Einstein Classification Prediction Confidence

Determine Your Level of Automation

When your model is active, Einstein predicts the value of every field that’s selected in your model. You can control how much prediction automation you’re comfortable with.

For each field in your model, customize the level at which Einstein selects or saves field values. For Einstein Case Classification, three levels of prediction automation are available. For Einstein Case Wrap-Up, two levels are available.

• Recommend Top Values (less automated): Available for Case Classification and Case Wrap-Up. By default, Einstein recommends the top three field values. The service rep selects and saves a value, so a confidence threshold isn’t needed.
• Select Best Value (more automated): Available for Case Classification and Case Wrap-Up. When a prediction meets your confidence threshold, Einstein shows the field with the best value selected. The service rep confirms and saves the value.
• Automate Value (fully automated): Available for Case Classification only. When a prediction meets the confidence threshold, Einstein saves the best value for the field on the incoming case. No service rep review is needed. You can have Einstein Case Routing run your case assignment rules on auto-updated cases.

Learn About Confidence Thresholds

Prediction confidence measures the likelihood that Einstein’s best prediction for the field value is correct. The higher the confidence level of a prediction, the more likely that it’s correct.

Einstein classification apps measure prediction confidence as a percentage. For each field in your model, you set a prediction confidence threshold from 0% to 100% for the Select Best Value or Automate Value options. The threshold represents the minimum required confidence level for Einstein to select or save a predicted field value. Because Einstein Case Classification can save a prediction automatically, the confidence threshold for a field’s Automate Value must be higher than its Select Best Value threshold.

Confidence thresholds create a clear, stepped path for Einstein to follow when predicting a field’s value. For example, for Einstein Case Classification, if the prediction confidence level is-

• Above the field’s Automate Value threshold, Einstein updates and saves the field.
• Under the field’s Automate Value threshold but above the Select Best Value threshold, Einstein selects the best value.
• Under both thresholds, Einstein recommends the top three values but doesn’t select or save one.

In other words, the more automated the action, the higher the prediction confidence needed. If a prediction doesn’t meet one option’s threshold, Einstein falls back to the next, less automated option. For Case Wrap-Up, if the prediction’s confidence level is above the Select Best Value threshold, Einstein selects the best value. If the level is below that threshold, Einstein instead recommends the top three values.

Choose Your Confidence Threshold

Your automation sweet spot depends on your business processes and comfort with risk. A lower confidence threshold means more predictions but potentially less accuracy. In general, set a relatively low threshold for Select Best Value because the service rep can review and correct the value if needed. Because errors can be more costly in full automation in Einstein Case Classification, start with a high threshold for Automate Value.

To test automating field predictions with Einstein, start with the default top-three recommendations where the service rep must select and save the field value. After you’ve monitored the impact, try turning on Select Best Value, Automate Value, or both for the field. You can update prediction settings at any time without rebuilding your model.