Salesforce has spent much of the past year testing how to charge for the AI agents it sells through Agentforce, and it is now leaning toward the licensing approach that enterprise customers understand best: seats. After consideration of per-conversation and consumption-style pricing, Salesforce executives say customers have pushed for predictability, particularly as AI pilots move from experimentation into production workflows.
In recent remarks to investors and customers, CEO Marc Benioff has described the Agentic Enterprise License Agreement (AELA) as the preferred licensing path for large deployments. The idea is familiar to any IT buyer negotiating big software contracts: companies buy a package tied to user volumes, and get a menu of rights and flexibility inside a single agreement. But Salesforce’s seat-based posture comes with a caveat that is rapidly becoming the industry standard: “seat-based” does not necessarily mean unlimited.
A Pricing Approach in Flux
Analysts tracking enterprise AI monetization say many vendors now pair per-user licensing with an underlying meter, typically expressed as AI credits, units, or fair-use thresholds, to keep model costs from getting unexpectedly high. In other words, seats set the budget envelope, while embedded caps preserve the vendor’s ability to throttle heavy usage or charge for overages. Salesforce’s approach, which is in flux, follows that pattern: a seat-based construct designed to give procurement teams cost certainty, with an internal currency that can require top-ups if usage exceeds contracted limits.
Salesforce appears to be keeping multiple pricing paths open, because not every agentic workload behaves like a human user, and perhaps because the AI agent field is still so new. A sales assistant embedded in a rep’s workflow is naturally tied to a person, so per-seat pricing maps cleanly. But an automation agent that runs continuously (monitoring service tickets, updating records) looks more like a high-volume system process. For those cases, Salesforce has positioned utility models that charge for what the platform does rather than how many humans log in.
Bowing to Customer Demands
For enterprises, the headline is that the biggest deals are trending toward AELA-style agreements that look and feel like classic enterprise licensing: a negotiated, seat-anchored commitment that bundles rights to deploy Agentforce broadly, often alongside Data Cloud (now branded as Data 360). The pitch is this provides peace of mind: fewer surprise bills while teams learn how employees actually use agents at scale.
But buyers should read the fine print. Even in a seat-centric agreement, costs can rise through embedded usage limits, add-on credit purchases, and premium SKUs. Salesforce has been rolling out higher-tier Agentforce packaging, including premium editions aimed at broader, cross-cloud deployments, which can drive spend upward even when the model is per user.
The bottom line is that Salesforce is not choosing between seat-based and consumption-based pricing so much as blending them. Seats are becoming the primary commercial wrapper because enterprises want stable budgeting and a clear ROI baseline. Additionally, Salesforce is preserving metering mechanisms, via credits, actions, and fair-use controls, to ensure that the economics still work when agent activity ramps up.
