Salesforce gives Agentforce seven named agents and a runtime that works for weeks unattended
Casey, Paige, Carter, Hunter, Marshall, Piper and Fin ship as pre-configured agents for specific jobs rather than a platform to build one, with Hunter piloting a "long-horizon" runtime that pursues a sales pipeline across sessions instead of a single chat.
Salesforce on September 11 rolled out seven named Agentforce agents, each scoped to one job rather than sold as a general-purpose platform: Casey (customer service, across voice, SMS, WhatsApp and web chat), Paige (IT and HR requests via Slack and internal portals), Carter (shopper discovery and in-chat checkout), Hunter (outbound sales pipeline work), Marshall (supply-chain and back-office orchestration), Piper (inbound lead qualification) and Fin (customer-experience workflows across channels). Six are generally available now; Hunter stays in pilot until November.
The pitch shifts from platform to product
ppc.land's coverage frames the shift precisely: Salesforce is moving from selling an agent-building platform to selling pre-configured agents for named jobs. Hunter is the one built to show off Salesforce's new long-horizon runtime, meant to let an agent pursue objectives across days and weeks rather than completing a single interaction, through three mechanisms -- memory that carries context and progress between sessions so a plan survives the end of a chat, durable execution that holds a plan while allowing course-correction, and dynamic steering that adapts to feedback mid-run.
The numbers Salesforce chose to publish
Salesforce's own release cites specific customer figures rather than aggregate claims: 50% of Engine's chat inquiries are fully resolved by its help agent Eva; 60% of Perk's sales pipeline is built by Hunter; 70% of Autism Queensland's administrative requests are resolved by Paige; 90% of Hibbett's core shopper journeys are handled by its AI agent, live in six weeks; Asana's website agent Piper drove 4x its prior conversation volume; and -- notably, since Anthropic is both an AI lab and a Salesforce customer -- 79% of the conversations Fin sees in Anthropic's own support queue are resolved autonomously. Salesforce also disclosed a platform-wide usage figure: 7 billion "Agentic Work Units" delivered across Agentforce and Slack, with 3.2 billion of those in Q2 alone. As ppc.land also cautions, a resolution rate like that "depends entirely on which contacts enter the funnel" -- a company can route easy tickets to the agent and hard ones to humans and still report a high number.
Named, pre-configured agents with published resolution rates are exactly the kind of claim Merit AC's own scoring exists to check independently -- a vendor's self-reported resolution percentage isn't the same as a customer's own usage and outcome data showing the same thing.