Agnost AI, a YC S26 company, offers product analytics for chat and voice agents by extracting user feedback from conversations. Co-founders Shubham and Parth identified a gap in metrics for language-based products, where traditional clickstream data is less effective.

Agnost AI analyzes conversations to detect behavioral failures like rage-prompting, rephrasing, and feature requests that users often don't explicitly state. The platform clusters conversations into product-specific intents, including feature requests and bugs, dynamically evolving with user data. For instance, one AI video editor company discovered 70 users wanted auto-subtitles through varied phrasing in their chats.

The system processes millions of messages efficiently using a combination of techniques, including embedding clustering and smaller classifiers, to avoid excessive LLM use. Agnost AI is currently ingesting approximately 1 million messages daily, with pricing tiers including a free Starter plan, a $499/month Pro plan, and custom Enterprise options. They are SOC 2 Type 1 compliant.