TwinLadder logoTwinLadder
TwinLadder
TwinLadder logoTwinLadder
Back to Archive
TwinLadder Intelligence
Issue #13

TwinLadder Weekly

August 2025

TwinLadder Weekly

Issue #13 | August 2025

M&A Due Diligence: Where AI Excels (and Where It Doesn't)

Luminance claims to revolutionize M&A review. We investigated. Here's what the data actually shows—and what experienced deal lawyers say about trusting AI findings.

The Promise vs. Reality

Every M&A AI vendor promises the same thing: faster due diligence, fewer missed issues, lower costs. Luminance claims their platform can analyze thousands of documents in hours rather than weeks.

But what does that actually mean for deal lawyers? When can you trust AI findings, and when does human review remain essential?

We examined Luminance's actual capabilities, talked to lawyers who've used it on live deals, and identified where AI genuinely delivers—and where it falls short.

What Luminance Actually Does

Luminance is built on a proprietary Legal Large Language Model called LITE (Legal Inference Transformation Engine). The system has been trained on over 150 million verified legal documents.

Core Capabilities

Document Classification: Luminance automatically categorizes uploaded documents by type—contracts, corporate records, regulatory filings, correspondence. This alone can save significant initial triage time.

Concept Extraction: The platform identifies over 1,000 pre-defined legal concepts: change of control provisions, assignment restrictions, termination rights, IP ownership clauses. It surfaces these across massive document sets.

Anomaly Detection: Luminance's pattern recognition excels at finding outliers—the one contract with non-standard terms, the lease with unusual restrictions, the employment agreement with atypical non-compete language.

Data Room Analytics: Dashboard views show document counts by category, flagged issues by severity, and review progress across team members.

The Performance Claims

Luminance marketing cites dramatic improvements. In one case study with Bird & Bird, the firm analyzed 3,600 documents per hour using Luminance compared to 79 documents per hour manually. A team of two associates reviewed contracts relating to 20,000 employees in just two weeks.

Those numbers sound impressive. Let's examine what they actually mean.

Where AI Genuinely Excels

1. High-Volume Document Triage

The most consistent success: initial document classification and prioritization.

Traditional approach: Associates open each document, identify its type, and categorize for review. For a data room with 10,000 documents, this alone takes days.

AI approach: Luminance classifies documents within hours of upload. Associates start substantive review immediately, focusing on categories that matter most.

Verdict: Clear win. AI handles this faster and often more consistently than tired associates at 2 AM.

2. Finding Known Issues Across Large Document Sets

AI excels when you know what you're looking for.

Example: "Flag every contract with a change of control provision that could be triggered by this acquisition."

In a deal with 500 contracts, manually checking each one takes weeks. Luminance surfaces all instances in minutes. Human review then validates and assesses severity.

Verdict: Significant efficiency gain. AI doesn't miss clauses buried on page 47 of an obscure license agreement.

3. Anomaly Detection

Pattern recognition is where Luminance's technology genuinely differentiates.

Example: Reviewing 200 vendor contracts. AI identifies that 198 follow standard terms, but two have unusual limitation of liability provisions.

Human review of 200 contracts might miss those anomalies. AI surfaces them automatically.

Verdict: High value. Anomaly detection catches issues that pattern-fatigue causes humans to miss.

4. Tracking Review Progress

For large deals with multiple reviewers, AI platforms provide real-time visibility into coverage.

Traditional approach: Email updates, spreadsheet tracking, coordination meetings.

AI approach: Dashboard shows exactly which documents have been reviewed, by whom, with what findings.

Verdict: Operational improvement. Not glamorous, but genuinely useful for deal management.

Where AI Falls Short

1. Understanding Business Context

AI can identify a change of control provision. It cannot assess whether that provision actually matters for this specific transaction.

Example: Luminance flags a change of control clause in a minor software license. Experienced deal lawyer knows the $50K/year software is easily replaceable—low materiality. AI treats it the same as a critical manufacturing contract.

Reality: AI surfaces issues. Humans assess materiality. The volume of AI flags can overwhelm if not properly filtered.

2. Cross-Document Analysis

AI examines documents individually. Understanding how provisions interact across a document set requires human synthesis.

Example: The asset purchase agreement excludes certain IP. The license agreement grants rights to that IP. The disclosure schedules reference both. Understanding the complete picture requires connecting dots across documents.

Reality: Luminance cannot generate contracts from scratch without pre-existing templates and struggles with complex analytical synthesis. It's a powerful search tool, not a thinking tool.

3. Negotiation Strategy

Identifying an issue and knowing what to do about it are different skills.

Example: AI flags 47 contracts with inadequate indemnification provisions. Experienced lawyer knows which are worth renegotiating, which to address through escrow, and which to accept as deal cost.

Reality: AI supports analysis but doesn't replace strategic judgment.

4. Non-Standard Document Formats

Luminance only works with Microsoft Word, requiring users to convert documents from other formats before uploading. Scanned PDFs, handwritten amendments, and legacy document formats create friction.

Reality: Real data rooms contain document soup. Format conversion adds time and introduces potential errors.

The Honest Assessment

After examining Luminance's actual capabilities and limitations, here's the bottom line:

Task AI Capability Human Necessity
Document classification Excellent Spot-check only
Known-issue identification Very good Validation required
Anomaly detection Excellent Assessment required
Materiality assessment Poor Essential
Strategic analysis Not applicable Essential
Deal negotiation Not applicable Essential

AI makes due diligence more efficient. It doesn't make human judgment obsolete.

Tool Review: M&A Due Diligence Platforms

Comparing the major players in AI-assisted deal review

Luminance

Market Position: Ranked "Best for M&A" among AI contract review tools (86/100 score)

Strengths:

  • Purpose-built for high-volume due diligence
  • Strong anomaly detection
  • 150M+ document training corpus
  • Good integration with virtual data rooms

Limitations:

  • Microsoft Word only (format conversion required)
  • Cannot generate documents without templates
  • Significant setup time for new matter types
  • Enterprise pricing limits accessibility

Funding: $75M Series C (February 2025)

Best For: Large deals with high document volumes where anomaly detection matters

Rating: 4/5 for M&A due diligence specifically

Kira Systems

Market Position: Established pioneer in contract analysis

Strengths:

  • Mature platform with proven track record
  • Strong clause extraction accuracy
  • Good training capabilities for custom provisions
  • Well-understood by experienced deal teams

Limitations:

  • Interface feels dated compared to newer tools
  • Less emphasis on generative AI features
  • Learning curve for customization

Best For: Teams with established workflows who value stability

Rating: 3.5/5 for M&A due diligence

Harvey (M&A Module)

Market Position: Rapidly gaining share in M&A workflows

Strengths:

  • Generative AI capabilities beyond extraction
  • Natural language querying
  • Strong integration with existing workflows
  • Active development and rapid feature releases

Limitations:

  • Newer to dedicated M&A use cases
  • Less specialized than purpose-built tools
  • Enterprise focus limits small deal accessibility

Best For: Firms already using Harvey who want consolidated platform

Rating: 4/5 for firms in Harvey ecosystem

What's Working: Due Diligence Success Stories

Success Story #1: The Employment Review

Deal type: Tech acquisition, 500 employees across 12 jurisdictions

Challenge: Review employment agreements, equity grants, and compensation arrangements for all employees in compressed timeline.

AI approach: Luminance classified all employment documents by jurisdiction and type. AI flagged non-standard provisions: unusual non-competes, atypical equity terms, change of control provisions.

Result: "Two associates completed review in two weeks that would have taken six weeks manually. AI caught a non-standard severance provision we might have missed—triggered by the acquisition, worth $2M. That alone justified the platform cost."

Key insight: AI shines on high-volume, standardized document types where pattern recognition matters.

Success Story #2: The Contract Audit

Deal type: Private equity platform acquisition of professional services firm

Challenge: Audit 800+ client contracts for fee structures, termination rights, and assignment provisions.

AI approach: Harvey analyzed contract set for specific provisions. Surfaced all instances of rate lock provisions, notice requirements, and consent-to-assign language.

Result: "We identified that 40% of contracts had consent-to-assign requirements—much higher than expected. Informed deal structure and timeline. Without AI, we'd have estimated based on sampling."

Key insight: AI enables complete review rather than statistical sampling.

Hard Cases: Where AI Struggled

Hard Case #1: The Legacy Documents

Deal type: Manufacturing company acquisition

Problem: Data room contained scanned contracts from 1990s, handwritten amendments, and documents in multiple languages.

AI limitation: Luminance couldn't process non-Word formats without conversion. OCR introduced errors in scanned documents. Non-English documents required separate handling.

Outcome: "We spent more time on document preparation than we saved on review. For legacy document sets, AI adds friction rather than removing it."

Lesson: Assess data room quality before committing to AI approach.

Hard Case #2: The False Positive Flood

Deal type: Real estate portfolio acquisition

Problem: AI flagged hundreds of "issues" that experienced real estate lawyers immediately recognized as standard market terms.

AI limitation: System wasn't tuned to real estate norms. What looked anomalous to the AI was routine to practitioners.

Outcome: "Associates spent days validating AI flags that senior lawyers would have dismissed immediately. The tool created work rather than saving it."

Lesson: AI requires domain-specific tuning. Out-of-box configurations generate noise for specialized practice areas.

Hard Case #3: The Strategic Blind Spot

Deal type: Competitive acquisition

Problem: AI identified all material contracts and provisions correctly. But it couldn't identify which relationships were strategically important to preserve versus which could be transitioned.

AI limitation: No understanding of business context. A $10M contract with a strategic partner matters differently than a $10M commodity supplier contract.

Outcome: "AI gave us a complete inventory. We still needed experienced lawyers to assess what mattered. The AI work was necessary but not sufficient."

Lesson: AI is a powerful tool, not a replacement for deal judgment.

Reliability Corner

AI Due Diligence Accuracy Metrics

Metric Luminance Claims Independent Assessment
Document classification 95%+ Generally accurate for standard types
Provision extraction 90%+ Varies by clause complexity
Anomaly detection Not quantified Strong for pattern deviation
False positive rate Not disclosed Varies significantly by setup

When to Trust AI Findings

Trust Level Use Case
High confidence Document classification, standard clause identification
Verify before relying Anomaly flags, complex provision interpretation
Human essential Materiality assessment, strategic analysis

This Month's Reality Check

A June 2025 survey by Artificial Lawyer found that "while pioneers like Kira remain popular, Luminance is a strong competitor in M&A due diligence, and newer GenAI-native tools like Harvey are rapidly gaining ground."

The market is fragmenting. No single tool dominates. Choose based on your specific deal types and existing workflows.

Workflow of the Month: AI-Assisted Due Diligence Checklist

Use this for any M&A due diligence project involving AI tools:

AI-ASSISTED DUE DILIGENCE CHECKLIST
===================================

DEAL: _______________________________
DATE: _______________________________
LEAD ATTORNEY: ______________________
AI TOOL: ____________________________

PRE-REVIEW ASSESSMENT
---------------------
[ ] Data room quality assessed?
    Estimated document count: _______
    Format distribution: Word __% PDF __% Other __%
    Language mix: ___________________
[ ] AI tool appropriate for document types?
[ ] Custom training/configuration needed?
    Timeline for setup: _____________
[ ] Human review resources allocated?

DOCUMENT CLASSIFICATION PHASE
-----------------------------
[ ] AI classification complete?
    Date: __________________________
[ ] Classification accuracy spot-checked?
    Sample size: ______ Error rate: ______%
[ ] Misclassified documents corrected?
[ ] Review priorities established?

PROVISION EXTRACTION PHASE
--------------------------
Target provisions:
[ ] Change of control
[ ] Assignment/consent requirements
[ ] Termination rights
[ ] IP ownership/licensing
[ ] Indemnification
[ ] Limitation of liability
[ ] Non-compete/non-solicit
[ ] Other: _________________________

[ ] AI extraction complete?
[ ] Results exported for review?

ANOMALY REVIEW
--------------
[ ] Anomaly flags reviewed?
    Total flags: ______ Validated: ______
[ ] False positive rate acceptable?
    Rate: ______%
[ ] Material anomalies escalated?
    Count: ______

HUMAN VALIDATION
----------------
[ ] All AI flags validated by qualified reviewer?
[ ] Materiality assessments completed?
[ ] Cross-document analysis performed?
[ ] Strategic implications assessed?

QUALITY CONTROL
---------------
[ ] Sample of AI "no issue" documents manually reviewed?
    Sample size: ______ Issues found: ______
[ ] False negative rate acceptable?
[ ] Review coverage documented?

DELIVERABLES
------------
[ ] Due diligence report complete?
[ ] AI findings appropriately caveated?
[ ] Human review documented?

LESSONS LEARNED
---------------
What worked well:
_________________________________________
What to improve:
_________________________________________
Tool recommendation for similar deals:
_________________________________________

SIGNED OFF: _____________ DATE: _______

Time investment: 15 minutes to complete; ongoing updates during review Why it matters: AI doesn't eliminate quality control—it changes what you're checking.

Quick Hits

Market News:

Adoption Trends:

  • Transaction work is most frequently mentioned AI use case at law firms (25 mentions)
  • Banks and private equity firms showing significant interest in AI for transactional workflows
  • Due diligence platforms increasingly integrated with virtual data room providers

Coming Next Issue:

  • Harvey Reaches 42% of Am Law 100: Adoption Analysis

Ask the Community

M&A AI adoption varies significantly across deal types and firm sizes:

  1. For deal lawyers: What percentage of your due diligence is now AI-assisted?
  2. Technology assessment: Which AI platform performs best for your practice area?
  3. Quality question: Have you caught material issues through AI that manual review might have missed?
  4. Cautionary tales: Any deals where AI created problems rather than solving them?

Reply to share. Anonymized contributions welcome.

TwinLadder Weekly | Issue #13 | August 2025

Helping lawyers build AI capability through honest education.

Sources

Previous Issue
Issue #12
TwinLadder Nedēļas Apskats
Next Issue
Issue #13
TwinLadder Nedēļas Apskats