Conversational Insights and Quality AI

Conversational Insights helps you detect and visualize patterns in your contact center data. Understanding conversational data drives business value, improves operational efficiency, and provides a voice for customer feedback.

You can use Conversational Insights to analyze conversation data in order to:

  • Run machine learning analytics to gain additional information such as agent and caller sentiment, entity identification, and call topics. Automatically identify interesting interactions in need of further review.

  • Review a conversation with audio playback, transcript synchronization and analytics annotations.

Conversational Insights seamlessly integrates with other products, allowing you to import conversations from Dialogflow CX and Agent Assist.

Quality AI builds on Conversational Insights to enable comprehensive contact center quality evaluations, including individual conversations, live agents, and virtual agents.

Prerequisites

  • Enable Conversational Insights on a Google Cloud project. As a part of the process you must also enable the following APIs:

    • Speech-to-Text: to be able to generate transcripts from the call audio recordings.

    • Cloud Storage API: to be able to access the stored call recordings and transcripts from the Cloud Storage bucket. You can either use a pre-existing Cloud Storage bucket or create a new one.

Service account requirements

Before you can set up Conversational Insights, you need to create a Google Cloud Service Account and key in the project where you will be using Conversational Insights. In addition, you need to grant the following IAM roles to your service account:

For more information on granting roles to a service account see the Google IAM Manage Access documentation.

Configure new storage

Google recommends configuring a new Cloud Storage bucket specifically for Conversational Insights, with its own service account key. The following procedure explains how to do this.

  1. GCS Bucket Name: Enter your Cloud Storage bucket name.

  2. Service Account Key: Drop service account key (JSON). To upload a key you will need a JSON service account key from your Service account. For more information on creating (and downloading) a service account key see Create a service account key.

  3. Browse your files > Folder Path: Enter the path to your Cloud Storage bucket to connect your data to Conversational Insights. For more information on using variables in this folder path see the Developer Settings documentation.

  4. File destinations: Voice and chat preview file paths will be generated based on your folder path entered.

  5. Click Save.

Enable Conversational Insights

  1. Go to Developer Settings > Conversational Insights and click Setup Insights.

  2. Toggle Conversational Insights to On to enable it.

  3. Insights Project ID: Enter your Insights project ID.

  4. Region: Select the required Google Cloud region from the drop-down.

  5. Send call recordings: Toggle On to send call recordings to Conversational Insights. These will be sent as mp3 files.

  6. Send chat transcripts: Toggle On to send chat transcripts to Conversational Insights. These will be sent as JSON files.

  7. Server Setup: Select Existing Google Storage.

View Conversational Insights data

After your conversation data has been uploaded, you can view your Conversational Insights data in the Conversational Insights Console.

The conversation ID in Conversational Insights is either the Insights call or chat ID—for example, call-123 or chat-987. When a multiple-segment voice call is uploaded, the virtual agent segment and human agent segment are separated into two different conversations. The human agent segment conversation ID is postfixed with _2—for example, call-123_2.

You can't use the same Conversational Insights instance to consume conversations from multiple Google Cloud Contact Center as a Service instances. This leads to duplicate conversation IDs, which is not supported. Conversations with duplicate conversation ID fail to be uploaded to Conversational Insights.

Session metadata is also passed from Google Cloud Contact Center as a Service when the conversation is uploaded to Conversational Insights. The metadata values are provided as labels in the Conversational Insights information section when viewing a conversation.

The provided labels are imported:

Metadata key Calls Chats Description
hold_duration ✔ Duration that the end-user was put on hold for in seconds.
customer_phone_number ✔ The phone number of the end-user.
outbound_number ✔ The phone number that the outbound call was placed from (outbound calls only).
transfers ✔ ✔ The number of times the call was transferred.
has_feedback ✔ ✔ Whether the conversation had a feedback response.
fail_reason ✔ ✔ The reason for the failure of the conversation.
out_ticket_id ✔ ✔ The ticket ID of the conversation.
agent_assist_conversation_id_X* ✔ ✔ The Agent Assist conversation ID of the conversation.
dialogflow_conversation_id_X* ✔ ✔ The Dialogflow CX conversation ID of the conversation.
queue_priority_level ✔ ✔ The priority level of the conversation.
answer_type ✔ ✔ The type of the answer.
session_type_v2 ✔ ✔ The type of the session.
menu ✔ ✔ The menu of the conversation.

Other metadata values that are used within Conversational Insights include the following:

Metadata key Calls Chats Description
id ✔ ✔ The call or chat ID of the session used for the conversation ID.
agent_name ✔ ✔ The name of the human agent that handled the conversation.
agent_id ✔ ✔ The internal CCaaS ID that identified the human agent.
virtual_agent_id ✔ ✔ The internal CCaaS ID that identified the virtual agent.

Session metadata stored in Cloud Storage is updated 15 minutes after completion of the session.