Document analysis and extraction
Pull named fields out of contracts, invoices or reports into a table with verbatim evidence, locations and confidence, plus a list of red flags.
1List the fields you need
Write one field per line with a short definition, for example 'Notice period: days either party must give before terminating'. Definitions stop the model guessing which clause you mean.
3 steps and 1 ready-to-use prompts are in the full version
Get this resource with editable variables, examples and future updates as part of a download pack.
Get Everything Bundle — $49Example input
Document type: software services agreement (14 pages). Fields: Term, Auto-renewal, Notice period, Liability cap, Governing law, Data processing terms.
Example output
| Field | Value | Verbatim evidence | Location | Confidence | | Term | 24 months | "an initial term of twenty-four (24) months" | §9.1 | High | | Auto-renewal | Yes, 12-month periods | "shall renew automatically for successive twelve-month periods" | §9.2 | High | | Notice period | 90 days before renewal | "no less than ninety (90) days prior" | §9.2 | High | | Liability cap | NOT FOUND | | | | Red flags - §12.4 excludes liability for data loss "howsoever caused", which may undercut the security commitments in Schedule 2. Missing - No overall liability cap is stated.
Troubleshooting
- Values look plausible but no evidence quote is given: the model may be inferring. Re-run with 'Leave the row as NOT FOUND unless you can quote it.'
- Page numbers are wrong on uploaded PDFs: ask for section or clause numbers instead, which survive text extraction better.
- Tables in the document are garbled: paste that table separately or convert the PDF to text first.
Known limitations
- This is not legal or financial advice; a qualified person should review anything you rely on.
- OCR quality on scanned documents limits accuracy, and handwriting is often missed.
- Do not upload confidential documents to a tool your organisation has not approved.