Einstein Case Classification

Every business revolves around its customers and their experiences. Providing customers with a positive experience is always at the top shelf of every business to-do list. In Salesforce, the Service cloud maintains the growth of the customer lifecycle with the help of Support Agents, Einstein for Service, and various other features.

Einstein for Service contains Einstein case Classification, which is a powerful feature of Salesforce intended to Support Agents. Based on Historical Data, Einstein Case Classification fills field values on new cases. This removes the guesswork and saves time for agents to work on customer relationship development.

Now the question arises ‘Why should we use Einstein Case Classification?’ and ‘How is data of cases prepared?’. Clearing these basics of Case Classification will help us to easily create and maintain our cases.

Advantages of Einstein Case Classification


  • 1) Searching for field values is very time-consuming. With the help of automatic case routing, support agents save some time and use that time more efficiently on more important tasks.
  • 2) The predictive model for routing cases is more accurate than humans.
  • 3) Cases can be sent to the right support agent and resolved quickly due to automatic classification.
  • 4) Customer satisfaction score improves as in free time support agents can focus on building strong customer relationships.

For example, your business deals with laptops and has 2 support agents, you receive a case with some software issues. One of your support agents is good in dealing with cases related to Hardware and the other is good in dealing with Software cases. Then based on previous cases closed by them Einstein Case Classification will directly assign that case to an agent with good software case-solving capability. This will save time and improved data quality which in turn ensures better customer service.

Note: Always set up case fields, case notifications, and case assignment rules before setting up Einstein Case Classification. Follow the four-step setup process for Service Cloud:

  • 1) Clear Case Management
  • 2) Channels
  • 3) Knowledge
  • 4) AI and Bots

Setting up case classification is the last step in setting up Service Cloud because if case management, channels, and knowledge aren't set up first, the time AI saves for agents will be spent figuring out other processes.

Use Cases For Einstein Case Classification

  • Growth in Inquiry Volume is one of the major problems across multiple channels. We can use Einstein Case Classification along with Einstein Case Routing to control costs while maintaining customer satisfaction.
  • Keeping service agents motivated, engaged, and retained with your company is of utmost importance. Solving simple cases repetitively leads to a decline in agents' satisfaction towards work. With the help of Einstein Case Classification Service Agent can burn through their backlog of repetitive inquiries and have more time to work on more challenging tasks.
  • A new service agent lacks experience in solving cases and a lot of time of your senior employees is wasted on training him. Since Einstein Case Classification already keeps historical data, a new agent can close cases with less help from senior employees.

Complete Step by Step Tutorial for Einstein Case Classification

Before training your data it is necessary to understand that our data should not reflect human subjectivity and underlying social biases. If biased data is learned by Einstein Case Classification, a characteristic, factor or group can be over or underrepresented. So, data should be of sufficient size and representative of real-world use. Hence, avoid using low priority cases as a data set.

Step: 1 Prepare closed data for case classification

Preparing data is an important and time-consuming step for building a learning model. Populate the case fields that you want to predict correctly. Though the more samples per field you have better will be the result but the minimum value cannot be below 100. If you want to predict a field value you should have at least 400 closed cases with value in those fields in the past 6 months. 10,000 closed cases are considered ideal to accurately predict values on case fields.

Step: 2 Enable Einstein Case Classification

From Setup >> Quick Find box >> ‘Einstein Case Classification’ >> Enable Einstein Case Classification


einstein-ease-image1

Figure 1: Enable Einstein Case Classification. [Source: Trailhead]

Note: To Enable Einstein Case Classification make sure you check for Setup of Einstein Case Classification only in Lightning Experience of Enterprise, Performance, and Unlimited Editions only.

Step: 3 Configure Predictive Model

For configuring data in the predictive model you need to answer 4 Questions of Einstein Case Classification-

  • Question 1: Want to base predictions on all recently closed cases?
  • Question 2: Want Einstein to learn from all Cases?
  • Question 3: Which fields do you want your agents to get recommendations for?
  • Question 4: Does everything look good?

Einstein Case Classification provides options to customize the predictive model with the help of simple Questions. The first question that we are asked is if we want to base predictions on all recently closed cases or not. Here we will define our segment rather than reviewing all closed cases of the past six months as shown:


einstein-ease-image2

Figure 2: Choose fields to filter by [Source: Trailhead]

Click ‘Next’, which leads to our second question.

Since we have done advance planning to identify our business criteria, we will select ‘No, learn from specific cases’ as an answer to our second question of Einstein Case Classification. Define your example as shown:


einstein-ease-image3

Figure 3: Define your example case set [Source: Trailhead]

Click ‘Next’, and move on to third question.

Search for fields you want your agent to get recommendations for and click ‘Next’.


einstein-ease-image4

Figure 4: Add fields for recommendations [Source: Trailhead]

The number of closed cases lies in four ranges which will predict a field’s value

High: >10,000 closed cases
Medium: 1000-10,000 closed cases
Low: 400-1000 closed cases
Insufficient: < 400 closed cases


einstein-ease-image5

Figure 5: Include fields in your predictive model [Source: Trailhead]

At last click on ‘Finish’ if you have sufficient closed cases.

Step: 4 Build a Predictive Model

After you configure the predictive model, a new model appears on the setup page. Click the model name on the list. Click the ‘Setup’ tab. For making any changes in fields that you want to predict use the ‘Edit’ and ‘Remove’ button. Click ‘Build’ to start analyzing your closed cases.


einstein-ease-image6

Figure 6: Build a predictive model [Source: Trailhead]

Step: 5 Review Confidence of Predictive Model

Einstein Case Classification gives us three options to customize Field Prediction Settings:

  • 1) Recommend the field value only
  • 2) Populate the field automatically and let agent save after review
  • 3) Automatically save the field value


einstein-ease-image7

Figure 7: Review confidence of predictive model. [Source: Trailhead]

Click ‘Review’ on the Setup tab to start customizing. Enable the ‘Select Best Value’ option, this allows the Support agent to save the case. Now ask yourself, ‘How much Confidence you have that the best recommendation for the field value is correct?’. Drag the answer on your Confidence level for Select Best Value Slider.


einstein-ease-image8

Figure 8: Select Best Value. [Source: Trailhead]

Now adjust the Confidence level slider for saving the best value for Priority to the case, after enabling ‘Automate Value’ Toggle. As soon as you click ‘Save’, prediction settings appear in the field list.


einstein-ease-image9

Figure 9: Select Automate Value. [Source: Trailhead]

Note: Keep the confidence level of Automate value higher as saving requires more confidence as a change.

Step: 6 Activate Einstein Case Classification

Before clicking on the ‘Activate’ button make sure you have given support agent access to Einstein Case Classification and check if they are able to view Einstein Case Classification recommendations.
To give support agents access to Einstein Case Classification, follow the given path:

Setup >> Quick Find Box >>Select ‘Permission Sets’ >>Select ‘Einstein Case Classification User’ >> Click ‘Manage Assignments’ and assign users to the permission set.

Steps to add Case Classification to the Service Console

Lightning App Builder >> Service Console >> Open any case record page >> Click ‘Edit Page’ >> Drag the ‘Einstein Field Recommendations’ component onto the page >> Select type as ‘Case Classification’ >> Save and Activate changes


einstein-ease-image10

Figure 10: Add Case Classification to the Service Console. [Source: Trailhead]

Now you can Activate Einstein Case Classification from the Model Details Page by clicking on the ‘Activate’ button.


einstein-ease-image11

Figure 11: Activate Predictive Model. [Source: Trailhead]

CONCLUSION

Einstein Case Classification uses Machine Learning on case subject and description to auto-populate case record fields. At present only picklist and checkbox values are recommended by machine learning in new cases. To enhance the power of Einstein Case Classification uses Einstein Prediction Builder, which assists Service Agents in providing a more personalized customer experience with custom prediction models. But always remember to set up AI at last because if case management, channels, and knowledge are not set up first, the time AI saves for agents will be spent figuring out other processes.

Some of the resources that can be referred to develop a better understanding of Einstein Case Classification are:

For any query on Einstein Case Classification, contact support@astreait.com