Compare Einstein Bots with Agentforce Service Agent

Compare Conversational Building Blocks

AI agents can understand the underlying intent of messages and generate contextually relevant responses out of the box. They engage in fluid conversations and use conversation context to make decisions.

To navigate conversations, Einstein bots use predefined dialogs, steps, and utterances, and they respond to customers with scripted messages. They’re built on intent models that can require extensive setup and training. To review how bots work, see What Makes Bots Tick?.

Instead of a pre-defined conversation flow, Agentforce Service agents navigate conversations with the help of a reasoning engine and an LLM. And while bots’ main building blocks are dialogs and steps, agents are built with natural language topics and actions.

• Topics: Agent topics define the range of capabilities that your agent can handle. A topic is a collection of actions that an agent can use to perform a job to be done. Topics also contain topic instructions that help the agent make decisions and use actions effectively. Adding topics to an agent helps it perform different jobs.
• Actions: Agent actions are the tools that the agent can use to perform the job. Similar to dialog action steps in Einstein Bots, actions can complete specific tasks by taking inputs and producing outputs. Each agent action is built on a reference action, such as a prompt template, flow, or invocable Apex class.
• Reasoning Engine: Agents use a reasoning engine to choose and launch topics and actions during a conversation. Similar to a bot’s intent model, the reasoning engine helps an agent interpret users’ interactions and utterances. However, unlike intent models, the reasoning engine calls an LLM trained on vast amounts of data.

Compare Availability and Billing