Intencion launches product analytics for AI agents, capturing user intent, tool calls, and outcomes to guide weekly improvements

Intencion is a product analytics platform built for teams running customer-facing AI agents in production. It captures every run end-to-end: what the user wanted, the steps the agent took, and how it ended. Installation requires one line of code and adds about 1 ms of latency, flushed in the background.

The platform maps user intent to agent actions and outcomes, then ranks what to fix and what to build next by frequency. Unlike observability tools (which track what the system did), Intencion tells you whether the user got what they came for and shows emerging requests your agent cannot handle yet. It works alongside existing tracing and analytics tools and is published for TypeScript and Python.

In the illustrative example, "Change my plan" resolves 61 percent overall, but "can't verify identity" causes 43 percent of its failures on the path verify_user → lookup_account → escalate. Fixing that one step weekly, the intent climbs from 78% resolved in Week 1 to 93% in Week 4 (+15 points in a month).