A payer's record of itself lives in fragments, spread across other people's systems: claims with one vendor, eligibility with another, prior years with a contractor whose contract has ended.

Each extract is partial, and where they overlap they disagree. Every reconciliation comes down to three questions. Is this the same patient? Do these records conflict? Which source is more reliable?

HARROW is the record the payer owns: those fragments brought together inside a boundary the payer or its sponsor controls. Nothing migrated. Nothing replaced. Nothing turned off.

Every value keeps where it came from: the source, the raw bytes it arrived as, and a hash over those bytes. Where sources disagree, both values stay. When a use needs one, it picks by a named rule, and the pick is recorded.

Three weeks later, the eligibility vendor corrects a coverage date. The old date already fed a risk score, a quality measure and a payment check. The correction lands beside the old value, and HARROW follows it forward: every answer that used the old date goes back to its owner to reopen.

The machine decides structure. People decide everything else. Disagreements no rule settles, and every answer a correction reaches, go to your reviewers with the machine evidence beside them.

Deterministic outputs, built to be checked: receipts, a fixed PDF evidence memo, an XLSX universe with the hard rows still in the denominator, and a replay manifest a reviewer can re-derive without taking the record's word.

The fragments already exist. The vendors already run. What is missing is the record the payer owns. Request the briefing →