← All work
Operating estimate·ai
Sourced estimating for property claims
Enter an address, get a priced and defensible estimate back immediately — with every figure labeled by where it came from: measured on site, model-assumed, catalog-priced, or public record.
The problem
Estimating tools hand you a number and no provenance. When a carrier and a contractor disagree about that number, neither can show their work, so the dispute is about trust rather than evidence — and the estimate is the artifact the entire claim settles against.
My role
Product and engineering on the estimating intelligence layer.
Decisions worth defending
- Four provenance classes on every line item — measured, assumed, catalog, public record — so the reader knows what to check before sending, rather than trusting or distrusting the whole document.
- Assumptions show their reasoning and become editable the moment they are wrong for the job; a cached catalog states how stale it is wherever staleness changes a number, never silently.
- Priced per estimate rather than per seat, because the estimate is what has a real cost — and the security page describes only what is built, marking single sign-on explicitly as not yet available.
Where it stands
Live, with a public try-it flow that needs no account. Cloud, on-premise, and hybrid deployment against one estimating pipeline.