Aviary

About us

AviaryAI builds AI voice and SMS agents for credit unions, community banks, and insurance providers. Our agents call members and customers, and the same generative AI carries the conversation over text, so someone who does not answer the phone can finish the same task by message.

We handle the outbound work financial institutions never have enough staff for: onboarding, card activation, dormancy reactivation, early collections, and loan servicing.

We went through Y Combinator in 2022. We are structured as a credit union service organization (CUSO), with credit unions among our owners, and several of our investors are also our customers. That arrangement keeps us honest about what is worth building.

How we work

We work in person. Sales, operations, and our forward deployed engineers (FDEs) are in Chicago. Platform engineering is in New York.

Regulated is the constraint every one of us works under. Our customers answer to examiners, so an agent that says the wrong thing on a recorded call is not a bug report, it is a compliance event. Engineers build for auditability. Product decisions get made with a risk committee in the room. Client teams walk institutions through security review before a single call goes out. Whatever your function, that constraint shapes the work.

We run a forward deployed model. Rather than handing the product over a wall, the people who build it sit inside real customer deployments, configure live workflows, and hear what is broken from the person it is broken for. Most people here talk to customers regardless of their title.

Benefits

• Medical, dental, and vision coverage

• 401(k)

• In office meal stipend

• Unlimited PTO

• Parental leave

• Equity for full time employees

Why people join

The market is real and badly underserved. Credit unions and community banks compete against institutions with hundreds of times their technology budget. What we build changes what a 200 person credit union is capable of. That is not a mission statement, it is the business model.

The AI problems here are the hard ones, and they are already in production. Millions of calls have already run with real members on the other end. Real time latency. What an agent does when it does not know the answer. When to hand a call to a person. How you prove to an examiner what was said. None of these are demo problems, and none of them belong to a single team.