Make AI assess your bot responses

AI is great at processing a lot of information and producing useful summaries or making suggestions. We can employ it to suggest improvements to the bot flow based on the bot response data, understand the emotional aspect of the conversations, etc This can be done using the OpenAI Integration Automation.

Step 1. Add and set up the OpenAI integration. Navigate to the organization integrations section:

Step 2. Open the automation section of the integration settings:

• select a bot, responses will be assessed for
• select a period for which the responses will be selected for the assessment
• adjust parameters sent to AI to get the result.

In our example, we ask a model to suggest improvements to the bot flow based on the bot responses:

{
  "messages": [
    {
      "role": "system",
      "content": "You're a chatbot expert"
    },
    {
      "role": "system",
      "content": "Review the chat transcripts and suggest an improvement to the bot flow"
    },
    {
      "role": "user",
      "content": "Chat transcripts"
    },
    {
      "role": "user",
      "content": "{{chatTranscriptsMarkdown}}"
    }
  ]
}

As you can see, variables are supported in the parameters. Namely, we use the chatTranscriptsMarkdown variable that will be resolved to the bot responses text transcripts in the Markdown format.

We’re done with the settings!

Once the action is executed, the result will be sent to the organization’s finance email that you set in the organization’s settings.

An email will look like this:

Edit this page