What do my users actually want from my agent?
Most user feedback is invisible. Your users tell your agent what they want in their own language, and that signal is buried in transcripts that nobody reads. Flowlines extracts intent from every conversation, classifies it across recurring patterns, and segments users by what they actually asked for.
CSAT thumbs and NPS surveys are dwarfed by the data your agent already collects every day. Every conversation is a feature request, a missing capability, or a misaligned expectation, in the user's own words.
Flowlines reads your existing traces and runs intent extraction on every session. Recurring intents get clustered. Edge intents surface as discoveries. Each cluster has a name, a frequency, a cohort breakdown, and a list of representative quotes.
What you get instead of CSAT thumbs:
Top intents this week. Ranked by session count, with delta vs prior period. New intents. Patterns that emerged in the last 7 days. Failed intents. Intents the agent received but could not satisfy. Cohort-specific intents. What at-risk users are asking for that happy users aren't.
A pricing question the agent never answered
The intents view surfaces a cluster of sessions where users asked about annual_billing_discount. The agent doesn't have a policy for it, so it deflected or hallucinated. Many of those sessions ended in session_abandon.
That's a feature request with a quantified cost, surfaced in the language users actually used. No survey, no NPS, no feedback widget.
Intent extraction per session
On ingestion, each session runs through an intent classifier that emits one or more intent labels with confidence scores. Recurring labels become first-class entities on the dashboard, with their own cohort breakdowns and outcome distributions.
You can promote any extracted intent into a custom signal so future sessions matching that intent fire an alert.
Last updated 2026-05-28