Skip to main content

Build and Activate Your Reply Recommendations Model

Learning Objectives

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

  • Prepare data for Einstein Reply Recommendations.
  • Learn how to curate replies for agents.
  • Understand how to edit and activate your Einstein model.

Before You Begin

In this module, we assume you're a Salesforce admin with the proper permissions to enable Einstein Reply Recommendations. If you’re not, that’s OK. Read along to learn how your admin would take the steps in a production org. Don't try to follow these steps in your Trailhead Playground. Einstein Reply Recommendations isn’t available in the Trailhead Playground.

Build Your Model

To get started, Maria must enable Einstein Reply Recommendations. Then she’s ready to build a model that scans past chats and generates a list of commonly used replies. Here are the steps she follows.

  1. From Setup, in the Quick Find box, enter Einstein Reply Recommendation, and select Einstein Reply Recommendations.

  2. Click the toggle to enable Einstein Reply Recommendations.

    Enabled toggle for Einstein Reply Recommendations.

  3. On the Einstein Reply Recommendations Setup page, click Let's Go.

    Einstein Reply Recommendations set up flow chart: build model, generate replies, activate model, and monitor recommendations.

    If you don’t have enough closed chats to build a model, Einstein lets you know.

  4. To build your model and generate a list of up to 100 common replies, click Build Model.

    Build your model screen with Build Model button.

Building can take up to 48 hours, but it doesn't affect the app's performance for agents.

After you start a build, you can't pause or cancel it. You can leave the Setup page and come back to it later at any time to check the build status.

Keep your recommendations fresh with enhanced learning and automatic model retraining. The model that recommends replies to agents as a chat unfolds in the console and is automatically retrained each month. The updated model also learns from agents’ feedback so that they can get more accurate, useful replies that evolve alongside your service setup.

If you want to stop showing recommendations to agents for any reason, you can click Pause Recommendations on the Einstein Reply Recommendations Setup page.

When the model is built, the Einstein Reply Recommendations Setup page shows the replies that Einstein generated. Milestone complete!

A list of generated Reply Text with the option to activate the model.

Review and Publish Generated Replies

You’ve got all of Einstein’s generated replies in front of you, and now the reply power is in your hands. To ensure a smooth rollout, it’s a best practice to scan through these replies in this original format, also called reply text, and decide which replies you want Einstein to recommend to agents. As you review, keep an eye out for any changes you might need to make to the text. Your replies should:

  • Accurately represent your customer service policies.
  • Be up to date; for instance, they shouldn’t mention a product that’s no longer offered.
  • Not include sensitive customer material, such as a customer’s name, phone number, or account number.

After you’ve finished your review, it’s now time to publish--or convert--the replies to quick text so Einstein can recommend them. You need a quick text record to edit while Einstein keeps the reply text hard copy on the back end.

A generated quick text entry with the reply message: Is there anything else I can help you with today?

Quick text is a tool that lets you save snippets of reusable text for your company to use in all sorts of support scenarios. Combined with Einstein Reply Recommendations, it lets you:

  • Assign the reply a category, such as “greetings,” and make a folder for it.
  • Make the reply available not only in Einstein Reply Recommendations, but also as searchable quick text in the Chat and Messaging channels.
  • Update your reply text and settings as your business grows.

If you publish one reply at a time, you can edit the reply text during publishing. If you publish multiple replies at once, you can edit the reply text from the Quick Text page in Setup after publishing. After you publish and fine-tune all of the replies that you want to recommend to your team, it’s time to give agents access and activate your model.


Divide replies into different quick text folders to control which agents see which replies as recommendations. For example, sometimes Tier 1 and Tier 3 agents need different sets of replies because they follow different terms of service.

Activate Your Reply Recommendations Model

After Maria enables Einstein Reply Recommendations, she opens it up to her support team by giving them:

  • Einstein Reply Recommendations User permission set.
  • Access to the quick text folder containing the published replies.

She and her team of support agents can now see the Einstein Replies component on the Service Console’s Chat and Messaging tabs. However, the component doesn’t yet show any recommended replies during chats. Maria navigates to Setup and activates the reply recommendation model. 

  1. On the Einstein Reply Recommendations page in Setup, confirm that all replies you want to use list a status of Published to Quick Text.
  2. Click Activate.

    Set Up Einstein Reply Recommendations page showing confirmation of activation.

All Ursa Major Solar’s agents can now see reply recommendations when they’re chatting with customers. 


Keep learning for
Sign up for an account to continue.
What’s in it for you?
  • Get personalized recommendations for your career goals
  • Practice your skills with hands-on challenges and quizzes
  • Track and share your progress with employers
  • Connect to mentorship and career opportunities