AI for Contract Review

AI-powered redlining: finding every deviation from your standard terms

Traditional redlining is a comparison exercise: you place the counterparty's draft next to your firm's standard template and identify every point of deviation. This is tedious for short contracts and practically impossible to do comprehensively for long, complex agreements where deviations may be buried in defined terms, cross-references, or boilerplate.

AI transforms this workflow. Instead of manually comparing documents paragraph by paragraph, you provide the AI with both the counterparty's draft and your standard template, along with instructions about which deviations matter and which are acceptable.

The AI can identify deviations at multiple levels:

• Substantive deviations — where the counterparty's language changes the legal effect of a provision (e.g., shifting indemnification from mutual to one-sided)
• Definitional deviations — where defined terms are altered in ways that change scope (e.g., expanding "Confidential Information" to exclude categories your standard template protects)
• Structural deviations — where provisions are moved, split, or combined in ways that change their interaction with other clauses
• Missing provisions — where your standard template includes a clause that the counterparty's draft omits entirely (e.g., no data protection addendum, no most-favoured-nation clause)

The output is not just a tracked-changes document. It is an annotated analysis: here is what changed, here is why it matters, and here is how it compares to your firm's standard position.

Pulling specific provisions across hundreds of contracts

Clause extraction is the workflow that demonstrates AI's volume advantage most clearly. Consider these scenarios:

A private equity firm is acquiring a target company. The data room contains 800 contracts. The deal team needs to identify every contract that contains a change-of-control provision, an anti-assignment clause, or a termination-for-convenience right — because each of these could be triggered by the acquisition.

A corporate legal department has just learned about a new data privacy regulation. They need to review their entire vendor contract portfolio — 2,000 agreements — to determine which contracts contain data processing provisions and whether those provisions comply with the new requirements.

A litigation team is preparing for trial in a breach-of-contract case. They need to compare the indemnification language across every version of the agreement exchanged during negotiation — 14 drafts over six months — to establish what each party understood and agreed to at each stage.

In each scenario, a human team would spend days or weeks reading contracts. AI can extract the relevant provisions from the entire set in hours, presenting results in a structured format: contract name, relevant clause, page number, and a plain-language summary of the provision's effect.

The key to effective clause extraction is specificity in your instructions. Do not ask AI to "review these contracts." Ask it to "extract every indemnification provision from each contract, identify whether the indemnification is mutual or one-sided, note any carve-outs or caps, and flag any provision that differs from the following standard language."

Automated risk identification against your negotiation playbook

Every sophisticated legal department and law firm has a negotiation playbook — a set of rules defining preferred positions, acceptable fallback positions, and deal-breaker terms for each contract type. The challenge is enforcement: when a junior associate or contract manager is reviewing their twentieth contract of the week, playbook compliance becomes inconsistent.

AI enforces playbook rules with perfect consistency. You encode your playbook into the AI's instructions:

• Preferred position: Mutual indemnification capped at the contract value
• Acceptable fallback: Mutual indemnification capped at 2x the contract value
• Escalation required: Any one-sided indemnification or uncapped liability
• Deal-breaker: Unlimited indemnification for IP infringement without mutual obligation

The AI reads every contract against these rules and generates a risk assessment: green (matches preferred position), yellow (within acceptable range), orange (requires escalation), red (deal-breaker term identified).

This is not theoretical. Firms are deploying this workflow today and seeing dramatic improvements in consistency. A contract that a tired associate might wave through after a long day gets flagged by AI every single time.

Transforming due diligence from weeks to days

M&A due diligence is contract review at its most intense. A typical mid-market transaction involves reviewing 300-1,000 contracts in the target's data room. A large transaction can involve thousands. The deal team needs to extract critical information from every contract:

• Change-of-control and assignment provisions that could be triggered by the transaction
• Key commercial terms — pricing, term, renewal, termination rights
• Liability provisions — indemnification, limitation of liability, warranty disclaimers
• IP-related provisions — licence grants, IP ownership, restrictions on use
• Employment-related provisions — non-compete clauses, severance obligations, change-of-control bonuses
• Regulatory provisions — data privacy, export control, sanctions compliance

Traditionally, this work is done by a team of junior associates and paralegals working around the clock during the exclusivity period. The quality varies with fatigue. The cost is substantial. And the risk of missing a critical provision in contract number 487 is real.

AI transforms this workflow in two ways. First, it extracts the information from each contract faster than any human team. Second, and more importantly, it does so with perfect consistency. The extraction criteria applied to contract number 1 are applied identically to contract number 800.

The output is a structured due diligence report: a spreadsheet or table with one row per contract and columns for each extraction category, with links back to the specific provisions in each document. The deal team reviews the structured output and focuses their judgment on the contracts and provisions that actually warrant attention — rather than spending days reading contracts that turn out to be unremarkable.

Prompt template: NDA review

Here is a prompt template for AI-assisted review of a non-disclosure agreement. Adapt it to your firm's specific standards and playbook positions.

You are a corporate attorney reviewing a non-disclosure agreement (NDA).
Review the attached NDA and provide analysis on the following points:

1. PARTIES AND STRUCTURE
   - Identify all parties. Is this mutual or one-way?
   - What is the effective date and term?

2. DEFINITION OF CONFIDENTIAL INFORMATION
   - How is "Confidential Information" defined?
   - Are there carve-outs (publicly available information, independently
     developed, etc.)? List each carve-out.
   - Does the definition cover oral disclosures? If so, what marking or
     confirmation requirements exist?

3. PERMITTED USE AND DISCLOSURE
   - What is the permitted purpose for using Confidential Information?
   - Who may receive Confidential Information (employees, advisers,
     affiliates)?
   - Are there "need to know" limitations?

4. TERM AND SURVIVAL
   - What is the term of the agreement?
   - How long do confidentiality obligations survive after termination?
   - Flag if survival period is less than 2 years or greater than 5 years.

5. RETURN/DESTRUCTION OBLIGATIONS
   - What are the obligations upon termination?
   - Are there exceptions for archival copies or regulatory retention?

6. REMEDIES
   - Does the NDA include injunctive relief provisions?
   - Are there liquidated damages or penalty provisions?

7. RISK FLAGS
   - Flag any non-standard provisions (non-solicitation, non-compete,
     standstill, exclusivity).
   - Flag any provision that is one-sided when you would expect mutual
     obligations.
   - Flag any governing law or dispute resolution provisions that differ
     from [YOUR PREFERRED JURISDICTION].

Present your analysis in a structured table format with columns:
Provision | Summary | Risk Level (Green/Yellow/Red) | Notes

Prompt template: SaaS agreement review

SaaS agreements present unique risks around data ownership, service levels, and vendor lock-in. This template targets the provisions that matter most.

You are a technology transactions attorney reviewing a SaaS
(Software-as-a-Service) agreement. Analyse the attached agreement and
report on the following:

1. SERVICE DESCRIPTION AND SCOPE
   - What services are included? What is excluded?
   - Are there usage limitations (users, data volume, API calls)?

2. DATA RIGHTS AND OWNERSHIP
   - Who owns the data input into the platform?
   - Does the vendor claim any rights to customer data (anonymised,
     aggregated, or otherwise)?
   - What happens to customer data upon termination? Is there a data
     export/portability provision?
   - What is the data deletion timeline after termination?

3. SERVICE LEVELS (SLA)
   - What uptime commitment is provided (e.g., 99.9%)?
   - What are the remedies for SLA failure (credits, termination right)?
   - Are there exclusions from the SLA calculation?

4. SECURITY AND COMPLIANCE
   - What security certifications does the vendor maintain (SOC 2,
     ISO 27001)?
   - Is there a data processing addendum (DPA)?
   - Does the agreement address data breach notification? What is the
     notification timeline?

5. LIMITATION OF LIABILITY
   - What is the liability cap? Flag if less than 12 months of fees.
   - Are there carve-outs from the liability cap (IP infringement,
     confidentiality breach, data breach)?
   - Is there a consequential damages exclusion? Flag if mutual vs
     one-sided.

6. TERM, RENEWAL, AND TERMINATION
   - What is the initial term? Is there auto-renewal?
   - What is the notice period for non-renewal?
   - What are the termination for convenience rights (if any)?
   - What are the termination for cause triggers and cure periods?

7. IP AND INDEMNIFICATION
   - Does the vendor provide IP indemnification?
   - Is there mutual indemnification or one-sided only?
   - What is the process for IP claims (notice, control of defence)?

8. RISK FLAGS
   - Flag any provision where the vendor's liability is uncapped.
   - Flag any data ownership language that is ambiguous.
   - Flag any auto-renewal with notice period shorter than 60 days.
   - Flag any unilateral right of the vendor to modify terms or pricing.
   - Flag if governing law differs from [YOUR PREFERRED JURISDICTION].

Present findings as: Provision | Contract Language (quote) |
Standard Market Position | Risk Level | Recommended Action

Prompt template: employment contract review

Employment contracts require attention to jurisdictional enforceability, particularly around restrictive covenants. This template targets the provisions that generate the most litigation.

You are an employment law attorney reviewing an employment agreement.
Analyse the attached contract and report on the following:

1. POSITION AND COMPENSATION
   - What is the role, reporting structure, and employment classification?
   - What is the base compensation, bonus structure, and equity (if any)?
   - Are there clawback provisions on bonuses or equity?

2. TERM AND TERMINATION
   - Is this at-will or fixed-term employment?
   - What are the termination provisions (with cause, without cause,
     resignation, mutual)?
   - What constitutes "cause" for termination? Is the definition
     reasonable and specific?
   - Is there a notice period for termination?

3. RESTRICTIVE COVENANTS
   - Non-compete: What is the scope (duration, geography, activity)?
     Flag if duration exceeds 12 months or geography is unreasonable.
   - Non-solicitation: Does it cover clients, employees, or both?
     What is the duration?
   - Non-disclosure/confidentiality: What is covered? What is the
     survival period?
   - Flag any restrictive covenant that may be unenforceable in
     [EMPLOYEE'S STATE/JURISDICTION] based on current enforceability
     standards.

4. INTELLECTUAL PROPERTY
   - Is there an invention assignment clause?
   - Does it cover inventions created outside work hours or using
     personal resources?
   - Is there a carve-out for prior inventions?
   - Does the clause comply with state-specific protections (e.g.,
     California Labor Code Section 2870)?

5. SEVERANCE AND CHANGE OF CONTROL
   - Are there severance provisions upon termination without cause?
   - Is there a change-of-control provision (acceleration of equity,
     enhanced severance)?
   - Is severance conditioned on a release of claims?

6. DISPUTE RESOLUTION
   - Is there a mandatory arbitration clause?
   - Does the arbitration clause cover statutory claims (discrimination,
     wage/hour)?
   - Is there a class/collective action waiver?
   - What is the governing law and venue?

7. RISK FLAGS
   - Flag any provision that is materially one-sided against the employee.
   - Flag any restrictive covenant that may face enforceability challenges.
   - Flag any provision that conflicts with applicable labour law.
   - Flag if the agreement lacks standard protections (e.g., no severance
     upon termination without cause, no IP carve-out for prior inventions).

Present findings as: Provision | Key Terms | Enforceability Risk |
Recommended Revisions