Voice AI Identifies Knowledge Gaps in Conversations

TL;DR

HighLevel's Voice AI agents now flag questions they can't answer as 'knowledge gaps'. This is a major update for anyone using or planning to use Voice AI.

Who this affects

HighLevel has not confirmed who this affects. It is relevant to users who deploy Voice AI for handling incoming calls and those who manage knowledge bases for AI agents.

What changed

When a caller asks a question that the Voice AI agent cannot answer using its existing knowledge base, that specific question will automatically appear as a 'knowledge gap'. HighLevel will group similar questions together, organizing them by topic and subtopic, to provide a structured overview of what the AI agents are missing.

HighLevel has not confirmed any limits, plan gating, or rollout status for this feature.

Where to find it

HighLevel has not published the exact location for this feature yet. It is expected to live within the Voice AI configuration or reporting sections, likely accessible from the sub-account dashboard where Voice AI agents are managed.

How agencies will use this

Agency owners can leverage this to refine their clients' Voice AI deployments. For instance, an agency setting up a Voice AI agent for a chiropractor client might find that the AI consistently flags questions about specific types of insurance not covered in the initial knowledge base. The agency would then navigate to the identified knowledge gaps, see a grouped list of insurance-related inquiries, and add the necessary information (e.g., policy details, accepted providers) to the Voice AI agent's knowledge base. This allows the agency to quickly identify and address deficiencies in the AI's understanding, improving its ability to handle calls and reducing the need for human intervention. This process allows for continuous improvement of the AI's effectiveness for a specific sub-account.

Our take

This is a major update. Understanding where an AI agent fails is crucial for improving its performance. This feature directly addresses that, providing actionable insights into knowledge base deficiencies. It reduces the manual effort of reviewing transcripts to find missed questions. For agencies serious about deploying effective AI solutions, this is a significant step forward in making those solutions more robust and reliable.