Enable Smart Responses
Updated
This document outlines the steps that must be followed by the IC & MS team to implement Smart Responses.
Raise a Support Ticket
To enable Smart Responses, raise a support ticket with the Partner ID and the target environment in which you want Smart Responses to be enabled.
Permissions
To be able to create, edit, and deploy the pipeline, you must have the View, Edit, and Delete permissions under AI+ Studio Agent Assist within AI+ Studio.
Editing the Rule and Pipeline
Once the support team confirms that the feature has been enabled, go to the Sprinklr Platform and verify the following:
Navigate to the pipeline deployments page for Smart Responses, and you will find 2 newly deployed pipelines:
Default Smart Reply Private Pipeline
Default Smart Reply Public Pipeline
Click the Options icon of the pipelines and click Edit Pipeline. You will be directed to the pre-configured default pipeline on the canvas.
Now, on the pipeline canvas, hover over the Prompt node and click Edit Element.
On the left side of the Prompt window, you will see Provider and, below it, Model.
Gemma model (selected by default): For all languages except Spanish, Bengali, Danish, Hebrew, Hungarian, Norwegian Bokmål, Dutch, Norwegian, Romanian, Swedish, and Urdu, use the Gemma model hosted under the Sprinklr In-house provider with the name agent-copilot-gemma-26B.
GPT-4.1 mini model: For Spanish, Bengali, Danish, Hebrew, Hungarian, Norwegian Bokmål, Dutch, Norwegian, Romanian, Swedish, and Urdu, use the GPT-4.1 mini model hosted under the Sprinklr Azure OpenAI provider with the name gpt-4.1-mini-2025-04-14/sprinklraiplus/intuition.
In the System Prompt, replace the placeholders <Brand_Name> and <Write about 2 lines of description about the brand>. You can use any LLM or the brand's website to obtain the description.
Example for the Niva Bupa partner:
"You are a customer care agent working for Niva Bupa. Niva Bupa is a leading health insurance provider in India, offering comprehensive and customer-centric medical insurance solutions. It focuses on hassle-free claims, a wide hospital network, and innovative digital services to make healthcare access simple, fast, and reliable for individuals and families."
After updating the Model (if required) and the System Prompt, click Save in the top-right corner of the Prompt node. Then, click Save and Deploy, located at the bottom-right corner of the Canvas, to make the updates live.
Steps to Follow on Rule Engine
Find the inbound rule named Default Smart Reply AI+ Studio Prediction.
Open this rule and edit it to meet the partner's requirements, as it contains pre-selected filters by default, such as the language set to English and the Channel/Message Types configured for their respective pipelines. You can also add the specific accounts where you want Smart Responses to populate.
After adjusting all the filters, enable the rule. You can then verify that Smart Responses are populating on a live case.
Multiple Deployments for different use cases
On the Smart Response Pipeline Deployments page, you can create multiple deployments to cater to different client use cases, such as providing a specific Smart Response tone for a particular channel or account type.