Salesforce lets customers pay for AI by revenue grown or costs cut

The Information reports that Salesforce is changing how it charges for AI, and that the shift is more complicated than it looks. Laura Bratton and Kevin McLaughlin published the piece on 30 August. It is behind a paywall, so what follows is a small number of facts from their reporting plus our own view of what it means.

Three things from the article, which is all this item takes from it:

1. Software companies selling AI are moving away from subscription fees, towards charging on how much customers use it and whether it helps their business.
2. Salesforce is starting to let businesses choose how they want to pay for its Agentforce product.
3. That includes negotiating custom contracts priced on how much the AI grows revenue by helping salespeople close more deals, or cuts costs by automating more customer service interactions.

The reporters' own framing is that Salesforce shows how complicated this transition may be. Read their piece for the detail.

Why this affects engineering

A subscription needs one number: how many seats. Usage pricing needs one more: how many calls. Outcome pricing needs a definition of the outcome, an agreed way to measure it, and a system both parties trust to produce the figure.

That last part is not a commercial problem. It is an instrumentation problem, and it becomes the job of whoever builds the system.

Take "helped a salesperson close a deal". Which deals count? A deal the agent drafted an email for, one where it found the contact, one that would have closed anyway? Somebody has to write that rule down, in code, in a way that produces the same answer when the customer runs it and when the vendor runs it.

Cost reduction is not easier. "Automated a customer service interaction" needs a definition of an interaction, a rule for one that gets escalated to a person halfway through, and a decision about a conversation the customer abandons.

The pattern to watch

The reporters place Salesforce inside a wider move away from subscription fees, towards charging on how much a customer uses the software and whether it helps their business. If your company sells software with AI in it, expect the question within a year, and expect it to arrive from a customer rather than from your own pricing team.

The thing worth building before then is measurement you would be willing to show the other side. Not a dashboard, a defensible count with the rules written down.

Both parties reading the same number and agreeing on what it means is the hard part, and it is engineering work.

Source

• How Salesforce Is Overhauling the Way It Charges for AI, Laura Bratton and Kevin McLaughlin, The Information, 30 August 2026, subscription required