Salesforce has made clear it wants to move beyond charging for software seats and instead bill customers for the outcomes its AI agents deliver. The problem, according to a report from The Register, is that the company has not yet figured out how to make that work in practice. Its existing pricing models already pull in multiple directions, and none of them cleanly maps to the value an AI agent might create.
Customers today can buy access through traditional per-seat licenses, which become awkward when an AI agent does the work of many humans. They can also use Flex Credits, a consumption-based model that tracks usage but can be hard to predict, or sign all-you-can-eat contracts that offer certainty but can lead to overuse and strained infrastructure. Each approach has its own logic, but together they create a fragmented landscape.
The Register describes the result as an 'anxiety-filled architecture,' where customers are unsure whether they are paying for the right thing or will be hit with unexpected costs. Salesforce's ambition to charge for outcomes is clear, but the company still needs to define what an outcome is, how to measure it, and how to package it into a contract that both sides trust. Until then, the billing model remains a work in progress. The source offers no alternative view; it simply documents the gap between Salesforce's stated direction and its current reality.