Powerful AI,
with people in charge.

People approve the money

AI drafts quotes, invoices and insurance claims, and books routine visits. Anything that commits a business to a price, charges a customer or can't be quietly undone waits for a person to approve it. Read more in Why AI should never send an invoice on its own.

AI says it's AI

When our voice agent answers a call or replies to a message, it identifies itself as an AI assistant. Anyone can ask for a person at any time, and the hand-off keeps the full conversation so nobody has to repeat themselves.

Customer data stays the customer's

Outside AI providers process data only to complete a task and never train their models on our customers' data. Our own models are tuned only on data we have permission to use. Each task sends a model only the information it needs.

Evaluated before it's trusted

No model takes on live work until it has passed a task-specific evaluation built from real, permitted job data and checked by people. New models start on a small share of work and keep being measured after they go live. See One task, many models.

Clear sources

Every AI draft arrives with what it was built from: the call transcript, the photos, the rates it used and anything it was unsure of. People can see why a draft says what it says, and correct it.

Secure by default

Data is encrypted in transit and at rest, access is limited to the people who need it, and we name every company that processes data for us on our Security page.

Always improving

We review these principles as our software and the technology change. Questions or concerns: hello@vilozi.com.