Salesforce introduced a set of pre-built Agentforce agents on Friday, each packaged with the skills, actions and data models its job requires, and paired them with a runtime built to pursue goals across weeks rather than single interactions.
The company says it has delivered 7 billion agentic work units across Agentforce and Slack, including 3.2 billion in its second quarter alone. The agents connect to Customer 360, so they work with the customer data and business processes a company already runs.
Salesforce says six are generally available. Casey covers service over voice, SMS, WhatsApp and web chat. Paige fields IT and HR requests from Slack and employee portals. Carter guides shoppers through comparison and in-chat checkout. Marshall runs back-office supply chain processes and records an audit trail of every action. Piper chases inbound leads for sales and marketing. Fin takes the hardest customer-experience workflows, using an operations agent and models trained specifically for that domain. A seventh, the outbound seller Hunter, is in pilot, with general availability planned for November.
The long-horizon runtime rests on three pieces: memory that preserves progress across sessions, durable execution that lets a plan resume or change course, and dynamic steering that adapts behavior to a user’s feedback. Hunter is the first agent running on it.
Early numbers from named customers followed the launch. Engine’s help agent resolves half of its chat inquiries. Hunter builds 60 percent of Perk’s pipeline. Paige handles 70 percent of administrative requests at Autism Queensland, and Hibbett AI covers 90 percent of core shopper journeys. At Anthropic itself, Fin resolves 79 percent of the conversations it sees.
