AI Agent Builder "select type" screenshot

Agent Builder

Create and customize agents with an AI-powered, low code AI agent builder. Just write a description of the agent, and we help you build it. Want more? Pro-code tools can take your agents even further.

Define jobs to be done with topics.

Agent Builder gives you pre-built topics that agents can take advantage of that handle common prompts and use cases. These topics are automatically created by Agent Builder by understanding the metadata and semantic similarity of your data and knowledge articles.

AI Agent builder displays drafted topics like Event Scheduling and User Interaction and Preferences, allowing the user to select for editing or approval.

If you can describe it, Agentforce can do it. Using everyday language, you can simply describe how you want an agent to execute a task. You can tell your agent what actions to take, what actions not to take, what data it should have access to, what systems to work with, and when to escalate to a human.

Within AI Agent builder, the Event Scheduling topic name and its description, await user review and approval.

Every agent starts with a role that describes the goal or objective of an agent on your team. The role, or persona, is defined in order to provide clarity around the types of jobs your agent should be doing around the clock — whether that's an agent with the role of creating campaigns or troubleshooting support.

In the AI Agent builder interface, various fields allow an Agentforce user to define an agent's name, role, company, and permissions.
AI Agent builder displays drafted topics like Event Scheduling and User Interaction and Preferences, allowing the user to select for editing or approval.
Within AI Agent builder, the Event Scheduling topic name and its description, await user review and approval.
In the AI Agent builder interface, various fields allow an Agentforce user to define an agent's name, role, company, and permissions.

Create and assign intelligent actions.

Assign preconfigured prompt templates that agents can use to retrieve specific information, generate content, or summarize a response. These prompts can be prebuilt in Prompt Builder and configured in Agent Builder.

A preview within AI Agent builder showing a prompt template for an agent to generate an SMS message for a customer, alongside the agent's resulting response.

Agents can take advantage of prebuilt Flows to automate business processes. Agents can be configured to invoke Flows and use their output to help respond to the intial prompt. Agents can also be invoked from within a Flow to help automate even more work.

AI Agent builder displays a template-triggered prompt flow, outlining the steps: Get Experiences, Experiences Loop, and Add Experience to Prompt.

Agent Builder gives agents access to any business logic or automations built with Apex as an invocable action. Agents can trigger existing Apex calls to help automate processes and trigger custom logic.

In AI Agent builder, a list of action names and their descriptions are presented, allowing an Agentforce user to checkmark which ones to add to a topic.
A preview within AI Agent builder showing a prompt template for an agent to generate an SMS message for a customer, alongside the agent's resulting response.
AI Agent builder displays a template-triggered prompt flow, outlining the steps: Get Experiences, Experiences Loop, and Add Experience to Prompt.
In AI Agent builder, a list of action names and their descriptions are presented, allowing an Agentforce user to checkmark which ones to add to a topic.

Test and monitor agent responses.

Agent Builder lets you preview how your agents will respond to queries in the Plan Tracer. You can submit test prompts and see how your agent reasons through the prompt, assigns a topic, and then chooses the correct actions.

The AI Agent builder displays a preview of a user prompt, alongside the defined rules for how an agent should respond when a user asks about off-topic subjects.

Agents can be configured to handoff to humans seamlessly when the conversation veers to a certain topic or when the agent cannot sufficiently answer the prompt. Agents can escalate to humans for pricing inquires, severity escalations, off-topic questions, and more.

In AI Agent builder, a topic named 'Human Handoff' is used to define instructions and actions for when agents should transfer a case to a human.
The AI Agent builder displays a preview of a user prompt, alongside the defined rules for how an agent should respond when a user asks about off-topic subjects.
In AI Agent builder, a topic named 'Human Handoff' is used to define instructions and actions for when agents should transfer a case to a human.

AI Agent Builder FAQs

AI agents are developed through data collection, model training, natural language processing, and reinforcement learning. By leveraging these techniques, developers can create intelligent agents capable of understanding human language, making informed decisions, and taking actions to achieve specific goals. Using Salesforce's Agent Builder, building an agent is simple and low-code.

Agentforce can be applied across various business functions to improve efficiency and customer experience. From automating customer support to streamlining sales processes, generating marketing content, and maintaining e-commerce websites, Agentforce can be built and customized for any industry or use case.

Salesforce has pre-built Agentforce for Service, to help companies resolve customer cases 24/7, and Merchant Agent to optimize storefronts. Soon, there will be two Sales Agents, Sales Development Representative and Sales Coach agent to help optimize sales pipelines.

An AI agent builder is a tool that simplifies the creation and deployment of AI agents, like Salesforce's Agent Builder. It is a user-friendly, low-code tool to manage data sources, create jobs to be done and actions, and monitor agent performance.