Why AI changes contract review

A senior lawyer spends between 60 and 90 minutes reviewing a standard service agreement. An M&A contract can take days. AI does not replace legal judgment, but it transforms the process in three concrete ways:

• Triage speed: it identifies in seconds the clauses that require human attention, separating the routine from the critical.
• Consistency: it applies the same review criteria to every contract, without fatigue or variability due to workload.
• Coverage: it detects omissions. A lawyer may overlook a limitation of liability clause buried in an annex. AI does not get tired.

At top-tier firms like Clifford Chance, Allen & Overy, or Freshfields, contract teams already integrate AI tools into their workflow. Not as a substitute for the lawyer. As a first layer of analysis that allows the professional to focus on what truly matters: legal judgment.

Key clause extraction

The first practical use of AI in contract review is the systematic extraction of clauses. Instead of reading 40 pages looking for the indemnification clause, you ask the AI to locate it, extract it, and summarize it.

Critical clauses you should always extract

Regardless of the contract type, there is a set of clauses that every lawyer must review. This is the minimum checklist:

1. Scope and subject matter: what exactly is being contracted, what is excluded.
2. Price and payment terms: amount, deadlines, late payment penalties.
3. Term and renewal: initial period, automatic renewal conditions, notice periods.
4. Limitation of liability: caps, exclusions, required insurance.
5. Indemnification: who indemnifies whom, for what, and within what limits.
6. Confidentiality: scope, duration, exceptions.
7. Intellectual property: ownership of work product, licenses, assignments.
8. Termination: grounds, consequences, transition periods.
9. Governing law and jurisdiction: competent courts, arbitration.
10. Data protection: roles (controller/processor), safeguards, international transfers.

Structured extraction prompt

This prompt works with any AI model (Claude, GPT-4, Gemini). Paste it along with the contract text:

Act as a senior commercial lawyer reviewing a contract.

Extract the following clauses from the attached contract. For each one:
1. Clause number and exact title
2. Resumen in 2-3 lines
3. Risk assessment: LOW / MEDIUM / HIGH
4. Reason for the risk assessment

Clauses to extract:
- Scope and subject matter
- Price and payment terms
- Term and renewal
- Limitation of liability
- Indemnification
- Confidentiality
- Intellectual property
- Termination
- Governing law and jurisdiction
- Data protection

If any clause is NOT present in the contract, flag it
explicitly as "MISSING CLAUSE" with HIGH risk.

Output format: table with columns
[Clause | No. | Resumen | Risk | Reason]

The key is asking it to identify missing clauses. A contract that does not mention limitation of liability is more dangerous than one that has an unfavorable one. Omission is always a red flag.

Extraction by contract type

Adapt the clause list to the contract type:

• SaaS/technology: add SLA, uptime, backup, data portability, exit plan.
• Employment: add non-compete, exclusivity, employee inventions, bonus and clawback.
• Real estate: add encumbrances, easements, conditions precedent, earnest money.
• Distribution: add territory, exclusivity, minimum targets, stock, trademarks.

Identifying contractual risks

Extracting clauses is the first step. The second is assessing the risks they contain. AI can analyze each clause against known risk patterns.

Risk analysis pattern

Analyze the following contract from the perspective of [PARTY: buyer/seller/provider/client].

Identify all contractual risks, organized as follows:

CRITICAL RISKS (require renegotiation):
- Unilateral clauses that excessively favor the other party
- Unlimited or disproportionate liability
- Penalty clauses without reciprocity
- IP assignment without consideration

SIGNIFICANT RISKS (require modification):
- Insufficient notice periods
- Automatic renewal without clear opt-out
- Unfavorable jurisdiction
- Absence of industry-standard clauses

MINOR RISKS (acceptable with review):
- Ambiguities in definitions
- Payment terms not aligned with standard practice
- Incomplete force majeure clauses

For each risk, indicate:
1. Affected clause (number and literal text)
2. Specific risk to my party
3. Suggested alternative wording

Automatic red flags

Configure your prompt so the AI specifically looks for these warning signs:

• Asymmetries: unilateral termination rights, non-reciprocal penalties, information obligations for only one party.
• Unfair terms: in B2C contracts, verify against EU Directive 93/13/EEC on unfair terms in consumer contracts.
• Excessive lock-in: disproportionate commitment periods, unlimited automatic renewals.
• Concealed assignments: subcontracting or contract assignment rights without consent.
• Unilateral modifications: clauses allowing changes to terms without agreement.
• Exotic jurisdictions: arbitration forums in countries with opaque legal systems.

Specialized NDA analysis

Non-disclosure agreements are the most reviewed and worst negotiated contract. Most professionals sign them without reading. This is a mistake. A poorly drafted NDA can block business opportunities or expose critical information.

Critical points of an NDA

Review this NDA from the perspective of [receiving/disclosing party].
Analyze these critical points:

1. DEFINITION OF CONFIDENTIAL INFORMATION
   - Is it too broad? (everything is confidential = nothing is confidential)
   - Does it include standard exclusions? (public information, previously known,
     independently developed, received from third parties)

2. OBLIGATIONS OF THE RECIPIENT
   - Are they reciprocal or unilateral?
   - Do they include return/destruction obligations?
   - Do they allow disclosure to advisors/employees?

3. DURATION
   - NDA period vs confidentiality period (they should be different)
   - Is it reasonable? (2-5 years typical, perpetual = red flag)

4. EXCEPTIONS FOR LEGALLY COMPELLED DISCLOSURE
   - Does it allow disclosure if required by a court/regulator?
   - Does it require notification before disclosure?

5. REMEDIES FOR BREACH
   - Injunctive relief provisions?
   - Penalty clause? Reasonable amount?
   - Damages indemnification?

6. HIDDEN NON-COMPETE
   - Does the NDA restrict doing business with competitors?
   - Does it limit the use of general industry knowledge?

Result: table with [Point | Status | Risk | Recommendation]

The most common NDA error: confidential information definitions so broad they cover publicly available information. This makes the NDA unenforceable. If everything is confidential, a court will not enforce it.

Contract version comparison

In a contract negotiation, parties exchange multiple versions. Manual redlining is tedious and error-prone. AI can compare versions exhaustively.

Comparison prompt

Compare these two contract versions:

VERSION 1 (our proposal):
[paste version 1 text]

VERSION 2 (counterproposal received):
[paste version 2 text]

Generate a change report with:

1. SUBSTANTIVE CHANGES (affect rights or obligations):
   - Modified clause
   - Original text vs new text
   - Legal impact on our party
   - Acceptable / Requires negotiation / Unacceptable

2. MINOR CHANGES (wording, formatting):
   - Resumen list

3. DELETED CLAUSES:
   - Which clauses from V1 do not appear in V2?
   - Implications

4. NEW CLAUSES:
   - Which clauses in V2 did not exist in V1?
   - Impact analysis

5. EXECUTIVE SUMMARY:
   - Counterparty's negotiating position (what are they seeking?)
   - Recommendation: accept / negotiate these points / reject

Section 5 is the most valuable. Asking AI to interpret the other party's negotiating position based on the changes they propose gives you strategic advantage. If they remove the indemnification clause and reduce the limitation of liability, their priority is limiting their financial exposure. You can use that information to negotiate.

Negotiation tracking

For multi-round negotiations, use this prompt at the start of each session:

This is round [N] of the [type] contract negotiation.
Parties: [Party A] and [Party B]. I represent [Party A/B].

Points already agreed in previous rounds:
[list]

Points still under negotiation:
[list]

Analyze the new version received. Focus only on the pending
points. Ignore the ones already agreed. For each pending point:
- Have they moved position? In what direction?
- What is the distance between our positions?
- Suggested counterproposal

Prompt templates for contract review

These are tested prompts you can use directly in your daily practice:

Quick review (5 minutes)

Review this contract quickly. Give me only:
1. The 3 most dangerous clauses for [my party]
2. The 3 most critical missing clauses
3. Overall risk score: 1 (low) to 10 (critical)
4. Recommendation: sign / negotiate / do not sign

Review for non-lawyers

I am a [CEO/CTO/Sales Director], not a lawyer. Explain this contract
in plain language:
1. What am I agreeing to exactly?
2. What can and cannot I do?
3. How much could it cost me if something goes wrong?
4. Can I get out? How? What does it cost?
5. What should I be worried about?
No legal jargon. Direct answers.

Alternative clause drafting

Clause [N] of the contract states:
"[clause text]"

This clause is unfavorable for [my party] because [reason].

Draft 3 alternative versions:
1. Ideal version for my party (maximalist position)
2. Balanced version (reasonable compromise)
3. Minimum acceptable version (red line)

Each version must be legally sound under [applicable law].

Complete workflow

An AI-assisted contract review workflow has 5 phases:

1. Ingestion: copy the contract text (or upload the PDF if your tool supports it). Remove headers, footers, and automatic numbering that could confuse the AI.
2. Extraction: use the clause extraction prompt to obtain a contract map.
3. Risk analysis: apply the risk identification prompt from your perspective.
4. Human review: the lawyer reviews the AI analysis, validates the alerts, discards false positives, and adds context the AI does not have (business relationship, precedents, strategy).
5. Documentation: generate a review memo with conclusions, recommendations, and negotiation points.

Critical limitations: this is NOT legal advice

Before integrating AI into your practice, you must understand what it cannot do:

• It does not know the business context. AI does not know that this vendor is the only one that can deliver on time, or that the client represents 40% of your revenue. Business context radically changes which risks are acceptable.
• It does not have up-to-date case law. AI models have a knowledge cutoff. They do not know the latest Supreme Court ruling on penalty clauses. For case law analysis, you need specialized legal databases.
• It can invent clauses. Hallucinations exist in the legal domain. AI can cite Civil Code articles that do not exist or invent references to repealed laws. Always verify legal citations.
• It does not replace professional liability. If a client suffers harm from following AI advice without professional review, the liability falls on the lawyer who used it, not the AI.
• Confidentiality. If you use a cloud-based AI (ChatGPT, Claude API), the contract text passes through external servers. This may violate your duty of confidentiality to the client. Use on-premise tools or tools with contractual guarantees of non-retention.

Ejercicio practico

I am in-house counsel reviewing an IT/cloud services contract.
The vendor is [name]. My company has 200 employees, sector [X].

Perform a complete review of the attached contract:
1. Extract all key clauses in a table
2. Identify risks from my perspective (client)
3. Flag missing clauses that should be present
4. Generate an executive memo for leadership
5. Suggest 3 priority negotiation points

Applicable law: [your jurisdiction].
Format: professional, direct, no unnecessary jargon.

Puntos clave