⚠️ Disclaimer: this is a review for understanding, NOT legal advice. For important contracts — see a lawyer. Personal data / trade secrets are best run through a local model (the "private" step).
Ready-made: grasp a contract and its buried risks in minutes. From an impenetrable text you get a plain-words summary, a list of red flags and questions for a lawyer.
What you get
- The gist in plain words: who, what, to whom, when and for how much.
- A map of obligations, deadlines, penalties and termination terms.
- A list of red flags and one-sided clauses.
- A short 5–7 point summary plus questions for the lawyer.
What you need
Tools from the base: ChatGPT/Claude (review), the "Contract review clause-by-clause + risks" prompt, Ollama (private local review for sensitive docs), the "Local / private" combo (how to run offline). Any OCR for scans. Free with ChatGPT/Claude free for non-sensitive docs; local with Ollama for confidential ones.
Step by step
- Text from the file — extract the text (PDF/DOCX → text); OCR for a scan. OK when the text is clean with paragraphs intact.
- Clause-by-clause review — the "Contract review clause-by-clause + risks" prompt in ChatGPT/Claude: gist, each party's obligations, deadlines/amounts, penalties, termination, red flags, what to clarify. OK when each section is filled and risks point to specific clauses.
- Private (optional) — run the same review on Ollama locally so sensitive text never leaves for the cloud (see the "Local / private" combo). OK when it runs on localhost with the network off.
- Clarifying questions — ask the AI to list only the disputable/risky clauses and phrase a concrete question to the other party for each. OK when you get a short list with ready questions.
- Wrap-up — ask for a 5–7 point summary plus questions for the lawyer. The final decision and signature are on you. OK when the summary fits the screen.
Free vs fast (paid)
Free/private: ChatGPT/Claude free or local Ollama, split big docs into parts. Faster/sharper: a top paid model with a larger context window for big or critical documents. For NDA/secret material — always local (Ollama).
Common problems
Doc too large → split into sections, then ask for an overall summary, or use a large-context model. OCR garbage → higher-DPI scan/better OCR, verify key numbers manually. Vague "all fine" output → demand references to specific clauses and ask the clarifying questions. Confidentiality worry → never upload secrets, use a local model. Tempted to decide "by AI" → remember the disclaimer; a lawyer has the final word.
Time & money
10–30 min per document (extract → review → summary); a big contract with splitting up to an hour. Cost: free (ChatGPT/Claude free or local Ollama) or a little per token for a top model on large/important docs. It speeds up understanding but does NOT replace a lawyer.
🧩 The same conveyor fits any long document (offer, policy, spec): extract → review with the prompt → privately if needed → summary.