# Complex documents to cited decision brief

> Convert complex files into structured, traceable evidence for a human decision without relying on unverified document summaries.

- Canonical: https://gptnavi.com/workflows/complex-documents-to-cited-decision-brief
- Category: Research
- Difficulty: Advanced
- Setup time: 3 hours
- Estimated time saved: 4-10 hours
- Last materially updated: 2026-08-24
- Best for: Consultants, Operations teams, Procurement teams, Research teams
- Tools: Cohere Parse, LlamaParse, Reducto, PageIndex, Notion

## Quick answer

Use parsing and document AI to accelerate reading, not to fabricate certainty. Important claims should remain linked to a page, source excerpt, or original file for a responsible reviewer.

## When to use it

- RFP comparison
- Policy review
- Contract intake
- Due diligence brief

## Steps

1. **Set a source and confidentiality policy** — Define allowed file types, access owner, retention, questions to answer, decision deadline, and claims requiring specialist review. Tool: Notion. Expected output: A document-review brief.
2. **Parse layout-sensitive files** — Convert source documents into structured text, tables, forms, captions, and page references while retaining the original files. Tool: Cohere Parse. Expected output: A structured source package.
3. **Cross-check difficult formats** — Run a second parsing route for documents where tables, scans, or layout structure are material to the decision. Tool: LlamaParse. Expected output: A parsing comparison note.
4. **Extract decision fields** — Pull only the schema needed for comparison, and label uncertain, missing, or conflicting fields instead of guessing. Tool: Reducto. Expected output: A reviewable decision table.
5. **Answer with traceability** — Ask questions over the approved source set and require every important conclusion to show its original context before the human owner approves it. Tool: PageIndex. Expected output: A cited decision brief.

## Prompt templates

### Document comparison schema

Design a comparison schema for these documents. Include decision criteria, exact fields to extract, source page or section requirement, confidence, missing-data flag, conflict flag, and questions for a specialist reviewer. Documents and decision: [paste]

### Cited decision brief

Create a concise decision brief using only the supplied document evidence. Separate sourced facts, interpretation, unknowns, conflicts, and recommended next checks. Every important claim needs a source reference. Evidence: [paste]

## Common mistakes

- Treating extracted text as legally authoritative
- Discarding source-page references
- Using document AI with confidential files without an approved data policy

## Related workflows

- https://gptnavi.com/workflows/multi-agent-research-to-decision-brief
- https://gptnavi.com/workflows/governed-agent-with-connected-tools
- https://gptnavi.com/workflows/llm-evaluation-before-production