AI Reports

This is a brand-new category of reports, released to accompany AI Agents in Gnatta. Got thoughts or feedback? We’d love to hear about it.

Quick links in this article:

With this collection of reports, you'll be able to explore the performance and impact of your AI Agents across your contact centre. This article will explore the tiles available, and how to set them up on your own Gnatta domain.

Introducing AI Agent reports in Gnatta!

Key Definitions

There are 41 tiles available in this reporting category.

Overview Tiles

The Overview Tiles provide a high-level summary of how AI Agents are being used across your operation. They are designed to answer two distinct questions:

Coverage: How much of your inbound volume involves AI?
Effectiveness: When AI is involved, how often does it resolve interactions end-to-end?

The tiles are split into Agent Involvement Rates and Agent Resolution Rates, each focusing on a different aspect of AI performance.

Overview Tiles for AI Agents

Agent Involvement Rates - This tile measures AI coverage by comparing AI Agent involvement against your total workable volume. Metrics included:

All Workable - The total volume of all new and reopened interactions within the selected time period.
AI Involved -The total number of interactions where an AI Agent was assigned at any point in the journey. This includes interactions that were:
- Fully resolved by AI, and
- Escalated from AI to a human advisor.
% AI Involvement of All Workable - The proportion of all workable interactions that involved AI. This answers the question: What percentage of my inbound volume was touched by AI, regardless of outcome?

Agent Resolution Rates - This tile evaluates AI effectiveness, focusing on how many interactions AI Agents resolve without human escalation. Metrics included:

All Workable - The total volume of all new and reopened interactions within the selected time period.
New Inbound - The volume of newly created interactions only, excluding reopened interactions.
AI Involved - The total number of interactions where an AI Agent was assigned, regardless of whether the interaction was resolved by AI or escalated.
Fully Handled - The total number of interactions resolved end-to-end by AI Agents, without escalation to a human advisor.
% Fully Handled of All Workable - The proportion of all workable interactions that were fully resolved by AI. This shows: How much of my entire operation is being resolved by AI?
% Fully Handled of AI Involved - The proportion of AI-involved interactions that were fully resolved without escalation. This shows: How effective is AI when it is allowed to handle an interaction?

The distinction between the last two metrics is critical:

% All Workable tells you how much of your operation AI is being exposed to.
% AI Involved tells you how well AI performs when it is involved.

Command Metrics

About Command Failure: Commands are only logged as Failed where the workflow has completed but returned a "failure" status via the Update AI Agent action (e.g., order lookup found cancelled order, refund request outside policy window). All other commands (including Execute Only) will be logged as Succeeded.

An Update AI Agent action toggled to failure

Commands Triggered - Total commands activated by AI Agents
- by agent
- by data fields
Commands Succeeded - Commands that completed successfully
- by agent
- by data fields
Commands Failed - Commands that encountered a ‘failed’ toggle in an Update AI Agent action
- by agent
- by data fields
% Commands Succeeded/Failed - Success and failure rates for command execution

Interaction and Message Volume

In Gnatta, messages and interactions are different things! An interaction is a collection of messages and conversations with a customer around a single issue. A message is a single piece of communication belonging to a single conversation - such as an individual text message.

AI Agent reporting tiles showing Interactions and Interactions by Data

Interactions - Total interactions handled by AI Agents
- by agent
- by data fields
Messages Sent - Communication volume from AI Agents
- by agent
- by data fields
Average Messages Sent - Messages per interaction ratios

Handling Time

Average Handle Time - Time AI Agents spend on interactions
- by agent
- by data fields

Human Escalation Metrics

Fully Handled interactions represent complete autonomous customer service - the AI Agent resolved the customer's issue from start to finish. Partially Handled means the AI Agent provided initial assistance before escalating to a human agent.

Use these tiles to evaluate your AI’s ability to work independently!

Fully Handled - Interactions resolved entirely by AI without human intervention
- by agent
- by data fields
Partially Handled - Interactions where AI handled initial contact before human escalation
- by agent
- by data fields
Escalation Confidence Scores - AI confidence levels when escalating to humans
- overall averages
- for escalated interactions only
- by agent and data fields

Explore your AI Agent data

To analyse your AI Agent performance, start by creating a dedicated report. Navigate to Reporting in your Gnatta domain and create a new report specifically for AI Agent metrics.

Be sure to set the report date range, channels and queues appropriately using the filter options at the top!

Next, find Add Tile in the top right, and set the category to AI. This will list all of the AI Agent tiles currently available.

Create a new report to store your AI Agent metrics!

We recommend starting with the AI Agent - Interactions tile to understand total interactions handled by your AI Agent (regardless of whether they were escalated to a human or not).

Let’s create this as a line chart to track volume over time.

Create your first AI Agent reporting tile

If we wanted to see that same data, but split by a custom data field (such as reason for contact!), then we could use the AI Agents - Interactions by Data tile. We’re going to create this one as a bar chart so it’s really easy to see the proportions between each possible reason for contact.

Change the frequency of the data points between Daily/Weekly/Monthly/Total on your tiles to get a better visual, depending on your use case. We’re going to change ours to Total on the bar chart, to get a clean view.

Add more tiles to keep exploring your AI Agent data!

If you want to drill right down into the data, you can use the three-dot option menu next to a tile to view the interactions or export your data.

Use the three-dot menu to drill down into your data and see individual interactions

Finally - remember to save your report and any changes you’ve made to the filters and data range will stay put, so you can come back to your new report whenever you like.

The Report Save button is only available if you make changes to the filters. New tiles are saved automatically.

Save the report to keep your filters applied!