Customer Zero

We didn't pilot Ruby. We rebuilt our own company on her.

The restoration billing company that built Ruby now runs on her. Fourteen agents work its claims. Everything below is our own operating record, not a projection.

100,000 claims$1B in claim dollars under management1.4M tasks worked to closure1M calls and texts1.2M emails221 contractors

8x

output per person

Measured against 2024, while claim volume rose 50%.

75%

agent share of the work

Share of all task closures done by an agent rather than a person, from zero in 2024.

16 days

faster to the first carrier dollar, yield held

Yield per claim held through a period of significant carrier hardening.

How it operates

One layer decides. Many execute.

Nothing runs on a hunch. Every action traces to the same written procedure a person would follow.

Autopilot with a named exception path. Managers review exceptions, not files.

Every decision is graded. An agent scores the other agents; corrections are read on the next run.

One face per claim. A counterparty deals with the same name from first notice to paid.

Voice is just another channel. A call is answered with the claim file already open.

Service as software. The people carry the judgment the agents hand up.

The desks

Five desks counterparties meet. Eight they never hear about.

Each desk is an agent with one job, shown here by its role.

Submissions
Builds the packet and submits the invoice.
Carrier desk
Works the file, answers the comparative estimate.
Collections
Works the property-owner balance, in English or Spanish.
Voice
Answers and places calls with the file open.
Correspondence
Drafts outbound email and holds it for a person.
Accounts receivable
Finds the check, posts it to the right claim.
Comp mining
Reads every carrier estimate to the line.
Documentation grading
Scores the packet before the carrier sees a defect.
Quality assurance
Grades the other agents.
File hygiene, routing, prediction, reporting
Four more desks working the books.
What actually changed

Every agent runs on the same three things.

One claim record, one set of written procedures, and one graded feedback loop that reads its own corrections. That is the product we sell as Ruby. We ran it on our own company first.

Run your team. We'll run the rest.