We’re exploring a voice-AI tool — not for tracking the usual stuff like sentiment or response times — but something that digs much deeper. The goal is to uncover hidden insights and tribal knowledge buried across 100% of voice/video agent conversations.
Think of it like what Snowflake or Databricks did for raw data — we want to do that for raw conversations. Turn unstructured voice interactions into something structured, explorable, and actually useful — a knowledge layer on top of voice that helps teams learn what’s really going on.
We’re trying to surface stuff like:
Emerging topics or friction points users keep bringing up
Places where agents drift, hallucinate, or make assumptions
Repeated “off-script” asks or internal hacks no one documented
And patterns that only become clear when you look across 1000s of calls
Instead of just “monitoring,” the idea is to build a discovery engine — something you can query, learn from, and grow with over time.
We’re trying to figure out:
Does this feel like a real gap in the voice AI landscape?
Are current tools just scratching the surface?
What would you want to extract if you had full access to every convo your agent ever had — no tooling limitations?
Is this something you'd actually find valuable day-to-day, or just interesting in theory?
Example: We noticed users repeatedly referencing a process that wasn’t in any KB or doc — just something teams passed around informally. Our goal is to surface that kind of buried, high-leverage knowledge automatically.
Would love your gut reaction — whether you’d use something like this, or not. Brutal honesty welcome