite migration. That missing flag is why 12 accounts read as on plan.

Who builds the outcomes line if the vendor bills on tokens?

Product. When the vendor prices on outcomes, the count comes off the invoice. When the vendor prices on tokens or seats, someone has to define the outcome, count it, and divide the spend by it, and that definition is a product decision.

Is the OpenAI Marketplace bad for buyers?

I do not think so. A buyer with a commitment would rather apply it to an agent that closes loans than leave it idle, and Sierra's outcome pricing means the invoice carries a count. The caution is about reporting: do not let a rising drawdown percent stand in for that count.

Is Commit Drawdown a Good Measure of AI Adoption

Is commit drawdown a good measure of AI adoption?

THE SHORT ANSWER

No. Drawdown measures how much of a spend commitment has been consumed, and says nothing about what it bought or whether anyone uses it. In my own per-seat sunset, 12 accounts were classified as on plan because they were paying their committed minimums; their real usage was below the minimum, nobody escalated, and two were lost after the legacy tier was removed. The OpenAI Marketplace, where Sierra says eligible customers can apply part of an OpenAI commitment to building with Sierra, makes a commitment easier to consume and raises the percent without changing its meaning. Report three lines instead: committed and drawn, outcomes bought with the price of each, and active utilization flagged separately for anything paid for and not used.

Sierra joined the OpenAI Marketplace this week, which means an outcome-priced agent vendor can be paid from a model-vendor commitment, and Drawdown Is Not Adoption is my argument for retiring the percent that will be reported afterwards. The number behind it, 12 accounts that were on plan and not using what they paid for, is Break 1 in Field Report: What Broke When We Killed Our Per-Seat Tier. Why Sierra's pricing deserves the benefit of the doubt is in The Sierra Playbook: What They Actually Do Differently.

It belongs to the AI Business Models argument: the pricing model has to move to the outcome, and a budget measured in drawdown has not moved yet.

Last reviewed 2026-09-30 · 1 min read

Is commit drawdown a good measure of AI adoption?

SOURCES

THE LONG VERSION

RELATED ANSWERS