AI model

Overview

Ortto's AI model boosts efficiency and accuracy in managing customer inquiries within the Talk inbox. It suggests tailored responses based on imported resources and agent interactions, streamlining the customer service process. Automatic updates keep the AI current, and manual updates integrate new content sources for ongoing improvement in response quality.

Accessing AI model

To access the AI model, go to Settings > Inbox > AI model

Managing AI model

The AI model page allows users to:

• Setup your AI model
• How the AI model works
• AI performance

Set up your AI model

1. Navigate to Inbox > AI model.
2. Initially, the model will be empty. Click Setup to begin.
3. On the AI settings page, click Add source to import relevant materials like your organization's help documentation or marketing resources. You can also integrate data from linked sources such as Zendesk.
4. Once content is imported, Save your changes to finalize the setup.

How the AI model works

Versioning and updates

• The AI model starts with version 1.0 upon content import and automatically updates every 14 days. Updates labeled Live are ready for immediate use.

TIP: If you've updated your content sources and prefer not to wait for the next scheduled update, you can initiate the update manually. Simply click Update now in the banner at the top of the page.

Data accumulation

• The model aggregates data from integrated sources, saved replies, and insights gathered from agent interactions. Each data point contributes to training the AI.

Agent utilization

• When agents utilize AI suggestions—either as-is or with modifications—it aids in refining the AI model with each iteration.

Manual updates

• Depending on your Ortto plan, you have a set number of manual updates per month. It's recommended to initiate a manual update, particularly when introducing new content sources or making substantial updates to your help documentation that you want the AI to integrate.

AI performance

The Performance section evaluates AI effectiveness based on AI suggestions used by agents within a 7-day rolling window and considers the following:

• Satisfactory: Number of AI suggestions effectively used.
• Unsatisfactory: Number of AI suggestions generated but not used.