Audit against a reference

Blueprint compares each document against a template, checklist, or standard: gaps, matches, and mismatches per source.

Blueprint is the compliance / conformance workflow. You provide a reference (a template document, a checklist, a set of rules) and Blueprint checks every source in the dataset against it. The output is a per-document audit report: what matches, what's missing, what conflicts.

What you get

One report per source in the dataset. Each report breaks down the comparison: items in the reference that match the document, items missing from the document, items in the document that conflict with the reference, and items that need human review.

Reports are downloadable individually or as a ZIP, and can be saved into your workspace as notes.

Run it

  1. 1

    Build the dataset of documents to audit.

  2. 2

    Open the workflow runner and pick Blueprint.

  3. 3

    Provide a reference (see below).

  4. 4

    Optionally add custom instructions narrowing the audit scope or the severity of findings.

  5. 5

    Click Run.

Reference options

Two shapes of reference work:

  • A document. Upload a template document (a master contract, a policy template, a standard operating procedure) and Blueprint uses it as the spec.
  • A checklist or rule list. Paste a list of items each source must contain or comply with. Bullet points work; numbered lists work; a free-form description works.

The reference can be one or both: a master template plus a list of extra rules.

Common patterns

  • Contract compliance: master template as reference; check every counterparty contract against it.
  • SOX / ISO controls: checklist of required controls; audit each policy document against the list.
  • Editorial style: brand guidelines as reference; check every published piece against it.
  • Code review against architecture: architecture doc as reference; check that each component spec aligns.

Tips

  • Be explicit about what counts as a finding. Custom instructions like "only flag items missing entirely; don't flag wording variations" sharpen the report.
  • Iterate on the reference. Run Blueprint on three sources first, look at the false positives, refine the reference or instructions, then scale.
  • Combine with Data Extractor. Extract the structured values first; audit the narrative against the reference second. Each does its job well.

What's next

Was this helpful?