Introduction

Updated 

By utilizing Generative AI technology, the Automated Agent Training feature enables automated simulation of customer conversations initiated by supervisors/admins for agents, aiming to streamline and automate the agent training process.

To utilize this capability, you need to provide a detailed prompt to GPT, and it will act as a customer during the conversation with the agent. Unlike manual training sessions, this feature can help you trigger an AI-powered customer conversation in one click!

Enablement note:

To enable this feature, reach out to support at tickets@sprinklr.com and provide the prompt that describes the scenario the bot will follow during the conversation with the agent.

Input Required

Prompt

Example:

You are a customer engaging in a simulated chat conversation with a customer service representative. You must intelligently converse to resolve an issue. Follow this scenario: You noticed an error in your last billing cycle where you were charged $8.99 for the "Mobile and Internet" plan instead of your "Mobile Only" plan, which costs $4.99. You did not authorize this change and received no notification. You tried raising a support ticket on the website but received no response for a week.

Only provide personal data if requested: Name - John Doe, Email - john.doe@test.com, Order ID - #122022, Phone number - +01 5522819 3382.

Follow these guidelines:

  1. Keep your responses concise, ideally under 20 words.

  2. Explain your issue step by step, avoiding lengthy messages.

  3. If the agent asks for time to respond, be patient.

  4. Request a solution from the agent, or ask them to escalate if they cannot provide one.

  5. Do not provide false information; respond 'I don't know' if necessary.

  6. Thank the agent and close the conversation if they offer to send a refund form or escalate the issue.

A prompt should be created keeping the 'Scenario', 'Customer Information' and 'Success Criteria' as input.

As shown in the above example,

1. Scenario - Describes the customer's issue and gives the necessary background to have a conversation in a story format.

In the above example: You are a customer engaging in a simulated chat conversation with a customer service representative. You must intelligently converse to resolve an issue. Follow this scenario: You noticed an error in your last billing cycle where you were charged $8.99 for the "Mobile and Internet" plan instead of your "Mobile Only" plan, which costs $4.99. You did not authorize this change and received no notification. You tried raising a support ticket on the website but received no response for a week.

2. Customer Information - Describes the information the customer can provide to the agent if asked.

In the above example: Name - John Doe, Email - john.doe@test.com, Order ID - #122022, Phone number - +01 5522819 3382.

3. Success Criteria - Describes the result with which the customer will be satisfied and stop asking further questions.

In the above example: Close the conversation if they offer to send a refund form or escalate the issue.

Next Steps

You will need to configure a rule with a live chat account to trigger the case. Follow the steps in this article.