Share with your CRO
Salesforce is betting that AI consumption, not seat expansion, becomes the next growth engine for its $46 billion business. Q2 FY 2027 results showed $11.35 billion in revenue, up 11% year-on-year, with AI and data ARR approaching $4 billion and Agentforce on pace for roughly $1.5 billion this year. CFO Robin Washington called NNAOV growth (net new annual order value, the forward bookings signal) the strongest in four years. Triple-digit AI bookings growth and a refill rate of 50% on Flex Credits suggest the consumption model is holding in production, not just in pilots.
What this means for your business
The refill behavior is the number that actually matters here. When half of Agentforce bookings come from customers replenishing credits after burning through them, that’s not a land-and-expand story, it’s a usage-compounds story. Revenue organizations already running Salesforce CRM are the ones closest to this inflection, because Agentforce agents resolving customer cases and updating pipeline records sit directly inside workflows those teams already own. If you’re in that installed base, the question isn’t whether to evaluate Agentforce; it’s whether your current contract structure captures or misses the consumption upside.
The Anthropic partnership and Claudeforce launch address a specific threat that enterprise software incumbents have mostly talked around rather than acted on. AI-native tools have been quietly pulling sellers out of CRM workflows by becoming the surface where research, drafting, and deal synthesis actually happen. Placing Salesforce data and context inside Claude’s interface inverts that dynamic, keeping Salesforce as the system of record while meeting users in the AI tool they’re already opening. Whether that holds depends entirely on data quality and governance inside the Salesforce environment, because a Claudeforce that surfaces stale or incomplete CRM data loses the argument fast.
Pricing flexibility is where Salesforce is making a structural bet that most enterprise buyers haven’t fully priced in yet. Outcome-based and consumption-based pricing for AI agents sounds like a procurement convenience, but it’s actually a margin negotiation you’ll run repeatedly as agent volume scales. The customers who define “agentic work units” (discrete tasks an AI agent completes in production, like resolving a support case or triggering a workflow) clearly in contract terms now will have more leverage when refill rates compound. I’d revise this read if Flex Credit refill rates stall in the back half of FY 2027, which would signal that early production deployments aren’t surviving contact with real enterprise data complexity.
Based on reporting from Salesforce Q2 FY 2027: Can Agentforce Drive Revenue Reacceleration?, originally published 2026-08-27 14:39:00.
