← Home/Legal · Due diligence
For legal, M&A and transaction teams

A Claude agent for due diligence

The agent works through the data room, extracts the key points of each document and drafts the report’s findings. Lawyers validate every conclusion.

Request an auditFrequently asked questions
Definition

A Claude agent for due diligence is an AI system that reviews the documents in a data room (contracts, corporate records, litigation, intellectual property, employment), extracts significant clauses and facts, compares them with the team’s review grid and drafts sourced findings. It operates in the orange zone: every finding is validated by a lawyer before it goes into the report.

How it works

Before / after the agent

01
Data room inventory
BeforeThe team discovers the data room folder by folder and builds the index by hand.
With the agentThe agent indexes the data room, classifies every document and flags what is missing against the request list.
02
Contract review
BeforeLawyers read every contract to find change of control, exclusivity or termination terms.
With the agentThe agent extracts those clauses from every contract, with the exact reference of the passage, following your review grid.
03
Risk identification
BeforeWeak signals get lost across hundreds of documents.
With the agentThe agent cross-checks documents, spots inconsistencies and ranks the risks identified by severity, using your criteria.
04
Due diligence report
BeforeThe report is pulled together in a rush, from everyone’s notes.
With the agentThe agent drafts sourced findings and a summary, the lawyer validates every conclusion, orange zone.
LOOP™ governance

Recommended trust zone

Orange zoneSupervision

A due diligence report informs an investment decision and engages the liability of those who sign it. The agent reads, extracts and proposes, lawyers validate every finding before it goes into the report, orange zone.

Related demo

See the agent in action

Legal Intelligence Agent
Contracts analysed and changes proposed, reviewed by a lawyer.
View the demo
Document Intelligence Suite
Key obligations and data extracted from your documents, with their source.
View the demo
Go further

Related resources

Discover the practice
Legal and compliance function
All the AI use cases for the legal function, from contract review to due diligence.
Discover
Discover Claude Ops
Claude Ops
The industrialisation programme that puts this agent, and others, into production on your business processes.
Discover
Frequently asked questions

Frequently asked questions

Can an AI agent write a due diligence report on its own?
No. The agent drafts the findings and a summary, each with a reference to the source document. Conclusions, risk assessment and signing the report stay human.
How is data room confidentiality preserved?
Trade secret protection only covers information whose holder has taken reasonable steps to keep it secret (EU Directive 2016/943, article 2; in France, Commercial Code, article L151-1). The agent’s environment is one of those steps: access limited to the deal team, hosting chosen during scoping, logging and retention defined in line with your confidentiality undertakings.
Are our documents used to train the models?
On Anthropic’s commercial offerings, the data you send is not used to train models by default. The retention period that applies to the model and offering used is checked and documented during scoping.
Which kinds of due diligence does the agent cover?
Legal due diligence first (contracts, corporate, litigation, intellectual property, employment), plus the document-heavy parts of financial or compliance due diligence. The review grid, materiality thresholds and report format remain your team’s.
What about personal data in the data room?
Employment contracts, executive files: a data room holds personal data subject to the GDPR. The agent only extracts what the review grid requires, and access stays limited to members of the deal team.
Next step

Let’s talk about your use case

A Claude Ignite audit maps your processes and prioritises the highest-impact use cases, this one or another.

Book a call →