Salesforce targets complex work with job-ready AI agents

Insider Spotlight

• Salesforce is positioning AI agents to handle complex, high-value work rather than simply answer questions or complete isolated tasks
• A new long-horizon runtime allows agents to pursue goals across days and weeks while retaining context and adapting plans
• Salesforce says specialized agents are already delivering results in customer service, sales, employee support and commerce
  
  
  

Salesforce is betting that the next phase of enterprise AI will hinge on whether agents can perform meaningful work across business functions, rather than simply respond to prompts. 

Its experience across thousands of agentic deployments suggests that AI needs to understand a job, operate across business processes and keep pursuing an objective as work becomes more complex.

Job-ready AI agents

Salesforce on Monday introduced a portfolio of job-ready Agentforce agents for sales, service, commerce, employee experience and back-office operations. Connected to Customer 360, they come with skills, actions and data models designed for specific jobs.

The portfolio includes Casey for customer service, Paige for IT and HR requests, Carter for shopping, Hunter for outbound sales, Marshall for supply-chain processes and Piper for inbound lead generation. Fin handles complex customer-experience workflows.

Durable execution 

Salesforce is targeting work that cannot be completed in a single conversation. Its new long-horizon Agentforce runtime enables agents to pursue objectives across days and weeks while incorporating new information and changing plans.

Memory preserves context across sessions, durable execution keeps plans running over time and dynamic steering lets agents adapt to user feedback. Hunter is the first agent using the runtime.

Customer results 

Salesforce said it has delivered 7 billion Agentic Work Units across Agentforce and Slack, including 3.2 billion in its second quarter.

The company also cited customer results: Engine's help agent fully resolves 50 percent of chat inquiries, Perk's Hunter builds 60 percent of its sales pipeline and Autism Queensland's Paige resolves 70 percent of administrative requests.

Anthropic, meanwhile, autonomously resolves 79 percent of conversations seen by Fin.

Participants in workflows

Salesforce is expanding Agentforce with Multi-Agent Orchestration, which coordinates specialized agents across roles and systems. AI Skills lets employees teach Agentforce Coworker how to perform tasks, while Agent Optimizer is designed to help companies test, analyze and improve agents.

The strategy reflects Salesforce's push to turn AI agents into persistent participants in business workflows, while keeping their actions within company rules, permissions and security controls. —Vanessa Hidalgo | Ed: Corrie S. Narisma