ทันเอไอ

YC-backed startup Agnost AI launches conversation analysis tool to detect 'unspoken' user frustration

New startup Agnost AI develops an automated tool for analyzing AI agent conversations, uncovering hidden user issues and needs that product teams would otherwise have to review manually.

Agnost AI, a startup backed by Y Combinator (YC S26), has launched a product analytics tool specifically for teams building chat and voice AI agents. It addresses the limitation of traditional metrics, like click tracking, which fail to measure the performance of conversational products.

At its core, Agnost AI analyzes real-world conversations between users and AI agents in production to identify "behavioral failures"—subtle indicators of issues that users don't explicitly report. Examples include "rageprompting" (cursing at the bot), repetitive attempts at the same query with different phrasing, correcting the AI's misunderstandings, asking for non-existent features, or abandoning the chat despite technically accurate responses.

Agnost AI integrates with existing conversation logs without needing additional instrumentation. The system identifies both explicit and implicit feedback, categorizing them into intents, feature requests, frustrations, and failure patterns. These insights are tagged and exported, enabling product teams to plan updates or fine-tune models. According to Y Combinator, the system can even generate Pull Requests (PRs) to improve prompts and agent tools.

Founders Shubham and Parth stated on Hacker News's Launch HN that their goal is to help developers understand user pain points that otherwise go unvoiced. The company reportedly raised $250,000 in pre-seed funding from Entrepreneurs First, though this remains unconfirmed by the company.