‎Similar Cases Overview | Sprinklr Help Center

Similar Cases Overview

Updated

1 year ago

Within the third pane of the Agent Console and Care Console, agents are provided with AI-generated recommendations of similar resolved cases handled in the past. Depending upon the intent of the conversation, these similar resolved cases appear within the Similar Cases section of the Smart Assist tab.

These AI-generated suggestions of similar cases help agents quickly resolve customer queries by referring to a similar case resolved previously.

Use Cases of Similar Cases

Decrease agent training time by providing them with historically resolved cases.
Increase productivity of the customer care agent thereby reducing AHT.
Increase customer satisfaction and net promoter score with timely, accurate case resolutions.

Note:

This capability is available in both old and new third panes of Agent Console.
Enabling this capability requires the brand's guidance in terms of the ideal cases they want the indexing to be done for. This will help ensure that the predicted cases provide the best help to the agent working with a similar issue. Also, the brands should have a minimum of 25K cases around those suggestions/filters.
You can get multiple models enabled for different sets of cases within a single partner/workspace. For example, a single partner environment may have different teams divided by market/geography/product/workspace and would like to get different case recommendations based on their respective set of cases. Raise a support ticket at tickets@sprinklr.com to get this capability enabled.
To learn more about getting Similar Cases enabled in your environment, please work with your Success Manager.

Before You Begin

To view the recommended similar cases in Agent Console and Care Console, a user must have the View permission under Similar Case.

Similar Cases Permission

Prerequisites for Training an AI Model for Similar Cases

Similar Case AI models are trained on historical case conversation data, and consider only those cases which have higher chances of resolution steps present within them. To achieve this, we recommend cases that are in Resolved/Closed status and created within the last 6 months (preferably). Additionally, you can tell us which cases to pick for model training based on some selection criteria.

For example, the following fields and values can be used.

Brand has responded = True

Case Status = Closed

Time range = Last 1 year

These filters are then used to fetch the case conversations which would make up the index for the AI model to recommend cases from.

Note:

The index (capped at 100K) of cases keeps getting updated on a weekly basis to include all the newly resolved cases to further improve the model.