Salesforce Rolls Out New AI Agents for Sales and Support Work
Salesforce introduced AI agents for sales, retail, customer support, internal help desks and supply chain tasks, alongside an updated Agentforce Coworker. Most features are available now; the rest arrive by year's end.
Three of the agents are directed at sales work. Hunter helps business-to-business salespeople find leads and draft outreach emails, and it carries those functions through later stages of the deal-making workflow. Salesforce says the agent automates tasks such as preparing salespeople for presentations and creating proposals.
Some business-to-business deals take weeks to close, and Salesforce says Hunter can operate across such spans through a module called the long-horizon runtime. The module allows the agent to develop long-term work plans and reuse data across chat sessions. At launch the runtime is available only in Hunter; Salesforce plans to add it to more of the agents introduced on Sept. 11 at a later stage.
Piper and Carter can be embedded in company websites. Piper greets business-to-business prospects and directs them to the most relevant salesperson. Carter helps online retailers answer consumers' product questions and includes an in-chat checkout widget.
For customer support teams, Salesforce released Casey, which automates tasks such as processing product return requests, and Fin, which addresses similar support use cases. Fin is based on technology Salesforce obtained earlier this year through a $3.6 billion acquisition.
An agent named Paige is designed for internal help desks and can field employees' human resources questions and technical support requests. A further agent, Marshall, automates repetitive tasks for supply chain teams.
Companies can customize the agents with a programming syntax called Agent Script. Salesforce says the syntax lets developers require that an agent carry out important tasks in a specific way, which the company says reduces the risk of hallucinations.
The new agents arrive with an update to Agentforce Coworker, which lets workers interact with business data such as customer records and doubles as a tool for managing agents created by colleagues. A feature called Multi-Agent Orchestration lets it divide a complex task among several specialized subagents. Another addition continuously optimizes those subagents' output quality, and a capability called AI Skills is intended to make it easier for workers to customize how Agentforce Coworker carries out tasks.