Topic Analyzer

Automatically analyze and categorize conversations by topic

Topic Analyzer automatically assigns topics to conversations based on content analysis, enabling better organization and insights without manual effort.

Topic Analyzer - Custom Instructions

Overview

Topic Analyzer is an add-on that:

• Automatically Categorizes Conversations: Assigns relevant topics based on content
• Enables Filtering: Search and filter by auto-generated topics
• Improves Analytics: Track topic trends automatically
• Reduces Manual Work: No need for agents to manually categorize

Prerequisites

System Requirements

• AGO platform deployed
• Topic Analyzer add-on enabled
• Topics configured in Conversation Labels

Configuration Required

• Enable in Settings > Topic Analyzer
• Configure custom instructions and settings

---

Enabling Topic Analyzer

Via Admin Interface

1. Navigate to Settings > Topic Analyzer
2. The page shows two tabs: Custom Instructions and Settings
3. Configure the analysis behavior using custom instructions
4. Adjust settings like maximum topics per conversation

Topic Analyzer can also be configured via API. See the Public API v1 Reference for endpoint details.

---

How It Works

Conversation Message → Content Analysis → Topic Matching → Auto-Assign Topics

1. A conversation is completed or analyzed on demand
2. The system analyzes conversation content using AI
3. Matching topics are identified based on configured instructions
4. Topics are automatically assigned to the conversation

Analysis Methods

Method	Description
Keyword Matching	Topics assigned when conversation content matches tag keywords (case-insensitive, whole-word matching)
AI Classification	An LLM analyzes the conversation and assigns the most relevant topics based on your custom instructions and tag descriptions

Both methods run together. Keyword matching provides fast, predictable tagging for known terms, while AI classification catches topics that keywords might miss.

---

Configuration

Custom Instructions

The Custom Instructions tab allows you to define how the AI should analyze conversations and assign topics. Use the editor to write specific instructions that guide the topic assignment process.

You can view a diff of your changes against the saved version before saving.

Settings

The Settings tab provides the following options:

Setting	Description
Maximum Topics per Conversation	Limit how many topics can be assigned (1-20)
Always assign at least one topic	Verify every conversation gets at least one topic

Managing Topics

Topics are managed in the Conversation Labels page under the Topics tab. Each topic can have:

• A name and description
• Keywords for matching
• A parent topic for hierarchical organization

Click Manage Topics in the Settings tab to navigate directly to the Conversation Labels page.

---

Running Topic Analysis

From the Conversation Sidebar

When viewing a conversation in Chat History:

1. Open the conversation sidebar
2. Find the Topics section
3. Click Run Topic Analysis to analyze the conversation
4. Topics are automatically assigned based on the analysis

Filtering by Topics

In Chat History, filter conversations by topic:

• Select topics in the filter panel
• Combine with other filters for targeted results
• Export filtered results

Analytics

Topics feed the Topics Analytics dashboard tab, which shows:

• Topic distribution treemap
• Daily topic trends
• Topics by agent breakdown
• CX Score by topic (when the Thread Evaluator is enabled)
• Escalation rate per topic

---

Practical Examples

Example 1: Support Category Analysis

Automatically categorize support conversations:

Topics to create:

• billing - Payment, invoices, subscriptions
• technical - Bugs, errors, how-to questions
• account - Login, password, profile issues
• feature-request - New feature suggestions

Result: View your Analytics dashboard to see which categories drive the most conversations.

Example 2: Escalation Tracking

Track conversations that require human intervention:

Topics:

• escalated - Applied when ticketing tool is used
• frustrated-customer - Applied by AI when sentiment is negative
• complex-issue - Multi-step problems requiring follow-up

Example 3: Product-Specific Routing

Categorize conversations by product area for team assignment:

Topics:

• product-a - Questions about Product A
• product-b - Questions about Product B
• integration - Third-party integration issues

Keyword matching:

Topic	Triggers when user mentions…
product-a	”Product A”, “feature X”, “module 1”
product-b	”Product B”, “dashboard”, “reports”
integration	”API”, “webhook”, “Zapier”, “integration”

Example 4: Hierarchical Topic Organization

Use parent topics for structured categorization:

Parent Topics:

• support (parent)
  • billing-support (child)
  • technical-support (child)
• sales (parent)
  • upsell-opportunity (child)
  • new-lead (child)

---

Best Practices

Practice	Recommendation
Naming	Use lowercase with hyphens: feature-request
Quantity	Aim for 10-20 core topics (too many reduces usefulness)
Hierarchy	Use parent topics to group related topics
Review	Consolidate and clean up topics quarterly
Monitoring	Check topic distribution weekly for trends

---

Related Documentation

• Thread Evaluator - Conversation quality assessment
• CX Score - Customer experience scoring
• Conversations - Search and filter conversations
• Topics Analytics - View topic distribution and trends
• Conversation Labels - Manage topics and admin tags