Salesforce announced on September 11 a new portfolio of seven named, role-specific Agentforce6.5 out of 10 agents, alongside a “long-horizon runtime” designed to let an agent keep working toward a goal over days or weeks rather than resetting after a single session.
Seven named agents, one runtime
The new lineup — Casey, Paige, Carter, Hunter, Marshall, Piper and Fin — are pitched as “job-ready” agents built for specific functions rather than general-purpose assistants. Casey handles customer service issues across voice, SMS, WhatsApp and web chat; Paige fields IT and HR requests through Slack and internal portals; the remaining agents cover sales, commerce, and supply-chain tasks. Most are entering pilot now, with staggered general-availability dates running into November.
The headline technical piece is Hunter, an outbound sales agent that becomes the first to run on Salesforce’s new long-horizon runtime. Rather than executing a bounded task and stopping, Hunter is meant to prospect for leads, draft outreach, and carry a deal through multiple stages over an extended period, using memory, durable execution and what Salesforce calls “dynamic steering” to keep working across interruptions and follow-ups. Salesforce said more agents in the portfolio will move onto the long-horizon runtime over time, and that customers will eventually be able to build their own long-horizon agents on the platform. Hunter is currently in pilot, with general availability targeted for November 2026.
Salesforce also expanded Agentforce Coworker, the company’s in-app AI layer, with a feature called AI Skills that lets an employee teach the assistant how to complete a task once and then reuse that know-how across the workforce and across different interfaces. AI Skills is in pilot now, with general availability planned for October 2026.
Scale claims
Salesforce said it has delivered 7 billion “Agentic Work Units” — its internal metric for agent-completed tasks — across Agentforce and Slack to date, including 3.2 billion in the second quarter alone, a figure the company points to as evidence that agent adoption inside its customer base is accelerating rather than plateauing.
Why it matters
The announcement pushes Salesforce further into “long-horizon” agent work — multi-day, goal-directed execution rather than single-turn task completion — a frontier that OpenAI’s newly opened Agents API and Cognition’s Devin are also racing toward. Naming individual agents for specific job functions, rather than shipping one general assistant, is a distinct product bet: it tries to make enterprise buyers comfortable evaluating and rolling out agents the way they would evaluate hiring for a role, with Hunter’s pilot-to-GA timeline serving as an early test of whether long-horizon autonomy holds up in production sales workflows.