‘Unit Economics’ Means Something Different to Buyers and Sellers

← Back · ACT I · Chapter 1: The Illusion of Value

Sixty-two percent of organizations surveyed said an unexpected AI cost materially altered a business decision over the past year. Among those, 40 percent required board-level escalation, 33 percent implemented emergency spending freezes, and 25 percent delayed or canceled an AI initiative outright. The figures come from Mavvrik’s 2026 State of AI Cost Governance Report, produced in partnership with Benchmarkit and reported by CFO Dive on 11 August. Mavvrik sells AI cost management, which is worth saying plainly, because the number that should worry a CFO most is the one no tool resolves: only 11 percent of organizations can forecast AI spending within plus or minus 10 percent, down from 15 percent the year before. Visibility into AI cost has never been better, and forecast accuracy went backwards anyway.

Gartner’s answer, cited in the same coverage, is to stop tracking AI cost in isolation and build what it calls AI unit economics, linking token consumption to measurable outputs and outcomes so spend can be sized to the value it produces rather than minimized on its own terms. That is the right diagnosis. On its own it is also unenforceable, because Gartner’s own framing concedes that the connection between token consumption and business outcomes “often remains unclear.” The recommended fix and the described problem rest on the same undefined word. The industry has since started on the definition itself. The Linux Foundation’s Tokenomics Foundation, launched in August 2026, published a draft definition of AI tokenomics on 1 September that ends on realizing measurable business value. That definition was written where the meters are. It still leaves the buyer to say what the value was, which is the argument in The New AI Economics Standard Measures Labor Replaced.

The Term Spread Before The Meaning Did

“Unit economics” is now doing three different jobs in the same sentence, depending on who is speaking. A vendor’s invoice defines the unit as a token. A finance team building a chargeback model defines it as a workflow. A board asking whether the AI initiative paid for itself defines it as a business outcome. All three get called unit economics. None of the three can be compared to the other two, because none of them starts from the same cost basis, and a ratio built on a different denominator is a different number wearing the same name.

This is not a new failure mode. A vocabulary spreads through procurement and finance conversations faster than its definition stabilizes, and in the gap, whoever controls the measurement controls the negotiation. The definition usually belongs to whoever built the tool, which is the argument in Define Success Before the Vendor’s Product Defines It for You. What is new is that the term now sits in the invoice rather than in the product philosophy.

The same report gives the gap a number from inside the buyer’s own walls. Ninety-eight percent of engineering organizations use AI coding assistants, and only 42 percent include developer AI tool spending in their AI cost reporting. Before a vendor is anywhere near the room, the people generating the cost and the people accountable for it are not counting the same things.

The Buyer Who Waits For The Vendor’s Definition Argues On The Vendor’s Terms

A buyer that adopts whatever cost basis a vendor’s billing system defaults to, tokens, seats, capacity blocks, has not measured unit economics. It has adopted the vendor’s accounting and called the result a metric. That is the same move a total-cost-of-ownership model makes when it prices only what its author can see, which is the subject of Vendor Total Cost of Ownership Models Miss Your Biggest Costs.

Gartner’s recommendation to size spend to value only works if the buyer, not the vendor, defines what a unit of value is before the invoice arrives. That definition belongs in the negotiation rather than in a finance exercise run after the contract is signed, so every vendor answers against one measure instead of each supplying its own. Stated up front, it does on cost what nothing yet requires on outcome claims: it fixes the standard of proof before the numbers arrive, which is the argument in The Numbers Behind Big Purchases Are the Only Ones Nobody Audits.

The Escalations Are The Symptom

Boards escalating on AI cost overruns are reacting to volatility, and the volatility is downstream of the vocabulary. The market adopted a term for measuring AI value before it agreed on what the term measures, and every vendor invoice since has been free to supply its own definition. A board that wants the escalations to stop will not get there with a better dashboard. The lever is to require, in writing, before the next AI contract is signed, whose definition of the unit the invoice will be measured against.

That requirement costs nothing to state and it changes what a vendor has to answer. It also converts a recurring surprise into a term of the deal, which is where an unpredictable cost belongs.

A cost line without an agreed definition is not a metric. It is a negotiating position dressed as math.

Related reading: AI Token Pricing: The Seller Sets Both the Price and the Quantity

Reference: Alexei Alexis, “1 in 4 companies delay or cancel an AI project over cost,” CFO Dive, 11 August 2026, reporting Mavvrik’s 2026 State of AI Cost Governance Report, produced with Benchmarkit, and citing Gartner research on AI unit economics. Mavvrik is a vendor in the AI cost management category and the survey is its own; the figures are attributed to it in text for that reason.

Keep reading: ACT I · Chapter 1: The Illusion of Value