Salesforce used its Dreamforce conference this week to detail how it plans to monetize an AI-agent-driven version of its platform, moving away from the seat-based subscription model that has defined enterprise software pricing for years.
Patrick Stokes, president of applications and marketing, described a shift in how users interact with Salesforce: instead of clicking through menus in discrete apps to find and apply data, users now prompt an AI agent — accessible through Claude, ChatGPT, Slack or Salesforce's own Agentforce Coworker — which pulls from Salesforce's data, business logic, security and permissions directly. This is the mechanism behind the company's Claudeforce partnership with Anthropic and its newly announced AIforce, which extends the same agent access to Slack and Agentforce Coworker.
CEO Marc Benioff said pricing preferences vary widely by customer: some want per-user pricing for predictability, others want per-agent, consumption, usage, transaction-outcome, or business-outcome pricing where Salesforce takes a cut of money saved or earned. Rather than standardizing, Salesforce says it has told its sales organization to offer maximum flexibility to write custom deals.
Analysts are split on what this means going forward. Barclays noted many customers are still comfortable with seat-based pricing because it offers cost visibility, while William Blair expects consumption or outcome-based pricing to become the dominant model. Salesforce left its financial guidance unchanged, including a fiscal 2030 revenue target of at least $63 billion.
For companies evaluating or renewing Salesforce contracts, the practical effect is that pricing is now negotiable and variable rather than fixed by seat count, with cost tied more directly to how much AI-agent work the platform performs.