Topic and Intent analysis

Overview

Reviewing and analyzing your conversation taxonomy can provide insight into trends in your end users’ behavior. Understanding these trends helps you identify opportunities and take action to improve your AI Agent’s performance.

How Intent Intelligence works

Your AI Agent analyzes each conversation and classifies it within a two-level taxonomy:

Topics group conversations by subject—for example, Billing.
Intents sit within a Topic and capture the specific reason behind a conversation—for example, Dispute a charge within Billing.

A conversation is assigned to the Intent that best matches it, along with that Intent’s parent Topic. On the Topics & Intents page, the taxonomy appears as an expandable table: each Topic can be opened to reveal the Intents within it, with performance metrics shown at both levels.

By default, the list is sorted by automated resolution (AR) opportunity, which represents the total opportunity a Topic or Intent has to affect your overall automated resolution rate. Topics and Intents containing the largest number of unresolved conversations have the greatest AR opportunity.

Use cases

Topics and Intents help you understand conversation patterns and prioritize improvements to your AI Agent.

Prioritize high-impact improvements: Identify Topics and Intents with the greatest AR opportunity to focus your efforts where they will most improve your overall automated resolution rate.
Uncover knowledge gaps: Review unresolved conversations within an Intent to find missing or unclear content in your knowledge base.
Monitor customer satisfaction trends: Track CSAT by Topic and Intent to identify where end users are dissatisfied and investigate root causes.
Pinpoint why customers reach out: Use Intents to move beyond broad subjects and understand the precise reason behind each conversation.
Analyze conversation volume patterns: Understand which Topics and Intents drive the most inquiries and how volume changes over time.

Capabilities & configuration

Intent Intelligence provides metrics, multilingual support, real-time classification, and configuration options for managing your taxonomy.

Multilingual support

Classification is multilingual and supports every language enabled for your AI Agent.

Example: If your AI Agent supports French and Spanish in addition to English, a conversation in any of those languages is classified into the same Topics and Intents.

Although classification is multilingual, Topic and Intent names and descriptions are in English.

Topic and Intent metrics

Click any Topic or Intent to view a more detailed breakdown of the data collected, then use this information to:

Explore volume, handoff, and customer satisfaction (CSAT) trends
Review inquiry summaries to quickly understand the key points of a conversation
Uncover opportunities for improvement based on unresolved inquiries
Review conversations that meet specific criteria using the available filters

Metric

Definition

Conversations

The number of conversations classified under this Topic or Intent.

AR rate

The percentage of conversations within this Topic or Intent that the AI Agent resolved automatically, without a handoff to a human agent.

AR opportunity

Estimates how much your overall automated resolution (AR) rate would improve if this Topic or Intent reached 100% AR. It is calculated as (1 − AR rate) × share of conversations, so high-volume, low-AR areas surface first.

CSAT

The percentage of conversations end users reviewed positively, out of all conversations they reviewed.

Understand CSAT reviews

Rating type	Negative review	Positive review
Numeric (5-point scale)	1, 2, or 3	4 or 5
Numeric (10-point scale)	1, 2, 3, 4, 5, or 6	7, 8, 9, or 10
Emoji (5-point scale)	😠, 🙁, or 😐	🙂 or 😍
Thumbs up/down (binary)	👎	👍

Containment rate

The percentage of conversations that did not result in a handoff to a human agent.

Real-time classification

Open conversations are classified about 10 minutes after the end user’s last message, and reclassified when new activity arrives—so your data reflects in-progress conversations rather than only those that have ended. Conversations handed off to a human agent are also classified.

System Topics

Two system Topics ensure every conversation is captured in your reporting. They are read-only and can’t be edited or deleted:

Other – Conversations that fall outside your defined Topics and Intents.
Unclassified – Conversations that are unclear or incomplete, where the end user’s intent can’t be determined.

Quick start

Review your taxonomy and identify improvement opportunities in a few steps.

On the Ada dashboard, go to Analytics > Topics & Intents.

Review Topics and Intents sorted by AR opportunity to find the areas with the greatest potential to improve your automated resolution rate.

Expand a Topic to see the Intents within it, and review the metrics shown for each—conversation volume, AR rate, AR opportunity, containment rate, and CSAT.

Click a Topic or Intent to open the Conversations view filtered to it, and review examples to identify areas for improvement.

For detailed instructions, see Implementation & usage.

Implementation & usage

Use Intent Intelligence to review insights, manage your taxonomy, act on AI-recommended Intents, and export performance data.

Review insights

Gain visibility into conversation patterns and performance metrics across your Topics and Intents.

To view your taxonomy:

On the Ada dashboard, go to Analytics > Topics & Intents.

From any Topic or Intent, you can review the conversations behind it to identify relevant actions to take to improve your AI Agent.

To see conversations for a specific Topic or Intent:

On the Topics & Intents page, click the Topic or Intent. You’ll be taken to the Conversations view with the matching filters applied, so you only see conversations for that Topic or Intent.

Manage Topics and Intents

Create and refine the Topics and Intents that matter most to your business. New Topics and Intents are active and available for classification immediately upon save.

To create a Topic:

On the Ada dashboard, go to Analytics > Topics & Intents.
Click New Topic.
Enter a Topic name and Topic description.
Add at least one Intent. Conversations are classified by the Intents within a Topic, so a Topic needs at least one Intent to classify conversations.
Click Create Topic.

To add or edit an Intent:

When creating or editing a Topic, click New Intent (or open an existing Intent), then define:

Intent name – A short, specific name for the reason behind the conversation.
Intent description – What the Intent covers.
What’s included – Phrases or situations that should match this Intent.
What’s excluded – Phrases or situations that should not match this Intent.

Editing a Topic or Intent definition affects future classifications only. Past conversations keep the label they were assigned, so your data may reflect a mix of results classified under the old and new definitions.

To delete a Topic or Intent, open its More options menu and select Delete. Historical conversations classified under it keep their label, but it will no longer be available for new classifications.

Review recommended Intents

When conversations don’t fit your existing taxonomy, your AI Agent surfaces recommended Intents (and, where needed, recommended Topics) for you to review—so the taxonomy grows from your real conversation data instead of being a fixed list.

To review recommendations:

On the Topics & Intents page, open the recommendations panel.
Review a recommended Intent’s volume and recent trend to judge whether it’s worth tracking.
Click Review, then:
- Approve the Intent to add it to your taxonomy (you can edit its name and definition first), or
- Dismiss it to remove it from suggestions.

Conversation titles

In the Conversations view, each conversation is titled with its primary Intent, giving you an at-a-glance summary of why the end user reached out. Conversations in a system Topic show as Other or Unclassified.

Filter and search

Narrow the table with the search box (“Search topics and intents”) and the available filters, including language, channel, handoff, AR status, CSAT, test user, and variable. You can also set the date range (the last 30 days by default, up to two years).

Export to CSV

Click Download CSV to export the current view of the Topics & Intents table, including each Topic and the Intents within it. The export reflects the date range, search, and filters you’ve applied.

Upgrading from Topics

If your account previously used Topics, a draft Topics & Intents taxonomy is generated from your historical conversation data so you can move to Intent Intelligence without starting from scratch.

To upgrade:

On the Topics & Intents page, review the draft taxonomy. Edit Topics and Intents as needed (see Manage Topics and Intents).
When you’re ready, click Upgrade. Upgrading is final: your legacy Topics are replaced by Topics & Intents.

Legacy Topic data remains available through the Data Export API. Accounts that have not upgraded are upgraded automatically on July 31, 2026.

For guidelines on writing effective Topic and Intent definitions, see Best practices.

Explore additional analytics and optimization options.

Conversations: View individual conversations and their Topic and Intent assignments.
Performance reports: Access detailed analytics on AI Agent performance.
Coaching: Improve AI Agent responses based on Topic and Intent insights.
Knowledge: Update content to address gaps identified through Intent Intelligence.
CSAT: Measure customer satisfaction across Topics and Intents.