Where document analysis time goes
Contract review, RFP response, due diligence reading, compliance reports, vendor agreements, board pack preparation — every business runs on documents that someone has to read carefully and pull insights from. That work is high-cost (often partner or director time), slow, and error-prone when done at the end of a long day.
A document analysis GPT pre-reads the document for you. It pulls out the terms specific to your firm's standard concerns, summarises in your house style, and flags the things you'd want a human to look at twice. The human's review time drops from 4 hours to 30 minutes — and the 30 minutes is focused on the issues that matter.
What we build
Contract Review Assistant
Upload a 200-page master services agreement; get back a 2-page summary of the IP, liability cap, indemnity, termination, and pricing terms — measured against your firm's standard playbook.
RFP Reader
Incoming 80-page RFP turned into a structured response checklist: required answers, evaluation criteria, deal-breaker terms, suggested win themes.
Compliance Document Review
ISO 27001 audit reports, SOC 2 attestations, PCI assessments — extracts findings, maps them to your control framework, generates the gap analysis.
Board Pack Pre-read
100-page board pack reduced to a 4-page director's brief: key decisions, financial highlights, risks, and the 6 questions a director should ask.
Document types we handle
- Contracts — MSAs, SOWs, NDAs, employment agreements, vendor terms
- Financial documents — annual reports, due diligence packs, valuations
- Compliance & audit — SOC 2, ISO 27001, PCI DSS, internal audit reports
- Tender documents — government RFTs, RFPs, RFIs, EOIs
- Insurance — policy wordings, claims documents, cover comparison
- Real estate — leases, contracts of sale, body corporate by-laws, strata reports
Accuracy and limitations
On contract review specifically, our deployments hit 94–97% accuracy on identifying named clauses (limitation of liability, indemnity, termination, IP) when measured against partner-reviewed ground truth. The 3–6% miss rate is concentrated in unusual clause structures or hand-amended scanned PDFs. Critically, the GPT reports its own confidence — flagging 'I'm not sure if this is the limitation clause; please review' rather than confidently producing wrong output.
The lawyer/accountant/director still owns the review. The GPT shortlist is 'here are the 12 things I think you should look at'. It doesn't approve contracts, sign off audit findings, or replace fiduciary judgement. Used right, it triples the throughput of the human reviewer.
Frequently asked questions
How does it handle scanned PDFs and handwritten amendments?
Reasonably well, with caveats. We use OCR (typically Azure Document Intelligence or AWS Textract) for scanned documents. Handwritten amendments work for clear handwriting; messy handwriting gets flagged 'human review required' rather than guessed. Most legal teams scan-and-amend less than 15% of documents these days, so this edge case is shrinking.
Can it handle 1,000-page documents?
Yes — we use chunked retrieval rather than stuffing the whole document into a single prompt. The trade-off is that asking 'find me everything about indemnification' across 1,000 pages will surface 90–95% of relevant passages on the first pass. We deploy a recall-test mode where you can verify it didn't miss anything important on a sampled basis.
Does this replace contract management software like Ironclad or DocuSign CLM?
No — different problems. CLM platforms manage the workflow (drafting, redlining, approval, signature). A document analysis GPT is the brain that helps you read documents fast. They work well together; we've integrated GPTs into Ironclad workflows for several legal teams.
What about confidentiality with sensitive documents?
Same architecture as our other deployments: Australian-region cloud, no training on your data, encrypted at rest and in transit, audit log of every document processed. We sign NDAs covering the GPT's access to confidential material.
Ready to build your custom GPT?
Get a free 30-minute scoping call. We'll map your use case, data sources, and ROI before you commit.
Start the Conversation