AIdapt gives you several ways to identify where your AI agent is underperforming and what to improve. This page walks you through the main workflows for finding actionable insights.

Start with the KPI Dashboard

The KPI Dashboard gives you an at-a-glance view of how your project is performing over time. Look for:

• A low or declining average quality score — signals systematic issues worth investigating
• Spikes in average handle time — longer conversations may indicate the bot is not resolving user needs efficiently
• High bot verbosity — overly long bot responses can reduce user satisfaction

Use the time range filters to compare performance before and after changes to your agent.

Find problematic conversations

The Conversation List lets you filter and browse individual conversations. To focus on problem areas:

• Sort by quality score (ascending) to surface the lowest-rated conversations
• Filter by tags to focus on a particular conversation type or topic
• Review the short summary to quickly understand what each conversation was about without reading the full transcript

Open a conversation to read its detailed summary and full transcript to understand what went wrong.

Use custom metrics to track specific issues

If you have a recurring issue to monitor — for example, whether the bot always offers a handover option — define a Custom Metric. Custom metrics run automatically on every conversation and appear in the Conversation List, making it easy to spot patterns at scale.

Annotate and track conversations manually

When you find a conversation worth following up on, you can:

• Add a manual rating and comment to flag it for your team
• Apply tags to group related conversations for later review