AI Contract Review: How to Use Claude for Due Diligence

Claude AI Can Review an Entire Contract in One Pass

Claude can review a complete contract in a single pass. With a context window of up to 1 million tokens on current models, attorneys upload an entire merger agreement, SaaS contract, or commercial lease and get a structured risk analysis in minutes. Most contracts fit in a single conversation without file splits (though you should still check that nothing was overlooked), and in our experience a disciplined first-pass review of a standard commercial agreement takes about 10 minutes of prompting plus your professional judgment on what it finds. Very large agreements with exhibits can exceed smaller models' context windows (Haiku 4.5, for example, has 200K tokens).

That last clause is the operating principle for everything below. Claude eliminates the mechanical scanning that precedes judgment, not judgment itself. On a commercial lease or vendor agreement, the first pass can shrink considerably. Claude finds the issues; you decide what to do about them.

What Claude Does Well on Contracts

Claude is strongest at the pattern-recognition layer of contract review: spotting one-sided provisions, extracting commercial terms into a clean summary, comparing two drafts clause by clause, and flagging what a contract of this type normally contains but this draft omits. It handles every contract type attorneys encounter:

• M&A agreements: APAs, SPAs, merger agreements with complex rep and warranty sections
• Commercial contracts: SaaS agreements, MSAs, vendor contracts, licensing deals
• Real estate: PSAs, commercial leases, title documents (see our guide to Claude AI for real estate lawyers)
• Employment: Executive agreements, non-competes, severance packages
• Finance: Loan agreements, security agreements, intercreditor agreements

How to Structure a Contract Review Prompt

Use the CRAFT prompting framework. "Review this contract" produces generic output. Structure your prompt instead:

Example prompt: "Context: We represent the buyer in a $25M acquisition of a logistics company. Delaware law governs. Role: Senior M&A attorney reviewing the seller's first draft APA. Ask: Identify every indemnification provision, survival period, and basket/cap structure. Flag deviations from market-standard terms for this deal size. Format: Table with Section Number, Provision Summary, Market Standard Comparison, Risk Level, Recommended Redline. Tone: Analytical, for a partner review memo."

The 10-Minute Contract Review Workflow

For standard commercial agreements, this five-step workflow compresses the mechanical first pass into about 10 minutes in our experience. The step times below are illustrative estimates; long or unusual agreements take longer. Run the steps in order; each builds on the one before it.

Step 1: Red Flag Scan (3 Minutes)

Start with a broad scan to identify the biggest risks. This is where Claude saves the most time.

Review this contract and identify red flags. I represent the [BUYER/TENANT/LICENSEE].

Contract type: [e.g., SaaS agreement, commercial lease, supply contract]
Deal value: [APPROXIMATE]

Flag:
1. One-sided indemnification
2. Unlimited liability exposure
3. Auto-renewal traps
4. IP assignment beyond what's needed
5. Non-compete or exclusivity restrictions
6. Missing termination rights
7. Unreasonable cure periods
8. Governing law or venue that disadvantages my client

For each red flag: quote the language, explain the risk, rate HIGH/MEDIUM/LOW, and suggest revision.

Contract:
[PASTE FULL CONTRACT]

This produces a prioritized issues list you can work from. For a ready-to-use version, see our Contract Red Flag Scanner prompt.

Step 2: Missing Provisions Check (2 Minutes)

The most dangerous contract issues are the ones that are not on the page.

This is a [CONTRACT TYPE] between [PARTY A] and [PARTY B].

Identify standard provisions that are MISSING from this contract:
1. Limitation of liability / damages cap
2. Representations and warranties
3. Confidentiality
4. Force majeure
5. Assignment restrictions
6. Insurance requirements
7. Dispute resolution mechanism
8. Data protection / privacy
9. Compliance with laws
10. Survival clause

For each missing provision, explain why it matters for this type of agreement and suggest standard language.

Contract:
[PASTE CONTRACT]

Step 3: Key Terms Extraction (2 Minutes)

Before you negotiate, you need a clean picture of what you are working with.

Extract the key commercial terms from this contract:

1. Price/payment terms and schedule
2. Term and renewal provisions
3. Termination rights and notice periods
4. Service levels or performance standards (SaaS/service contracts)
5. Liability caps and exclusions
6. Insurance minimums
7. IP ownership and licensing terms
8. Non-compete and exclusivity
9. Governing law and dispute resolution
10. Material conditions or closing requirements

Present as a term sheet I can share with the client.

Contract:
[PASTE CONTRACT]

Step 4: Clause Comparison (2 Minutes)

If you have a preferred template, Claude compares their draft to yours and identifies departures.

Compare these two versions of the [INDEMNIFICATION/LIABILITY/TERMINATION] clause:

Our standard:
[PASTE YOUR TEMPLATE CLAUSE]

Their draft:
[PASTE THEIR CLAUSE]

Identify:
1. Substantive differences in obligations or rights
2. Differences in defined terms that change meaning
3. Which version is more favorable to my client
4. Specific edits to bring their version closer to ours

Step 5: Generate Markup Comments (1 Minute)

After identifying issues, Claude drafts the actual markup comments you would add to the redline.

Based on the red flags identified, draft markup comments for a redline:

For each issue:
1. Quote the contract language
2. Write a brief comment explaining the concern (as you'd write it in a Word comment)
3. Suggest alternative language

Format the output so I can copy each comment directly into the document.

Issues:
[PASTE THE RED FLAG LIST FROM STEP 1]

When the 10-Minute Workflow Works Best

This approach is ideal for standard commercial contracts (vendor, SaaS, services, supply), commercial and residential leases, NDAs and confidentiality agreements, and employment agreements. For complex M&A agreements or heavily negotiated deals, use it as a first pass, then do a traditional section-by-section review of the critical provisions. Browse more tools on our contract law practice area page, including the Lease Review Analyzer.

Scaling Up: A Due Diligence Workflow

For larger transactions, chain multiple Claude conversations into a complete due diligence workflow:

• Pass 1: Risk identification. Upload the full agreement. Request a risk matrix by severity.
• Pass 2: Clause analysis. Drill into high-risk provisions with targeted questions.
• Pass 3: Redline generation. Draft alternative language for each flagged provision.
• Pass 4: Comparison. Upload the original and your client's standard terms. Get a gap analysis.

Verification: What Claude Cannot Do in Contract Review

AI contract review has clear limits, and each one maps to a verification habit:

• No legal advice: Claude identifies risks and drafts language. The attorney makes every strategic decision.
• Knowledge cutoff: Claude's built-in knowledge stops at a training cutoff that varies by model, so it may not reflect recent changes. Verify anything time-sensitive, such as a recent regulatory change, against primary sources.
• No context beyond the document: Claude knows nothing about prior dealings, business relationships, or negotiation dynamics unless you provide them.
• Unverified citations: If Claude references cases, statutes, or "market standard" benchmarks, verify them before relying on them. The prompts above ask for "standard language" and market-standard comparisons, and Claude has no deal database behind those answers unless you supply precedents or a market survey. Treat every output like a junior associate's first draft: useful, but reviewed before it reaches a client.

Confidentiality Considerations

Before uploading client contracts, confirm you are on a commercial plan. Claude's Team and Enterprise plans do not train on your documents by default (see Anthropic's model training policy), and the Enterprise plan adds custom data retention controls. One exception to know: under the same policy, if a user submits thumbs up/down feedback, Anthropic may use that conversation for training, and organization owners can turn this off with the Rate chats setting. Choosing the right plan is a starting point, not the whole answer: ABA Formal Opinion 512 (July 29, 2024) still expects you to understand the tool's risks, protect client information under Rule 1.6, and in some situations obtain informed client consent. See our guide to the ABA ethics rules for AI use. For plan details, see our Claude pricing breakdown for lawyers.