Claude for Lawyers
ReviewIntermediate15–30 min

Triage an Incoming Contract Against Your Standard Positions

Give Claude your standard contract positions, paste an incoming agreement, and get a severity-ranked list of every deviation with risks, proposed fixes, and a plain-English note for the business client.

Contract LawCorporateEmployment Law

When to use this

An incoming NDA, MSA, or vendor agreement lands in your inbox and the business team wants it signed yesterday. You need a fast first pass that catches the off-market clauses, ranks them by how much they actually matter, and gives you fixes you can drop into a redline. This is the contract-intake workflow most small legal teams reach for first.

Contract intake review is the single most-embedded AI workflow in small legal teams and in-house shops, and it is easy to see why: NDAs and vendor agreements arrive constantly, they mostly repeat the same handful of issues, and the slow part is comparing each one against the positions you already know you will take. Teams using AI for first-pass review routinely report cutting contract review time by roughly half.

The trick is giving Claude your standard positions up front. Your "playbook" does not need any special software — it can be a one-page list of acceptable ranges and escalation triggers for your key clauses (indemnity, limitation of liability, term and termination, governing law, confidentiality scope, non-solicit) that you paste in, or a saved Claude Project you reuse. Once Claude knows your guardrails, it can read an incoming contract and tell you exactly where it strays.

Everything Claude produces here is first-pass triage, not legal advice and not a substitute for your judgment. Treat the flags as a checklist to verify clause by clause against the actual text. You are responsible for every position taken and every word that goes into the redline.

The Workflow

  1. Load your standard positions (your playbook)

    Start the conversation by teaching Claude your guardrails. List your key clauses with an acceptable range and an escalation trigger for each. This one-pager is your 'playbook' — no special software required; paste it here or keep it as a saved Claude Project you reuse for every intake.

    Prompt
    You are helping me, a lawyer at a [SOLO / SMALL FIRM / SMALL IN-HOUSE TEAM], triage incoming contracts against my firm's standard positions. Everything you produce is first-pass triage that I will verify clause by clause — not final advice.
    
    Here is my standard positions playbook. For each clause: ACCEPTABLE = what I can sign without escalation; ESCALATE = what triggers negotiation or sign-off.
    
    - Limitation of liability: ACCEPTABLE = cap at [12 months' fees]; mutual. ESCALATE = uncapped, super-cap, or one-sided.
    - Indemnification: ACCEPTABLE = mutual, limited to third-party claims. ESCALATE = one-sided or covering first-party/consequential losses.
    - Term & termination: ACCEPTABLE = termination for convenience on [30 days]' notice. ESCALATE = auto-renew with no exit, or for-cause only.
    - Governing law / venue: ACCEPTABLE = [STATE]. ESCALATE = anything else.
    - Confidentiality: ACCEPTABLE = mutual, [2–3]-year term, standard carve-outs (public, independently developed, required by law). ESCALATE = perpetual, one-sided, or no carve-outs.
    - Non-solicit / non-compete: ACCEPTABLE = [none / mutual no-hire of named individuals]. ESCALATE = any non-compete or broad non-solicit.
    - [ADD YOUR OWN CLAUSES AND RANGES]
    
    Confirm you understand these positions. Do not analyze anything yet — I will paste the incoming contract in my next message.

    What you get: Claude restates your playbook in its own words, confirms the acceptable-vs-escalate logic for each clause, and waits for the contract. If anything is ambiguous it will ask before proceeding.

  2. Paste the incoming contract

    Drop in the full text of the agreement you received. Tell Claude what kind of contract it is and which side you represent, so it reads the risk allocation from your client's perspective.

    Prompt
    Here is the incoming contract. It is a [NDA / MSA / VENDOR AGREEMENT / SERVICES AGREEMENT] and I represent the [DISCLOSING PARTY / RECEIVING PARTY / CUSTOMER / VENDOR / COMPANY]. Read it in full against the playbook above. Do not summarize yet — just confirm you have read it and tell me how many clauses you intend to flag as deviations.
    
    [PASTE FULL CONTRACT TEXT]

    What you get: Claude confirms it has read the document, identifies the contract type and your side, and gives a quick count of how many clauses deviate from your standard positions before producing the full analysis.

  3. Flag every deviation, ranked by severity

    Ask Claude to produce the core triage table: each clause that strays from your standard positions, ranked by severity, with the specific risk to your client and a concrete proposed fix you could put into a redline.

    Prompt
    Now produce the triage. For EVERY clause that deviates from my standard positions, give me a row with:
    1. Clause name and section number
    2. Severity: HIGH (escalation trigger / deal-impacting), MEDIUM (off-market, negotiate), or LOW (cosmetic / nice-to-have)
    3. What it currently says (short quote or paraphrase)
    4. The risk to my client in one or two plain sentences
    5. A proposed fix — the specific change or replacement language I could redline
    
    Sort HIGH first, then MEDIUM, then LOW. After the table, list any clauses that are MISSING that my playbook expects (e.g., no termination-for-convenience right). Flag anything ambiguous as 'VERIFY' rather than guessing. Remember this is first-pass triage for me to confirm against the actual text.

    What you get: A severity-sorted table of deviations — each with section, current language, plain-English risk, and a concrete proposed fix — followed by a list of missing clauses and any VERIFY items where the text was ambiguous.

  4. Draft the redline summary and the client note

    Turn the triage into two deliverables in one go: an internal redline-ready summary you can work from, and a short plain-English note for the business client explaining what needs to change and why.

    Prompt
    Using the triage above, produce two things:
    
    A) REDLINE-READY SUMMARY (for me): a numbered list of the changes I will mark up, each as 'Section X — [change] — [one-line rationale]', ordered by severity. Keep it tight enough to work from while editing the document.
    
    B) NOTE TO THE BUSINESS CLIENT (plain English, no legalese): 4–6 sentences explaining the handful of issues that actually matter, what I'm asking the other side to change, and which points are must-haves vs. nice-to-haves. Frame it so a non-lawyer can decide how hard to push. Do not overstate certainty — note that I'm still confirming the details.

    What you get: Two clean outputs: an internal change list keyed to section numbers with one-line rationales, and a short, non-legalese client note that separates must-haves from nice-to-haves and stays appropriately hedged.

  5. Verify every flag against the actual text

    Before you send anything, read the contract yourself against Claude's output. Confirm each flagged section number and quote is accurate, check the VERIFY items, make sure no clause was missed, and adjust severity and fixes to fit the specific deal and client. The triage is your starting checklist — the positions you take are yours.

    What you get: A reviewed, corrected triage you trust: every section reference and quote confirmed against the document, ambiguous items resolved, severities and fixes adjusted to the real deal, and any AI miss or hallucinated clause caught before it reaches the redline or the client.

Example Output

Illustrative example — names, facts, and figures are fictional.

CONTRACT TRIAGE — Mutual NDA (received from Northwind Logistics)
Contract type: NDA · I represent: Disclosing Party · Deviations flagged: 4 (1 HIGH, 2 MEDIUM, 1 LOW)

[HIGH] Confidentiality Term — §5
Current: Obligations expire 1 year after disclosure.
Risk: Our roadmap and pricing stay sensitive well past 12 months; a 1-year tail leaves later-disclosed material unprotected.
Fix: Extend to 3 years from disclosure; carve out trade secrets to remain protected as long as they qualify.

[MEDIUM] Definition of Confidential Information — §1
Current: Covers only information marked "Confidential."
Risk: Oral disclosures and unmarked documents fall outside the definition.
Fix: Add oral/visual disclosures and a catch-all for info a reasonable person would treat as confidential.

[MEDIUM] Governing Law — §11
Current: Delaware.
Risk: Off our standard; adds out-of-state counsel cost on any dispute.
Fix: Change to [Home State]; fall back to mutual home-state filing if resisted.

[LOW] Notice — §12
Current: Notice by certified mail only.
Fix: Add email notice with confirmation. Cosmetic; do not hold up signature.

MISSING: No injunctive-relief clause — add a standard equitable-remedies provision. VERIFY: §7 return-vs-destroy language is ambiguous on backups.

NOTE TO CLIENT: This NDA is close to signable, with one issue that matters: confidentiality only lasts a year, which is too short for our roadmap — I'm asking for three years plus trade-secret protection. Two smaller cleanups (what counts as confidential, and governing law) are worth raising but not dealbreakers. I'm still confirming a couple of points in the return-of-information section before we send it back.

Tips

  • Save your standard positions as a Claude Project so every new intake starts pre-loaded — paste the contract and go, no re-teaching the playbook each time.
  • Tighten your ranges over time: when you accept or reject a deviation in real life, fold that decision back into the playbook so the triage tracks how you actually negotiate.
  • Always tell Claude which side you represent. The same clause can be a HIGH risk for the receiving party and a non-issue for the disclosing party.
  • Treat every section number and quoted clause as something to confirm against the document. AI can mis-cite a section or flag a clause that isn't there — verify before it reaches a redline or the client.

A note on confidentiality

Run contract intake only in Claude Team or Enterprise, where your inputs are not used to train models, and keep client and deal terms out of consumer tiers — using a confidential client agreement with a tool that trains on inputs can implicate your duty of confidentiality under ABA Formal Op. 512 and Model Rule 1.6.

Get New Workflows Like This Every Week

Join the free Claude for Lawyers newsletter — practical workflows, prompts, and guides for everyday practice.

Free weekly newsletter. Unsubscribe anytime.