What AI Is Genuinely Good At
Used correctly, AI turns contract review from a reading marathon into a structured triage. The tasks it does well are the mechanical ones:
🔎 Clause extraction
Find every indemnity, cap, assignment, and termination provision across hundreds of pages — reliably, in minutes instead of hours.
⚖️ Playbook comparison
Compare contract language against your firm's preferred clauses and flag where the counterparty's version deviates.
🚩 Risk flagging
Surface unusual terms — auto-renewals, one-sided termination, uncapped liability — so a lawyer decides, not the machine.
📄 Summaries
Produce deal summaries and obligation trackers that give a partner the lay of the land before they read the document.
🔁 Consistency checks
Compare definitions, dates, and cross-references inside one contract or across a portfolio of agreements.
Notice what these have in common: they're retrieval and organization tasks. The AI is a tireless associate that reads everything and lays out what it found — which is exactly the part of review that eats billable hours.
What AI Cannot Do
The honest list is shorter but matters more:
- Exercise judgment. A liability cap of 12× monthly fees isn't "good" or "bad" — it depends on your client's risk tolerance, leverage, and deal context. AI doesn't know any of that.
- Know your client. The model doesn't know the client's history with the counterparty, the negotiation posture, or the business reason for a clause.
- Negotiate. AI can draft fallback language. It can't read the room or trade terms in real time.
- Guarantee accuracy. General-purpose models hallucinate — they invent clauses, misstate standards, and fabricate citations. This is not hypothetical; it's the defining failure mode of the technology.
⚠️ The verification rule: if the AI can't show you the exact passage it relied on, its answer isn't a finding — it's a guess. Every output should be verifiable in the source document before it shapes advice.
The Two Risks That Change Everything
Two risks dominate every real-world deployment, and both have 2026 case law behind them:
Confidentiality. A contract is a bundle of client confidences. Paste it into a public chatbot and it can be collected, trained on, or disclosed under the provider's terms — the exact pattern courts found defeated privilege in United States v. Heppner (S.D.N.Y. 2026). ABA Formal Opinion 512 makes the standard clear: protect Rule 1.6 confidentiality or don't use the tool.
Competence. Opinion 512 also requires you to understand the tool's limitations and verify its output. A hallucinated clause summary that shapes a negotiation is a competence failure you own, not the model.
The architectural answer to both: local AI. When documents are indexed and answered on your own hardware — with citations back to the exact passage — nothing leaves your control and every output is checkable. See how a full local pipeline is built in Lawyer Assistant: A Privacy-First Legal AI Built on a Local RAG Pipeline.
How to Start Using AI for Contract Review
- Start with low-stakes documents. Pilot on form agreements and templates, not the deal that's closing Friday.
- Demand citations. A tool that can't show its source passage isn't production-ready.
- Keep data local. Prefer tools that run on your hardware — no third-party exposure, no training-pipeline risk.
- Build your playbook. Feed the AI your firm's preferred clauses so comparisons run against your standards.
- Review everything. AI triages; lawyers decide. Every flagged item and every summary crosses a human desk before it reaches a client.
- Write the policy. Approved tools, prohibited tools, review procedures — under Rule 5.1, leadership is responsible.
Frequently Asked Questions (FAQ)
Can AI review contracts?
Yes, in a limited sense. AI can extract clauses, compare language against a playbook, flag deviations and risky terms, and summarize sections. What it cannot do is exercise legal judgment: it doesn't know your client's risk tolerance, the deal context, or the negotiation strategy.
Is AI contract review accurate?
Accuracy depends on the system. General-purpose chatbots hallucinate and can miss clauses entirely. Retrieval-based systems that ground answers in the actual document with citations are far more reliable — but every output still needs a lawyer's review.
Is it ethical to use AI for contract review?
Yes, if you comply with your professional obligations. ABA Formal Opinion 512 requires competence (verify output), confidentiality (Rule 1.6), communication with the client, and firm policies. The safest architecture keeps client documents on your own machine.
What should lawyers never let AI do in contract review?
Never let AI make the final call on risk, never rely on it without verifying citations, never paste client contracts into public chatbots, and never use it without a firm policy and, where required, client informed consent.
How should a lawyer start using AI for contract review?
Start with low-stakes documents, use a tool that cites its sources, keep data local where possible, review every output, and build up a playbook of your firm's preferred clauses so the AI compares against your standards rather than generic ones.
⚖️ Need this built for your firm?
I design and deploy privacy-first local AI systems — private RAG, cited answers, on-premise LLMs for legal work. Contact me for a scoping conversation. Or start with the free, open-source Lawyer Assistant.