Get to Know Agentforce

Learning Objectives

After completing this unit, you’ll be able to:

Discuss how Agentforce (formerly Einstein Copilot) can benefit your business.
Explain how actions give an agent its capabilities.
Describe how Salesforce keeps your data secure when using Agentforce.

Before You Start

We recommend that you complete the The Einstein Trust Layer Trailhead module before starting this one. The work you do here builds on vocabulary and concepts you learn in this content.

Meet Your New AI Agent

Artificial intelligence (AI) has been a sea change in the world of business technology. It’s hard to find an application or an organization that isn’t transformed in some way by the accuracy, efficiency, and productivity gained by AI.

We’ve all interacted with business-focused AI assistants like chatbots to get information and complete basic or repetitive tasks. But, what if you could experience an AI assistant that was grounded in your structured and unstructured data? What if this assistant could respond to simple or complex business tasks using your company and customer data, communication history, Slack conversations, and more? What if this AI assistant was built into every Salesforce customer relationship management (CRM) app your business uses—and on mobile, too? Well, now you can have it all with Agentforce.

Agentforce takes the power of the Salesforce CRM and the convenience of an AI assistant, then adds the complex reasoning and natural language processing abilities of large language models (LLMs) like ChatGPT to understand and engage with you in a conversational way. Using our Agentforce platform, you can build powerful AI agents that assist your employees and customers with common business interactions.

What Is Agentforce?

Agentforce is a type of agent that helps your Salesforce users with business tasks. Agentforce is available on all of your Salesforce pages so you can seamlessly enhance your customer engagements and streamline the flow of work all across your organization. Agentforce uses your input, such as a prompt, and the trusted data that’s in your org to give accurate answers and execute the right actions.

An agent is semi-autonomous, meaning you type a request for a business task–like a question or an instruction–using natural language, and the agent takes it from there. The agent interprets your request, makes a plan to address it, and then calls the best action or actions to meet your needs. Based on what you ask, an agent might formulate sales or marketing guidance, auto-generate emails, summarize service work, visualize data, or write custom code. The more descriptive your request, the more detailed the response will be.

How Did Agentforce Get So Smart?

We mentioned that Agentforce picks the best action or actions to accomplish a task. Actions are how an agent gets things done. An agent includes a library of actions, which is basically a set of tasks the agent can do, such as summarizing information, getting answers from a knowledge base, or drafting emails. Every action in an agent is assigned to a topic.

Topics are a layer of organization and customization that help your agent make more accurate decisions and generate more relevant, predictable responses. If you peek inside a topic, you’ll find a collection of related actions and instructions telling Agentforce how to use them.

When a user enters a question or request, the agent selects a relevant topic and then launches one or more actions included within that topic. This keeps your agent focused on the actions and data that are most relevant to the current conversation. Another way to think of it: a topic is the job to be done (Manage an Account) and actions are the tasks related to that job (Generate an Email, Log Calls, Create a To-Do List). Together, they help the user accomplish their goal.

Agentforce comes out-of-the-box with many standard actions so you can become more efficient as soon as you open the Einstein panel. You can also use Agent Builder to create custom actions powered by features like flows, prompt templates, and Apex classes to address your unique business use cases that aren’t covered by standard actions.

Below are just a few examples of the many standard actions that are included in Agentforce. More actions will be included with each Salesforce release.

Action	Capability
Draft or Revise Email	Compose or refine persuasive sales emails.
Identify Record by Name	Map user input to the proper CRM entities.
Summarize Record	Summarize any CRM record.
Query Records	Search or query single and multiple CRM records using natural language.
Query Records with Aggregate	Get a total of records or record details based on query input.
Answer Questions with Knowledge	Summarize answers using data from a variety of knowledge articles.

Are Agents Just Fancy Bots?

You might be thinking, “This agent thing sounds a lot like a bot.” Or, “We already use Einstein bots, why do we need an agent?” The short answer is that bots are more predictable because they follow predefined conversational paths. Building rule-based dialogs and training a bot to understand user input can take a lot of time and expertise. Of course, there’s still room for Einstein bots to perform many types of static tasks across your Salesforce apps. But if you want to do more with flexibility and efficiency, your new AI assistant Agentforce is here to do just that.

Agentforce and Trust

Salesforce has long been known for its value of Trust, and its AI platform is no exception. Trust is baked into all of our AI products, tools, and assistants. Agentforce relies on multiple layers of security to safeguard your data from external entities, and to ensure that internally, data is only accessible to those who have permission to access it.

Agentforce Actions Respect Org and User Permission Settings

When a user enters a prompt into an agent and an action is called, that action will likely need some of your company or customer data to complete its steps. Agentforce actions recognize and respect the org and user permission settings of the requester and will only return relevant information that the user is already authorized to view. If the action launches a flow, it respects the users existing namespace access permissions. You don’t need to set additional permissions or take extra precautions to keep your org and data secure when using Agentforce.

The Einstein Trust Layer Protects Your Data

When you enter a request into Agentforce, it calls the right actions and accurately chooses data that is available in your org. Sensitive data that is detected by Einstein Trust Layer is masked to prevent it from being exposed to a large language model (LLM). The LLM, such as ChatGPT, uses the prompt plus the data it received to generate a relevant, accurate response.

That response, with data still masked, is sent back through the secure gateway of the Trust Layer for you to review. Then, the data is demasked so it becomes visible in the response, and an event log is created to track the process–from prompt to final response–so you have a record of everything that happened on its journey. A key feature of your prompt-to-response journey through the Einstein Trust Layer is the zero data retention policy that we have in place with our LLM partners. Your request, your data, and your response are immediately forgotten by the LLM so you don’t have to worry about your information showing up out in the wild.

To learn more about how the Einstein Trust Layer protects your company and customer data, visit The Einstein Trust Layer Trailhead module.

How Do I Get Agentforce for My Business?

Your Salesforce account executive has all of the information your business needs to start using Agentforce. They can guide you on licensing, explain what’s included with Agentforce, help you understand which actions would benefit your organization, and discuss Trust and data security at a deeper level. They can also keep you apprised of exciting release-related Salesforce AI features and updates.

In Summary

Agentforce for Salesforce is designed to be an engaging, conversational AI agent that helps employees complete tasks for a wide range of business use cases using Data Cloud and your grounded data. Getting started takes only a few minutes: It comes with preconfigured action types for common tasks, with more actions and enhancements coming with each release. Most importantly, Agentforce prioritizes the Salesforce core value of Trust in its design and functionality to keep your company and customer data safe. Explore the resources below to learn more about Agentforce and AI at Salesforce.

Looking for Help?

Discover More