Salesforce's AI portfolio carries three commercial layers: features bundled into Einstein 1 user licences, features metered against Data Cloud credits, and Agentforce conversations metered on a separate per-conversation basis. The pricing surface is deliberately opaque; the negotiation surface is correspondingly large. This article maps every AI line item Salesforce sells in 2026 and identifies the clauses buyers should land before signing.
Layer one: features bundled into the Einstein 1 super-edition. Einstein 1 Sales and Einstein 1 Service ($500+/user/mo list) include Einstein generative AI features (call summaries, email drafting, opportunity insights) for the entitled user. Layer two: AI-driven activity in Data Cloud (Einstein Studio model builds, Predictions, Copilot for Data Cloud) consumes Data Cloud credits at a higher rate than non-AI activity. Layer three: Agentforce — autonomous and assistive agents — is sold separately on a per-conversation or per-action commitment basis.
The over-spend pattern is layered too: buyers upgrade to Einstein 1 for a small number of power users, then commit to Data Cloud credits sized to optimistic AI adoption, then add Agentforce on top. Each layer alone may be defensible; the aggregate is rarely justified by realised adoption inside year one.
The unbundling math frequently shows à la carte at Enterprise + AI add-on is cheaper.
Einstein 1 Sales is priced as a bundle of Sales Cloud Enterprise, Data Cloud entitlement, Slack, Einstein AI features, and additional automation. For an enterprise where AI adoption is uneven across the seat base, the bundle is over-purchasing. A defensible alternative: Enterprise edition for the base seat, plus targeted Einstein AI add-ons for the power-user subset, plus a right-sized Data Cloud commitment. In our engagements, this unbundling typically lands at 25–45% lower TCO than blanket Einstein 1 licensing.
Salesforce will resist the unbundling on the basis that Einstein 1 provides a "complete platform". The buyer's case is the adoption telemetry: the proportion of seats that actually use the Einstein AI features the bundle provides. Converting that telemetry into a lower commitment is a matter of software contract negotiation discipline: you bring the adoption data, not Salesforce's platform narrative, to the renewal table.
Agentforce is metered on a per-conversation or per-action basis depending on the agent class. List pricing is volatile; the published rates in 2026 have moved twice already. The contract clause to land is rate lock-in for the contracted term and conversion math: how many "conversations" a given workflow actually consumes. Buyers signing Agentforce commitments without understanding the conversion math frequently double their effective per-action cost in year two.
Includes the Salesforce AI commercial-clause library and the conversion-math template.
Salesforce's standard Einstein and Agentforce terms grant Salesforce rights to use aggregated, de-identified customer data for model improvement. For regulated industries and for buyers with sensitive data flowing through the AI pipeline (financial services, healthcare, government), the standard terms are insufficient. The clauses to land:
The default contract is rarely the contract you should sign.
We have structured Einstein and Agentforce commercials across regulated and unregulated industries.
Each issue breaks down one vendor's latest pricing or audit move — and the exact counter — so you walk into your next renewal already knowing the number.
