How Check Works

How Check Works A data-flow diagram generated by Archify. 01 / Input 02 / Decompose 03 / Check 04 / Conclude 05 / Deliver Regulation · atomized into requirements · 01 / Input · KB-brain · read-only Regulation atomized into requirements KB-brain · read-only Document · human- or AI-authored · 01 / Input · the Case Document human- or AI-authored the Case Decomposition · claims, citations, data · 02 / Decompose · frozen by hash Decomposition claims, citations, data frozen by hash Requirement check · requirement by requirement · 03 / Check Requirement check requirement by requirement Citation grounding · verified by hash · 03 / Check Citation grounding verified by hash Findings + coverage · objections tied to their source · 04 / Conclude · founded opinion, never a verdict Findings + coverage objections tied to their source founded opinion, never a verdict Evidence-only export · for the auditor / third party · 05 / Deliver Evidence-only export for the auditor / third party Remediation export · for whoever revises the doc · 05 / Deliver Remediation export for whoever revises the doc submit anchored units requirements each citation status objection release release Legend primary data policy / PII data store data flow

Requirement-driven

  • • The regulation is broken into individual requirements
  • • Each requirement is held against the document, one by one
  • • The result is a founded opinion of where it stands, never a verdict

Grounded objections

  • • Every objection traces to a citable source of record
  • • A citation is verified by hash: extract, resolve, then verify membership
  • • We can show a citation is not in the corpus — never that a claim is false

Two exports, one evidence base

  • • A named person signs the opinion before it is released
  • • Evidence-only goes to the auditor or a third party
  • • Remediation goes to whoever revises the document — so the party that certifies never advised on the draft