Due diligence in the AI era

A typical due diligence in an M&A transaction involves reviewing between 500 and 5,000 documents. Contracts, minutes, financial statements, licenses, litigation records, permits, insurance policies. A team of 4 lawyers takes between 3 and 6 weeks. With AI, that same team can complete the first phase of review in 3-5 days.

Not because AI does the lawyer's work. Because it does the work that does not need a lawyer: reading, classifying, extracting standard data, identifying missing documents, and flagging anomalies. The lawyer focuses on interpreting, assessing risks, and advising.

What changes with AI

• Automatic classification: AI sorts thousands of documents by type (contract, minutes, invoice, license) in minutes.
• Data extraction: dates, amounts, parties, deadlines, terms. Without reading each document manually.
• Absence detection: identifies which documents should be in the data room and are not.
• Cross-referencing: detects inconsistencies between documents (a contract says one thing, the board minutes say another).

Bulk document review

The main challenge of a DD is volume. You cannot paste 5,000 documents into an AI chat. You need a system.

Batch processing strategy

Divide the data room into categories and process each with specialized prompts:

1. Phase 1: Classification. Process each document with a classification prompt that returns type, date, parties, and relevance.
2. Phase 2: Extraction. For each category, apply a specialized extraction prompt (contracts, minutes, financials, employment, litigation).
3. Phase 3: Anomalies. Cross-reference extracted data to detect inconsistencies.
4. Phase 4: Report. Consolidate findings into a structured report.

Document classification prompt

Classify the following document from an M&A transaction data room.

Return ONLY this information in JSON format:
{
  "type": "[contract/minutes/invoice/license/policy/financial_statement/
            employment_document/litigation/permit/correspondence/other]",
  "subtype": "[brief description]",
  "date": "[document date or 'not_identified']",
  "parties": ["party 1", "party 2"],
  "amount": "[if applicable, or 'not_applicable']",
  "in_force": "[yes/no/undetermined]",
  "dd_relevance": "[high/medium/low]",
  "summary": "[1-2 lines maximum]",
  "red_flags": ["list of anomalies detected, if any"]
}

Document:
[paste text]

When processing documents in batch, use JSON format. It allows you to consolidate results in a spreadsheet or database for subsequent analysis. A CSV with 2,000 classified rows is infinitely more useful than 2,000 free-text summaries.

Contract extraction prompt

Extract the following information from this data room contract:

1. Contract type
2. Parties (full legal name)
3. Execution date
4. Expiration date
5. Total / recurring amount
6. Change of control clause (yes/no, literal text)
7. Early termination clause (conditions)
8. Guarantees granted
9. Litigation or arbitration mentioned
10. Assignment restrictions

CRITICAL: If the contract has a change of control clause that
could be triggered by the M&A transaction, flag it as RED FLAG.

Format: structured JSON.

Due diligence in M&A transactions

In an M&A transaction, legal DD covers multiple areas. AI can assist in each with specific prompts.

Critical areas of legal DD in M&A

1. Corporate

Review the corporate documentation of [target company]:
- Articles of incorporation: share capital, corporate purpose, restrictions
- Current bylaws: share transfer clauses, qualified majority,
  drag-along and tag-along rights
- Board minutes: relevant decisions from the last 3 years
- Shareholders' register: complete ownership structure
- Powers of attorney: who has authority to bind the company

Identify:
1. Restrictions on share transfers
2. Shareholders' agreements mentioned
3. Authorizations required for the transaction
4. Delegated powers from the board of directors

2. Material contracts

Analyze the material contracts of [target company].
A contract is material if:
- Amount > [threshold] EUR/year
- Duration > 3 years
- Client representing > 10% of revenue
- Critical vendor with no alternative

For each material contract:
1. Does it have a change of control clause?
2. Would the transaction trigger that clause?
3. Consequences of triggering (termination, renegotiation, penalty)
4. Is prior consent required?
5. Risk of losing this contract post-transaction: LOW/MEDIUM/HIGH

3. Employment

Review the employment documentation of [target company]:
- Headcount: number of employees, contract types, seniority
- Applicable collective bargaining agreement
- Relevant individual agreements (golden parachutes, non-compete)
- Open employment litigation or contingencies
- Incentive plans with change of control clauses
- Union representation and works council agreements

Identify obligations that trigger upon a change of control
and quantify the potential cost.

4. Litigation and contingencies

Generate a legal contingencies matrix based on the documentation:

For each litigation or proceeding:
1. Type (judicial/arbitration/administrative)
2. Claimant and defendant
3. Claim (amount or action)
4. Procedural status
5. Probability of adverse outcome: high/medium/low
6. Estimated financial impact
7. Existing accounting provision (yes/no, amount)
8. Insurance coverage (yes/no)

Format: table sorted by descending financial impact.
At the end, total contingencies by probability.

KYC/AML: automated verification

Know Your Customer (KYC) and Anti-Money Laundering (AML) processes are mandatory for financial institutions, law firms (under the EU Anti-Money Laundering Directives), and many other obliged entities. AI can automate a large portion of the process.

Corporate identity verification

Analyze the following documentation of a potential corporate client
for KYC verification:

1. IDENTIFICATION
   - Full legal name and tax ID
   - Companies register: current registration?
   - Legal representatives: identified with sufficient powers?

2. ULTIMATE BENEFICIAL OWNERSHIP (UBO)
   - Identify all beneficial owners (>25% ownership)
   - Ownership chain to natural persons
   - Persons with effective control without majority ownership
   - PEPs (Politically Exposed Persons) in the structure

3. ACTIVITY
   - Corporate purpose consistent with declared activity?
   - Source of funds documented?
   - Countries of operation (FATF high-risk list?)

4. AML RED FLAGS
   - Complex corporate structures without commercial justification
   - Entities in opaque jurisdictions (tax havens)
   - Frequent ownership changes
   - Discrepancies between declared and documented activity
   - PEPs without enhanced justification

Result: KYC file with risk level (low/medium/high/very high)
and required due diligence (simplified/standard/enhanced).

Sanctions screening

Verify whether the following persons or entities appear on
international sanctions lists:

Persons/Entities to verify:
[list]

Lists to check:
- EU sanctions list (Regulation 269/2014 and subsequent)
- OFAC SDN List (USA)
- United Nations sanctions list
- HM Treasury (United Kingdom)
- FATF high-risk countries list

For each person/entity:
1. Exact or partial match found?
2. List on which they appear
3. Reason for the sanction
4. Date of inclusion
5. Implications for the business relationship

NOTE: This screening must ALWAYS be supplemented with professional
databases (World-Check, Dow Jones, LexisNexis). AI does not have
access to real-time updated lists.

Systematic red flag detection

One of the greatest advantages of AI in DD is the systematic detection of anomalies that a human reviewer might miss due to fatigue or volume.

Red flag detection prompt

Analyze the following data set extracted from the data room
and detect anomalies:

DATA:
[paste consolidated data: dates, amounts, parties, terms]

Look for these patterns:
1. TEMPORAL INCONSISTENCIES
   - Contracts signed before one of the parties was incorporated
   - Backdated documents
   - Undated documents

2. FINANCIAL INCONSISTENCIES
   - Contract amounts that do not match declared revenue
   - Insufficient provisions for open litigation
   - Related party transactions at non-market prices

3. CRITICAL ABSENCES
   - Contracts mentioned in minutes but not present in the data room
   - Litigation not included in the contingencies list
   - Key employees without a contract or without non-compete clauses

4. REGULATORY RISKS
   - Expired or soon-to-expire licenses
   - Documented regulatory breaches
   - Undisclosed administrative sanctions

5. RELATED PARTIES
   - Transactions with group companies
   - Directors with interests in competitors
   - Undeclared conflicts of interest

Format: prioritized list by severity with documentary evidence.

Findings organization and reporting

DD results must be presented in a structured manner. AI can generate different formats depending on the audience.

Executive report for the investment committee

Generate an executive due diligence legal report for the investment
committee. Maximum 3 pages.

Structure:
1. EXECUTIVE SUMMARY (1 paragraph)
   - Overall recommendation: proceed / proceed with conditions / do not proceed

2. CRITICAL FINDINGS (potential deal breakers)
   - Maximum 5, ordered by impact

3. SIGNIFICANT FINDINGS (require pre/post-closing action)
   - List with risk quantification

4. RECOMMENDED CONDITIONS PRECEDENT
   - What must be fulfilled before closing

5. SUGGESTED PRICE ADJUSTMENTS
   - Contingencies that should be reflected in the price

6. POST-CLOSING OBLIGATIONS
   - Legal, employment, contractual integration

Tone: direct, quantified, decision-oriented.
Do not include detailed analysis (that goes in the full report).

Findings data room organization

Organize the DD findings in the following folder structure:

DD-[Company]-[Date]/
├── 01-Executive-Resumen/
│   ├── executive-report.pdf
│   └── risk-matrix.xlsx
├── 02-Corporate/
├── 03-Material-Contracts/
│   ├── with-change-of-control/
│   └── without-change-of-control/
├── 04-Employment/
├── 05-Litigation-Contingencies/
├── 06-Tax/
├── 07-Regulatory-Licenses/
├── 08-Intellectual-Property/
├── 09-Real-Estate/
├── 10-Environmental/
└── 11-Pending-Documents/

For each folder, list the documents reviewed, the findings,
and the documents that are missing.

Prompt patterns for due diligence

Pattern: completeness checklist

Compare the list of documents present in the data room with the
standard DD checklist for a [type] transaction in the [X] sector.

Documents present: [list]

Identify:
1. Mandatory documents that are missing (CRITICAL)
2. Recommended documents that are missing (IMPORTANT)
3. Unnecessary documents included (can be removed)
4. Documents with upcoming expiration dates

Generate a supplementary document request ready to send
to the seller.

Pattern: trend analysis

Analyze the evolution of these indicators over the last 3-5 years:
[extracted financial/operational data]

Identify:
1. Positive trends (growth, margin improvement)
2. Negative trends (deterioration, risk concentration)
3. One-off anomalies that require explanation
4. Reasonable projection for the next 2 years

Context: this information is used to value the company in a
[type] transaction. Flag any data that could significantly
affect the valuation.

Tools and workflow

For a real DD, you will combine AI with specialized tools:

• Virtual data rooms: Datasite, Intralinks, Ansarada. AI does not replace them, but you can export documents and process them.
• Generative AI: Claude (best for long documents due to context window), GPT-4 (strong for financial analysis), Gemini (strong for multi-language).
• OCR: many data rooms contain scanned documents. Use Google Document AI or Azure Form Recognizer before feeding them to AI.
• Spreadsheets: consolidate all extraction in a master Excel. Pivot by document type, risk, date, amount.
• Legal databases: Westlaw, LexisNexis, Practical Law for verifying cited case law and regulations.

Ejercicio practico

Puntos clave