AI Document Reviewer

Analyze

Contract-ish prose, specs, or policies — risk-flag ambiguous clauses, define terms of art, and produce a reviewer checklist without pretending to be a lawyer.

Triage, do not certify

Useful before expensive humans — never instead of them.

Paste vendor MSAs, offer letters, privacy policies, technical RFCs, or other prose. You get structured risk flags with quoted snippets from the source text, inconsistent or undefined terms of art, missing standard clauses, and a prepared question list for counsel or the counterparty. The model is instructed to scream NOT LEGAL ADVICE because it cannot hold malpractice insurance — its job is to make you smarter and faster before you talk to someone who can. Output is shaped for forwarding to legal, not for replacing them.

How to brief a useful first-pass review

Five inputs that turn AI from generic to specific.

1. Paste the clause, section, or full short document — redact party names if confidentiality matters.
2. Pick the document type lens so the model knows what standard clauses to expect (or flag as missing).
3. State your reviewer goal concretely ("flag uncapped liability" beats "check this contract").
4. Add a jurisdiction hint when geography matters; many enforceability questions are state-specific.
5. Treat the output as questions for counsel, not answers — the disclaimer is structural, not decorative.

Document types it understands

Each lens drives different missing-clause expectations.

Vendor SaaS

MSA / order forms

Liability caps, auto-renewal, data residency, security commitments — the usual suspects.

Employment

Offers + non-competes

Equity terms, termination clauses, IP assignment, restrictive covenants by jurisdiction.

Privacy policy

GDPR / CCPA shape

Required disclosures by jurisdiction, ambiguous data-sharing language, consent-flow gaps.

Technical spec

RFC / design doc

RFC 2119 keyword audit (MUST/SHOULD/MAY), ambiguous requirements, missing edge cases.

Other prose

General review

Generic clarity and consistency check when the document doesn't fit a standard category.

Best for

Pre-counsel reviews where structure beats vibes.

• Vendor procurement reviews where engineers and PMs read MSAs before legal does
• Offer letter triage when employees want a sanity check before signing
• Privacy policy audits during product launches or regional expansions
• Technical RFC reviews where ambiguous SHOULDs become production incidents
• NDA reviews for design partners and beta customers
• DPA (data processing agreement) cross-checks against your actual data handling
• Pre-acquisition diligence reads on prospect contracts and customer agreements

Why quoted snippets matter more than summaries

Generic summarizers paraphrase risks into oblivion.

A risk flag without the original text is useless to legal — they have to find the clause anyway, and your summary may have softened or sharpened the language unintentionally. This template forces the model to quote the exact snippet that triggered each flag, so counsel can read both your interpretation and the source in the same view. Defined-term inconsistencies show both definitions inline. Missing-clause flags are explicit about what was expected for the document type. The output is a triage document, not a translation — trust is preserved by traceability.

Pro tips for faster legal handoffs

Habits that compound across procurement and legal cycles.

1. Always include the jurisdiction hint when reviewing employment, privacy, or regulated industry documents.
2. Forward the output verbatim to counsel along with the source PDF; the question list saves billable hours.
3. When the model surfaces missing standard clauses, verify by re-reading the source — sometimes they are present but worded oddly.
4. Use the technical-spec lens for any RFC or design doc with normative language (MUST, SHALL, etc.).
5. Pair with the AI URL Analyzer to spot-check linked policy pages or DPAs referenced in the document.
6. Maintain a checklist of recurring red flags (uncapped liability, IP assignment scope) so reviews stay consistent across team members.

Document Reviewer FAQ

Can I sign or commit based on this output?

No. This is triage assistance; licensed counsel signs. The output is structured to make your conversation with counsel faster, not to replace them.

Is this legal advice?

No — explicitly. The system prompt requires the model to disclaim and refer ambiguity to qualified attorneys. Treat every flag as a question, not an answer.

Will it invent clauses or misquote the source?

It is instructed to quote snippets verbatim from your pasted text. Verify any quoted clause against the source; if the model got it wrong, re-paste with cleaner formatting.

Does it understand my jurisdiction?

It uses jurisdiction hints to bias toward relevant frameworks (GDPR, CCPA, US state non-compete law). It does not provide jurisdiction-specific legal opinions — that is counsel's job.

Can it review entire contracts at once?

Better to review section-by-section for a long contract — focused passes produce sharper flags than one giant scan.

Which models power it?

Default reasoning-capable models for nuance. Switch to deeper models for complex commercial agreements or multi-jurisdiction policies.

How do I keep sensitive content private?

Redact party names and confidential numbers before pasting. The tool only sees what you upload, but the underlying model providers may have their own data handling — review your workspace settings.

Faster diligence loops

Ask smarter questions earlier.

Shrink the unknown-unknown surface before hour one on Zoom with legal. The output is structured to make counsel more effective, not to bypass them — and the time saved compounds across every contract your team touches.