Quick answer: AI can review a contract quickly against a standard you give it: extracting clauses, comparing them with your playbook, flagging deviations and missing clauses, and drafting a risk table. It cannot judge which deviations matter for this client and deal, spot documents that were never uploaded, or confirm stamping and execution. A lawyer checks every flag in the document itself.
Key takeaways
- AI review is only as good as the standard behind it, so write a playbook with standard positions, fallbacks, red lines and approvers before you automate anything.
- Load the whole contract, including schedules and documents it incorporates, because AI cannot flag what it was never given.
- Scanned copies cause quiet errors in numbers and dates, so check every figure in a liability cap or notice period against the image.
- Contracts often carry personal data protected under the DPDP Act, 2023, so they belong only in tools approved for confidential data.
- AI for Legal's 8-hour contract review and due diligence module ends with a lab that turns a sample commercial agreement into a playbook-based risk table.
AI can review a contract quickly against a standard you give it. It extracts clauses, compares them with your playbook, flags deviations and missing clauses, and drafts a risk table. It cannot tell you whether a deviation matters for this client and this deal, what the other side is really trying to achieve, or whether the document was properly stamped and executed. Treat it as a fast first reader whose work you check line by line, not as the reviewer.
What can AI do in contract review, and what can't it?
| Task | Where AI is useful | What still needs a lawyer |
|---|---|---|
| Clause extraction | Pulling indemnity, limitation of liability, termination, assignment, confidentiality and dispute resolution clauses into one table | Confirming nothing was missed, including clauses buried in schedules |
| Playbook comparison | Flagging every place the draft departs from your standard positions | Deciding which departures are acceptable in this deal |
| Missing clauses | Listing what your playbook expects and the draft lacks | Judging whether an omission is deliberate, harmless or serious |
| Defined terms | Listing each definition and where it is used | Seeing how a wide or narrow definition shifts risk elsewhere |
| Version comparison | Summarising what changed between two drafts | Reading the actual redline before agreeing to anything |
| Business summary | A one-page note for a non-lawyer | Making sure the note does not understate a risk |
| Stamping and execution | A checklist of points to verify | Applying the law for the instrument and the state |
The pattern is consistent. AI is quick at finding and sorting text. The decisions about meaning, materiality and negotiation stay with the person who will answer for the advice.
Why does the playbook matter more than the tool?
Without a playbook, an AI tool reviews against its own idea of a reasonable contract, which may be drawn from another jurisdiction, another industry or the other side's point of view. With one, it reviews against yours. A usable playbook needs four things for each clause: your standard position, the fallback you will accept, the red line you will not cross, and who can approve a departure.
An illustrative extract, not a recommendation:
| Clause | Standard position | Acceptable fallback | Red line | Who approves a fallback |
|---|---|---|---|---|
| Limitation of liability | Capped at fees paid in the previous 12 months, with carve-outs for fraud and breach of confidentiality | A higher multiple of fees | Uncapped liability on our side | Head of legal |
| Termination for convenience | Available to both parties on 90 days' notice | 60 days | Not available to us | Legal with the business owner |
| Assignment | Only with consent, except to group companies | Consent not to be unreasonably withheld | Free assignment by the counterparty | Legal |
| Dispute resolution | Arbitration, with the seat named | Courts at a named city | A foreign seat for a purely domestic contract | Head of legal |
Writing the playbook is slower than running the tool, and it is the part that makes the tool worth running. It also makes reviews consistent across a team, which is where most of the time saving comes from.
What does a sound AI review workflow look like?
- Clear the tool and the terms. Use only a tool approved for confidential data, and check that the NDA, engagement letter or data-room rules allow it.
- Load the complete document. The main body, every schedule and annexure, and any policy or standard terms it incorporates by reference. If a referenced document is missing, the AI will not know it exists.
- Check the text quality. Scanned PDFs can produce garbled text. A misread digit in a liability cap or a notice period is exactly the error that slips through, so compare figures with the image.
- Run extraction and comparison. Ask for the clause number against every finding, so each flag can be traced.
- Review each flag in the contract, not in the summary. Read the clause, the definitions it uses and anything it cross-refers to.
- Do a second pass for what the playbook does not cover. Unusual clauses, one-sided definitions, deemed acceptance provisions and cross-references that point to the wrong clause after renumbering.
- Write the risk table and recommendations yourself. Use the language your client or business team understands, and record which points you checked.
Where does AI contract review go wrong?
In predictable places, which is useful, because you can check them every time.
- Definitions that move risk. If "Losses" is defined to include indirect loss, an exclusion elsewhere may mean less than it says. AI often reviews clauses one at a time and misses the interaction.
- Cross-references after edits. A clause that refers to "clause 14.2" may point at the wrong text once the draft has been renumbered.
- Seat and venue. Summaries sometimes treat the seat of arbitration and the venue of hearings as the same thing. They are not, and the difference matters.
- "Standard" language. A tool may call a clause standard when it is standard somewhere else, or standard for the other side.
- Redaction that removes the point. Stripping names and numbers to protect confidentiality can also strip the commercial terms you needed reviewed. Redaction is not a substitute for an approved tool.
- Personal data. Employment contracts, customer agreements and loan documents carry details of individuals, protected under the Digital Personal Data Protection Act, 2023.
How do you know AI review is saving time?
Measure it before you claim it. Pick one contract type you see often, record how long a manual first review takes on a few recent examples, then review similar contracts with AI support and record the time again, along with the issues found and anything a second reviewer caught that the first pass missed. If the time falls and the misses do not rise, the workflow is working. The capstone of AI for Legal asks for exactly this kind of measured before and after on a workflow from your own practice area.
What does AI for Legal teach about contracts?
The AI for Legal professional certificate gives contract work its own module.
- Contract Review & Due Diligence (8 hours). Clause extraction and risk tables, comparing drafts against a negotiation playbook, flagging deviations from standard terms and missing clauses, due-diligence checklists and data rooms, and stamp duty and execution checklists to verify. Lab: review a sample commercial agreement against a playbook and produce a risk table. Tools: Claude, ChatGPT and NotebookLM.
- Drafting & Document Automation (6 hours) adds clause libraries, reusable templates, version control and red-flag review.
The programme lists a Contract Manager (AI-enabled) role at an indicative ₹8 to 18 LPA; that is a range, not a promise. The certificate runs 8 weeks and 48 hours on weekday evenings and Sunday mornings, IST, with recordings, and the next cohort starts on 10 January 2027. The fee is ₹21,999 including GST, or ₹3,667 a month for six months on 0% interest EMI, subject to approval; see fees for scholarships of up to 25% and the refund window. Labs use mock documents only, and the programme teaches skills for your own practice or team, not client acquisition, in line with Bar Council of India rules on advertising by advocates.
For contract work inside a transaction, read AI for corporate lawyers: contracts and diligence. For a team that reviews the same agreements every week, see AI for in-house counsel and legal teams.
What to do next
- Write a four-clause playbook for one contract type you review often, then test an approved tool against a mock or template agreement and note every flag it gets wrong.
- Attend the free masterclass AI for Lawyers: Faster Research and Drafting, Without the Risk, a one-hour live session at 7 pm IST on 21 November 2026.
- Read the contract review module, its lab and the refund policy before you decide on the AI for Legal certificate.