Skip to main content

Feeding External Insights to Conversational Bots

The AI models behind Graia CCaaS conversational bots rely on a vast amount of general data that can lead to suboptimal answers to very specific business scenarios. A convenient way to address this lack of subject matter expertise is by using Graia CCaaS Knowledge Bases. Often though you would need to provide actual values of dynamically changing data—e.g. current prices of services, goods, or data dips like account statements, etc. that are available in systems outside of Graia CCaaS. Integrating to these empowers conversational bots not just to fetch factual information specific to the business context of the ongoing conversations, but also to send data to these systems or even execute tasks outside of Graia CCaaS.

Integrating an external service to the conversational bots​

The external service needs to be able to send and receive data and specific events via an API you need to define as described in our Developer's Guide for integrating with Conversational Bots.

Configuring Graia CCaaS for integration​

You provide the signature of your integration API to Graia CCaaS—access details, parameters, and other bits—by specifying them in the portal.

  1. Make sure your external service can communicate with Graia CCaaS as per the above Integrating an external service to conversational bots section.

  2. As an administrator having the required permissions, navigate to the AI Administration > Bot Tools menu of the Graia CCaaS portal and click Create at the top of the list.

  3. Add a unique Name for your external integration (English letters, numbers, underscores, dashes and spaces, up to 64 characters) that will make sense to the person who administers the Graia CCaaS workflow. They will need to be able to decide which of these external integrations to select for which conversational bot node in the flow.

  4. Compose an AI instructional prompt in the Describe Tool to Bot field. This needs to be able to let the AI model behind the linked conversational bot know when/why it should trigger this integration.

  5. Click Add Parameter for each parameter the integration requires. This includes not just the precise Parameter Name but also their Type, Required flag and Description as well.

    Note that the property descriptions are also fed into the AI model behind the bot, offering you the ability to help it fully understand what piece of data is used for what purpose.

  6. Leave Configuration Type on API, and enter the API URL and the Api key.

  7. Under Advanced configurations, decide whether to turn on Hide the integration call from bot context window. When it's on, the bot generates the answer for the customer, then no longer keeps the response in its context.

  8. Click Create.

  9. Select the tool in the Integrations field of the relevant voice, chat or email workflow's conversational bot node.

note

All the fields, especially the descriptions and labels are fed into and control the AI model behind the conversational bot. Both the Describe Tool to Bot and the Parameter Descriptions must conform to the overall bot instructional guidelines for proper operation.

Using JavaScript​

Instead of using an API URL and key, you may set Configuration Type to JavaScript and define the logic here instead of relying on a middleware.

The functionality of the JavaScript integration for the conversational bot tools works the same way as the JavaScript node used in the workflows.

Last updated on