Agent

In Salesforce, the term Agent has two distinct meanings that coexist in modern orgs. (1) A service representative - a human user whose role is to handle customer inquiries through phone, chat, email, or messaging, typically working in the Service Console. (2) An Agentforce AI agent - an autonomous, LLM-powered worker defined in Agentforce Studio that performs tasks (resolve support questions, draft emails, update records, orchestrate workflows) using natural language and grounded CRM data. Both types are first-class on the platform; the context usually makes clear which is meant.

For most of Salesforce's history, the word agent meant one thing - a human service representative handling customer inquiries through phone, chat, email, or messaging in the Service Console. The Agentforce era added a second meaning. An Agentforce AI agent is an autonomous, LLM-powered worker defined in Agentforce Studio that performs tasks (resolve tickets, draft emails, update records, orchestrate workflows) using natural language and grounded CRM data. Both types are first-class on the platform now, and most modern conversations about agents apply to one or the other depending on context.

The reason this shared vocabulary is more consequential than it sounds is that the operating model for the two is genuinely different. Human agents need queues, presence states, ACW, training. AI agents need prompts, grounding rules, evaluation runs, monitoring. An admin who treats agent as one concept will end up making the wrong calls for both. Be explicit in documentation, distinguish the types in dashboards, and treat the architectural questions for each (capacity for humans, accuracy for AI) as separate concerns.