Turn on Einstein Classification Apps

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

• Implement Einstein classification apps.
• Adjust field prediction settings.
• Enable support agents to see and use Einstein's recommendations.

Einstein classification apps aren’t available in Trailhead playgrounds, but that’s OK. Keep reading to learn how an admin would set one up. 

Setting up a classification app is straightforward—it’s just the data preparation for building a learning model that can take some time. Since Ryan De Lyon has finished auditing closed-case data, Maria is ready to set up classification apps for Ursa Major Solar. Here are the main steps that she follows.

1. Review the data requirements, key concepts, and rollout tips. For Einstein Case Wrap-Up, make sure that you’ve set up Chat.
2. Turn on Einstein classification apps.
3. Configure a classification model. Choose an app type for the model, which case fields to predict, and closed cases for Einstein to learn from.
4. Build your model to let Einstein analyze your closed-case data.
5. Configure field prediction settings. Choose when Einstein recommends, selects, or saves field values. Optionally, for Einstein Case Classification, Einstein Case Routing can route updated cases to the right agent.
6. Give your agents access to the field recommendations component by assigning permissions and updating your console layouts.
7. Activate your model to start showing predictions to agents.
8. Maintain your model with the help of the prediction performance dashboard.

After you enable the classification apps, decide which closed cases Einstein should learn from, and choose fields for Einstein to predict. You can have as many as five predictive models for each app. In the Try Einstein version, you can have one model per app.

1. Click Get Started or New on the Einstein Classification setup page.
   
2. Select an app and enter a name for your model.
3. Click Next.
4. Decide the type of cases your model will focus on and which new cases get predictions. If you want, you can also narrow the scope of your model to a subset, or segment, of cases. Then, click Next.
   You can use segments to focus a model on a particular business unit. For example, use record-type-based segments to have one model predict field values on cases in your Enterprise division, and have another model cover cases in your Consumer division. Because Einstein learns from the words customers use to reach out to you, segments can provide helpful context for predictions.
   
   
5. Define criteria to identify example cases and Einstein will learn only from cases that meet your criteria. If you defined a segment, your example cases come from your segment. Then, click Next.
   Example cases help you point Einstein to cases whose completed fields and field values reflect your best practices. For help understanding when to use segments and example cases, check out Einstein Case Classification Key Concepts in the Salesforce Help.
   
   
6. Add the fields that you want to predict for your agents. Then, click Next.
   
7. Address any warning or error messages to make sure that you have enough data. If you have fewer than 400 closed cases in your segment or example case set or for a field in your model, gather more data or adjust your filters.
8. Click Finish and move on to building your model. Your new model appears on the Einstein Classification setup page.

After you configure your predictive model, its status changes to Ready to Build. Depending on the model’s scope, building may take a few hours or longer. The process runs in the cloud, so there’s no impact on performance for your agents.

1. On the Einstein Classification Setup page, select the model name.
2. Select the Setup tab.
   
   
3. The Fields to Predict section lists the field values Einstein will predict. To remove a field from the model, select Remove from the Action menu. To add fields, select Edit under Configure Data.
4. Click Build to build the model.

Einstein starts analyzing your closed cases and building the model for your selected fields. The model owner receives a notification when Einstein completes learning for each predicted field. 

To add more fields later, edit the model, add fields, and then rebuild. After the model finishes building, you can customize each field’s prediction settings and then activate the model to start showing recommendations in the Service Console.

After your Einstein classification model is built, it’s time to decide your level of prediction automation. With the lowest level of automation, Einstein recommends the top three field values for each field in your model. Or you can have Einstein select and save the best value automatically. 

You can update a field’s prediction settings at any time, whether or not the model is active. 

1. On the model’s Setup tab, select Edit under Configure Predictions and select a field. Alternatively, select Edit next to a field in the list.
   
2. To show the field with Einstein’s predicted best value already selected, turn on Select Best Value. Agents see a BEST label next to the value and must confirm then save it. Drag the slider to choose a prediction confidence threshold, which is your minimum required confidence level for selecting the best value. A prediction’s confidence level represents the likelihood that the recommendation for the field value is correct.
   
3. If you have the paid version of Einstein Case Classification, Einstein can update and save the field value without agent review. Turn on Automate Value, then drag the slider to choose a prediction confidence threshold for auto-updating the field. Because this update happens automatically, the Automate Value threshold must be higher than the Select Best Value threshold. For Einstein Case Wrap-Up, you can set the Select Best Value threshold only. Automate Value isn’t an option.
4. Select Save & Close. Your changes take effect immediately, and the prediction settings appear in the field list.
   

The Performance dashboard shows Maria how well Einstein’s predictions are working, and helps her decide which field values to automate.

From the model’s Overview tab, select a field and choose a date range to display performance information for closed cases.

The Top 3 Recommendations chart indicates how often one of the top three recommended values matches the final field value at the time the case is closed. The Top Recommendation chart shows how often the top-recommended value matches the final field value when the case is closed. When a case with field predictions is closed, the dashboard refreshes.

Now that Ursa Major Solar has implemented Einstein classification apps, support agents have more time to deliver exceptional customer experiences. Case data quality is sure to improve because of less human error determining field values. Sita and her team are excited to see where machine learning and AI can take their company next.

• Salesforce Help: Set Up Einstein Classification Apps
• Salesforce Help: Considerations for Choosing Fields to Predict
• Salesforce Help: Update Your Classification Model
• Salesforce Help: Use Einstein Classification Apps in the Console
• Salesforce Help: Set Up and Work with Service Console