About
Built by an insurer, not adapted for one
Heard comes out of Yasmina — a business already running insurance distribution with real carriers, real transactions and real customers.
Why we exist
The gap we kept hitting
Running insurance operations, the same pattern showed up everywhere. Customers called with questions that had documented answers. They called again to ask where their claim had got to. They called at 11pm after a collision and reached a voicemail.
Every one of those calls was answerable. Most were actionable. Almost none needed human judgement — and yet all of them consumed a person, because the alternative was an IVR tree that customers hate and that cannot do anything anyway.
We looked at the AI platforms available and found the same limitation each time: they could answer, but they could not act. An answer without an action is a handoff, and the handoff is where all the cost lives.
Our position
Vertical beats horizontal here
We are not trying to be the best general-purpose customer support AI. We would lose.
Everyone has access to the same frontier models. Nobody builds a durable advantage there. What is defensible is everything around the model: the integrations into policy and claims systems, the conversation logic an ops lead can actually change, and a data loop of claims, policies and customer behaviour from live traffic.
That last one is the part you cannot buy. It comes from operating in the industry rather than selling into it.
So the goal is narrow on purpose: be the AI operating layer for insurance operations. Customer and operations AI, for one industry, done properly — rather than general customer support AI for everyone, done adequately.
What we are building towards
Zero to a live agent in under a day
A business should be able to go from signing up to a production AI agent handling real customer calls in under a day, without writing a line of code. That is the bar we hold the product to.
- target automation rate
- 70%+
- voice response latency
- <1.5s
- signup to live agent
- <1 day
- founded
- 2026
Come and stress-test it
We would rather you tried to break it on a real conversation than watch a polished demo.