Adobe Says Its AI Agents Bill Only When They Work. Nobody Knows What That Means.
Adobe says it only bills when its AI agents work. The new pricing model covers 1,770+ enterprise customers, but Adobe has not published what counts as worked, and customers have not seen the definition either.
Adobe says it only charges for AI agents when they actually work. Nobody has checked what that means.
That is the gap at the center of Adobe's new enterprise pricing model, announced at its Summit conference last week. The company moved its Experience Cloud product to a credit-based and value-based system and said it would bill customers only when an AI agent completed a task successfully. The marketing is clear. The contract is not.
Adobe has not published what triggers a successful completion, how it defines a finished task, or how it separates an agent that produced a result from one that merely ran. Enterprise customers in the 1,770-plus already using the credit-based model have not seen a service-level agreement that defines those terms, according to people with knowledge of the contracts. Adobe declined to provide one for this article.
The pricing shift matters because outcome-based billing is only as honest as its definition of an outcome. If "worked" means an API call logged in a system, Adobe collects regardless of whether the task actually finished. If "worked" means the task produced the intended result, Adobe needs a way to measure that result, and the company has not described one publicly. The difference could be significant for the 1,770 enterprises handing over money under the new model.
Adobe is navigating this moment without a permanent chief executive. Shantanu Narayen departed in March, and the company has not named a successor. Leading a pricing model transition that requires precise contractual definitions is a different kind of challenge when the corner office is empty.
The company is also under pressure from a stock price that has fallen roughly 30 percent year to date, and from AI-native competitors including Anthropic, whose Claude Design product competes directly with Adobe's creative AI tools. The outcome-based pricing announcement is partly a signal to enterprise buyers that Adobe can compete on results rather than seat counts. Whether that signal matches the fine print is what customers are trying to find out.
Adobe said the new model reflects value delivered to customers and described it as aligned with enterprise interests. The company did not answer questions about whether the billing triggers are defined in the contracts, what mechanism verifies task completion, or whether an enterprise customer can dispute a charge on the grounds that an agent did not actually work.
Two questions will determine whether the marketing claim holds or collapses: whether Adobe publishes the definition of a successful task completion, and whether any enterprise customer has formally tested the billing dispute process. If the definition never arrives, the announcement will have done the work that contracts were supposed to do.
