Teaching a ChatBot Customer Service 101

This is the second post in our series on Azure Artificial Intelligence, and a continuation of Chatbots 101 – Quick and Easy Customer Service Chatbots. This post will build on our previous post, where we will now teach the chatbot customer service

In the previous post, we created a chatbot using Power Virtual Agent that integrates to the Dynamic 365 to handle a service ticket. However, customers’ needs are, in general, more than a ticket. They might be looking for quick answers for many common systems’ related questions, for example, how can I get my IP address in Windows? In general, there are Frequently Asked Questions (FAQs) available to address such questions. QnA maker allows you to bring-in and get trained on the Question and Answers from online FAQ manuals or files.  Then, QnA maker can respond to the similar questions through API which can be integrated to the services like a chatbot.

Creating a Question and Answer Service Button

For the demo purpose, we will work on bringing in the questions available at online FAQ https://www.microsoft.com/en-us/software-download/faq into our environment.

Go to https://www.qnamaker.ai/ and sign in with your AD credentials, and navigate to Create a knowledge base. There are three steps available to perform on this page.

Follow Step#1. Click ‘Create a QnA service’ button. This will take you to the Create screen – provide details and then hit Create.

Follow Step#2 and 3. Select your newly created QnA service for your Knowledge Base. Name your knowledge base.

Follow Step#4 and 5. Populate and create your knowledge base.

Once knowledge base is created, it will take you to the edit page for questions and answers. Adjust as needed and hit Save and train, then click Test and finally Publish.

Below are the test results. Please note that the questions asked during testing are “close” but “not exact” questions as in FAQ. The service is able to respond.

Below are the results of Publish. Note down the marked details, these will be used while integrating with Power Virtual Agent chatbot.

Cognitive Services

In addition, go to Cognitive Services in portal.azure.com and collect the following information.

Integrating Power Virtual Agent Bot with QnA Maker Knowledge Base.

Pre-requisite: An up and running Power Virtual Agent chatbot. If you do not have one already, follow Chatbots 101 – Quick and Easy Customer Service Chatbots (link provided in the references)

Create a new topic for the chatbot as below.

• Set trigger phrase as windows help.

• In authoring canvas, call an action and create a new flow for it.

• In the flow, add the QnA maker action, Generate Answers, and fill out the information as collected earlier.

The generate answer action should now be able to get the Answers from QnA maker, for the provided Input question from the chatbot.

Now, complete the flow, and save. The final flow will be as follows:

The finished topic will have the authoring canvas as follows:

Now, test the Power Virtual Agent chatbot.

The Power Virtual Agent chatbot is now integrated with the QnA maker.

This concludes today’s topic. Please stay tune for more posts.

If you need more information or looking for an intelligent customer service Bot solution on Azure/Office 365, contact Wintellisys!

References:

Chatbots 101 – Quick and Easy Customer Service Chatbots. https://wintellisys.com/azure-artificial-intelligence-ai-series-building-a-chatbot/

Learn QnA Maker. https://docs.microsoft.com/en-us/azure/cognitive-services/QnAMaker/