Measure and improve

Reviewing Conversations to find knowledge gaps

Reviewing conversations is a crucial step in identifying knowledge gaps within your AI system and improving the overall quality of customer interactions. By analyzing real conversations between customers and agents, you can uncover where the AI struggles to provide accurate or helpful answers, allowing you to target updates to your Living Knowledge base effectively.

why reviewing conversations matters

Conversations provide direct insight into how your AI performs in real-world scenarios. Negative feedback or low ratings from customers often highlight specific moments where the AI’s response was insufficient or incorrect. Regularly examining these conversations helps you spot recurring issues, such as misunderstood questions or missing information, that indicate gaps in your knowledge base. Addressing these gaps ensures users receive better, more reliable answers over time.

how to review conversations effectively

filtering and selecting conversations

Start by filtering conversations to focus on those with negative ratings or low customer satisfaction. These conversations are the most valuable for identifying knowledge gaps because they reveal where the AI failed to meet user expectations. You can also filter by conversation length, specific Moments, or customer segments to narrow down your review to relevant cases.

inspecting conversation details

Open the detail view of a conversation to read the full transcript. Each agent answer shows which Sources were used, any feedback received, and whether Procedures or Signals were triggered. This detailed view helps you understand the context of the AI’s response and pinpoint exactly where the knowledge gap occurred.

flagging and reporting issues

If you find a conversation that clearly demonstrates a knowledge gap or a bug, you can flag it for team review or report it to the Unless support team. Adding a contextual note with your flag helps others understand the problem and prioritize fixes.

addressing knowledge gaps found in conversations

updating living knowledge

When a content gap is identified, update your Living Knowledge base with the missing or corrected information. This ensures the AI has access to accurate data for future interactions. The Inbox feature can also surface content gaps detected during conversations, making it easier to track and resolve them.

promoting questions to quality reports

You can promote specific questions from real conversations into your Quality reports control set. This allows you to create targeted tests and monitor improvements related to previously identified knowledge gaps.

monitoring improvements

After making updates, continue to monitor conversation ratings and AI insights regularly. This ongoing review helps confirm that the changes have improved AI performance and that no new gaps have emerged.

conclusion

Reviewing conversations is an essential practice for maintaining and enhancing your AI’s knowledge base. By focusing on negatively rated interactions, inspecting detailed transcripts, and acting on identified gaps, you can ensure your AI delivers accurate, helpful answers that meet customer needs. Regular conversation analysis combined with timely updates to Living Knowledge supports continuous improvement and higher customer satisfaction.