As more organizations decide to use AI to automate their contact centers, many leaders have chosen Google Cloud’s Contact Center AI (CCAI) to power their new virtual agents. 

Here at Nuvalence, we’ve used CCAI to create powerful Dialogflow Agents for large-scale organizations. We’ve learned a lot along the way, and have created tools and best practices that have helped us significantly accelerate development over time. It’s knowledge that we want to share with other organizations who might be trying to do the same. 

So with this in mind, we are thrilled to introduce Nuvalence’s Agent Essentials for CCAI: an open-source virtual agent accelerator.*

How “Agent Essentials for CCAI” Turbocharges Call Center Automation

If you’ve read our recent blog post, “Fast-Track Your Dialogflow Virtual Agents with Auto-Generation,” you’re probably familiar with the complexity around creating new virtual agents and keeping them current. Agent Essentials for CCAI is an open-source suite of tools that tames some of that complexity while automating mundane tasks.

Let’s say your company just discovered the power of Google Cloud’s Contact Center AI (CCAI) tools, and now you have been tasked with developing a Dialogflow Agent to automate your call center. Your team’s product manager has identified a number of requirements for the Agent, which must be able to:

  • Answer over 100 different questions.
  • Break lengthy answers into more manageable chunks, and prompt the user if they want more information.
  • Repeat the last answer upon request.
  • Provide URLs for users to find more information when they mention “website” or “URL”.

Of course, all of this must be delivered as soon as possible. And just to keep things even more exciting, you must also provide support for several languages.

The good news? All of this can be done with a combination of spreadsheets and CCAI’s tools. The bad news? Setting up your virtual agent requires many manual steps for each question – everything from defining naming conventions, manually copy/pasting text from one location to another, manually updating intents, wiring up the flow, testing, and much more. Plus, every one of those manual steps introduces the potential for error. 

Ask yourself how much time you need to complete all of them, and multiply that by the number of questions you need to support. You’ll probably end up with a level of effort that equates to days of work.

You can create a level of automation described in our previous blog post, but why spend the time developing and testing that process when you can simply turn to Nuvalence’s Agent Essentials for CCAI, and create your virtual agent in minutes instead of days?

How It Works

When automating a call center, you need to consider two perspectives: the Agent itself, and the interaction between the Agent and external systems. Agent Essentials for CCAI accelerates the first perspective (the Agent itself).

  1. The Agent design: how to organize the flows, naming conventions, patterns, etc. If you’re new to Agent creation, check out our recent blog post on the topic.
  2. The Conversation design: determining the questions the Agent must answer, the training phrases for each question, keeping the answers succinct, etc.

The Agent Generator: Getting a Jump Start

Conversation design is an interactive and iterative process that involves brainstorming potential questions and answers, categorizing, reviewing, and refining them – and ideally, testing them. This collaborative process is fluid and agile, which means you need an overall view of categories and questions, the ability to rearrange them, split lengthy answers into chunks with follow up questions, etc. While you can see the intents, questions, and answers using the Dialogflow GUI, it is not the most convenient tool for this task – a spreadsheet is a much more universal and flexible way to visualize, filter, sort, and share information.

Agent Essentials for CCAI fits seamlessly into this workflow. The Agent Generator tool takes your pre-formatted Google Sheet containing intents, questions, responses, and follow-ups, and it auto-generates a fully-functional FAQ-style Agent for you. This way you can not only conveniently design your conversation flow, but also test the design along the way.

The Agent Translator: Simplifying Translations and Prosody

After a few iterations, your conversation should be stable enough for you to make changes directly to the Agent itself: implementing more complex flows, and interactions with backend systems, for example. 

Looking back at our example scenario earlier in this post, this is the point where you will incorporate the requirements for translations, as well as prosody.

  1. Support for additional languages: adding language-specific intents and responses.
  2. Prosody: when integrating with a telephony system, things like URLs and phone numbers require SSML tags for proper enunciation. For instance, you probably want to pause for a bit and then slow down when speaking those elements to ensure the user has time to write them down. 

Agent Essentials for CCAI has you covered: the Agent Translator tool enables you to export all intents and phrases for your current Agent to a Google Sheet, where you can conveniently make adjustments, including supporting new languages with a simple addition of a column. You can then import these changes, and Agent Essentials for CCAI merges them back into the Agent.

But that is not all! As part of the import process, Agent Essentials for CCAI also generates the audio output – it detects URLs and numbers, and it adds the proper SSML tags for clean enunciation over the phone. By automating this process, Agent Essentials for CCAI not only does the work for you, but it also ensures that the changes between the text and the audio outputs are always in sync.

Parting Thoughts

Agent Essentials for CCAI has helped us accelerate creation of Dialogflow virtual agents by:

  • Using a simple Google Sheet to drive the generation of a fully-functional Agent, therefore
  • Enabling quick prototyping and experimentation in the early stages of the conversation design, while
  • Simplifying the process of changing the Agent’s training phrases and responses, and 
  • Making it easy to support additional languages once the Agent is developed.

While these tools can help you accelerate the time needed to actually build your virtual agent, it’s important to remember that automating a contact center with AI is about much more than pipes and wires. Your team must invest in analysis, design, and best practices for implementation and maintenance. In an upcoming post, we’ll share insights we’ve learned around the ins and outs of conversational design. Stay tuned!

* NOTE: Agent Essentials for CCAI is an open-source project that was independently developed by Nuvalence engineers to accelerate creation of virtual agents using Google Cloud’s Contact Center AI (CCAI) and Dialogflow CX products. 

Check Out Nuvalence’s Agent Essentials for CCAI, now available in GitHub!