Reference and review architecture

This diagram shows the proposed review process. It is not a deployed system architecture or evidence of a completed workbook assessment.

flowchart TB
    accTitle: Reference and evidence review process
    accDescr: A dated reference informs applicability decisions, followed by evidence matching, classified findings and human review. The real workbook is used only in the separate approved Copilot environment.
    A["Dated reference"]
    B["Applicability"]
    C["Evidence matching"]
    D["Classified findings"]
    E["Human review"]
    A --> B
    B --> C
    C --> D
    D --> E

The wiki maintains source summaries, version history, review controls and the portable reference. The real Excel Hazard Log stays in the approved environment. Copilot receives the reference and workbook there, with web access disabled.

First establish the applicable capability, requirement, version and organisational role. Then compare both requirement-to-evidence and hazard-to-requirement coverage. Missing evidence remains distinct from demonstrated nonconformance. Manufacturer and deployer decisions remain separate. ^{001, 036, 037}

The arrows describe an original review method. They do not claim that file upload proves complete extraction, or that an AI finding grants release or deployment approval.

An editable version of the diagram is available as draw.io XML. The XML and Mermaid describe the same five stages.