Skip to content

Please submit the form below with your Attendee ID in 3 digits long format (e.g. if your attendee ID is 51, please enter 051) and click Save. All configuration items in the lab guide will be renamed with that prefix.



Lab 3: Seamless AI to Human Agent Handoffs

Objectives

This lab is designed to explore how to pass contextual intelligence from AI Agents to Webex Contact Center agents. It involves leveraging AI Summaries for Webex AI agent conversational transcripts . By completing this lab, you will gain practical skills and knowledge on how to provide the right context to agents to better handle customer queries.

By the end of this lab, you will:

  • Understand how to effectively transfer and hand off to Human Agents.
  • Learn how to provide the relevant call context to Human Agents.

Step by Step Guide

  • Before you start this lab, please make sure the webex contact center channel is set to correct Flow




Step 1: Transfer to Human Agent

  • Log into Control Hub using your contact center admin credentials. Select 'Contact Center' from the left panel and then navigate to Flows from the left panel.

  • Search For flow Wx1_VA_attendeeID where attendeeID is your attendeeID by using the search bar and then clicking on the cross launch icon




  • Switch the Edit button to on to enable Edit mode in the flow builder.

  • Drag and drop the Queue Contact activity onto the Flow from the left side panel

  • Drag and drop the Play Music activity onto the Flow from the left side panel.

  • connect the Escalated path from the Virtual Agent V2 activity to the Queue Contact activity.




  • Select the Play Music activity and choose the defaultmusic_on_hold.wav file from the Music File dropdown.

  • Select the Queue Contact activity and set the Queue name to attendeeID where attendeeID_Queue

  • Connect the Queue Contact activity to the Play Music activity.




  • Create a loop by connecting the Play Music activity back to itself - to create a music loop, following the diagram provided.

  • Connect the Failure path from the Queue Contact activity to the Disconnect Contact activity.

  • Connect the Failure path from the Play Music activity to the Disconnect Contact activity.




  • Enable the Validation toggle, verify there are no Flow Errors, and then publish the flow.




  • Log in to the agent desktop using the link https://desktop.wxcc-us1.cisco.com/ and the agent login wxcclabs+agent_IDattendeeID@gmail.com

  • select the Team shown

  • Set the Agent status to Availablefrom the top right corner.




  • Dial the support number assigned to your POD and during the conversation with the virtual Agent, say, "Please transfer me to an Agent." Answer the call on the agent desktop when you receive a ring notification and verify the trasciption is passed to Agent Desktop.

  • Once the call is answered, disconnect the call by clicking on the End button.




Step 2: Disable Virtual Agent Transcript

  • Open your flow Wx1_VA_attendeeID where attendeeID is your attendeeID

  • Select the Virtual Agent activity and, in the right side panel, scroll down and notice the option for Enable Conversation Transcript.

  • Disable the Virtual Agent transcript by unchecking Enable Conversation Transcript option for the Virtual Agent V2 activity, and publishing the flow again.




  • You can dial into the same support and observe that the conversation transcript is "Not available" on the Agent Desktop when Enable Conversation Transcript is unchecked.




Step 3: Routing Based on Last Intent

Important Prereq for Step 3

Before Starting this lab, please make sure to Enable the Virtual Agent transcript by checking Enable Conversation Transcript option for the Virtual Agent V2 activity, and publishing the flow again

  • Open your flow Wx1_VA_attendeeID where attendeeID is your attendeeID

  • In the global flow properties section, click on Add Flow Variable.

  • Enter a new variable named last_intent. Variable Type should be set as ‘String’ then click on ‘Save’ button.




  • Drag and drop the Parse activity to the flow from the left panel.

  • Configure the Parse activity with the following settings:

  • Output variable: VirtualAgentV2.MetaData

  • Content Type: JSON
  • Output Variable: last_intent
  • Path Expression: $.previous-intent.name

  • Connect the Escalated output from the Virtual Agent V2 activity to the Parse activity.




  • Add the Condition activity to the flow from the left panel.

  • Configure the expression in the Condition activity to {{ last_intent == "Book appointment" }}

  • Link the Parse activity to the Condition activity.




  • Insert the Play Message activity into the flow from the left panel.

  • In the Play Message activity settings, enable Text-to-Speech and select the Cisco cloud Text-to-Speech connector.

  • Click on Add Text-to-Speech Message.

  • Click the Delete button next to the Audio file dropdown to remove any existing audio file.

  • Enter the text "Routing to an agent skilled at booking an appointment" in the Text-to-Speech Message area.

  • Connect the True output of the Condition activity to the Play Message activity.

  • Connect the Play Message activity to the Queue Contact activity.

  • Direct the False output from the Condition activity to the Queue Contact activity.




  • Enable Flow Validation to check for errors, then publish the flow.




  • Log in to the agent desktop using the link https://desktop.wxcc-us1.cisco.com/ and the agent login wxcclabs+agent_IDattendeeID@gmail.com

  • Set the Agent status to Availablefrom the top right corner.




  • During your interaction with the virtual Agent, request a transfer by saying, "Please transfer me to an Agent." If the last intent was "Book appointment," you will hear the Text-to-Speech message: "Routing to an agent skilled at booking an appointment." Answer the call on the agent desktop when it rings.

  • End the call by clicking the End button, choose a wrap-up reason, and submit it with the Submit Wrap Up button.

Step 4: Virtual Agent Summary

  • Click on the AI assistant icon located on the top left navigation panel.




  • Dial the support number assigned to your POD and initiate a conversation with below

Sample Conversation

"I would like to Book an appointment"

What date are you considering for your visit

"Nov 20th"

Could tell us preferred time for your visit

"3PM"

Which doctor you want appointment with

"Dr John"

What is name of the patience

"Peter"

Could you tell us patience Date of Birth

"Please transfer me to an agent "

  • During the interaction with the virtual Agent, request a transfer by saying, "Please transfer me to an Agent." Answer the call on the agent desktop upon receiving the ring notification.
  • Observe that, after answering the call, a summary of the Virtual Agent interaction is now displayed on the agent desktop




Congratulations, you have officially completed the Intelligent Virtual Agent Handoffs lab! 🎉🎉