The problem
Due diligence on a business acquisition, property transaction, or financing deal is fundamentally a pattern-matching exercise. You review a large volume of documents looking for issues: the commercial contract that has an unusual termination right, the property lease with a personal guarantee, the employment agreement that does not have a non-compete, the regulatory licence with conditions attached.
The challenge is that the volume of documents is large, the deadline is usually tight, and the cost of missing something is significant. A missed issue in due diligence that materialises post-completion is a professional indemnity claim.
For a medium-complexity corporate acquisition, due diligence might involve reviewing 150 to 300 documents. The report that synthesises the findings and presents them in a digestible form for the client and deal team typically takes 8 to 15 hours to write, depending on complexity and how well organised the document review has been.
This workflow does not replace the document review itself — that requires legal judgement. It compresses the report writing and synthesis phase.
The system
Step 1: Maintain a structured review log
As you review each document, capture your findings in a standardised format:
- Document reference and name
- Category (contracts, property, employment, IP, regulatory, etc.)
- Key terms or provisions
- Issues identified (if any)
- Risk level: Red (material issue requiring renegotiation), Amber (notable issue to be addressed), Green (no issue)
- Recommended action
A spreadsheet or Notion database works well for this. The structured log is what powers the report generation.
Step 2: Upload category summaries to NotebookLM (NotebookLM)
Export your review log by category and upload each category as a document in NotebookLM. If you have key documents with specific issues, upload those too.
Ask NotebookLM to synthesise by category:
"Based on the document review log for the employment contracts category, what are the main issues identified? Summarise them by risk level and explain the practical implications for the buyer."
Repeat for each category. This gives you a synthesised narrative for each section of the report.
Step 3: Generate the executive summary (Claude)
Take your full issues list (all categories) and paste it into Claude:
"I am writing a due diligence report for a business acquisition. Here is a summary of all issues identified across the document review. Generate an executive summary that:
- Opens with a brief description of the scope of the review
- Summarises the key material issues (Red) and their implications
- Notes the significant but manageable issues (Amber)
- Gives an overall assessment of the risk profile
- Lists the recommended next steps before exchange
Write for a non-lawyer client and a commercial deal team. Be clear and direct — they need to understand what the issues are, not just that issues exist. UK spelling, no em dashes.
Issues summary: [paste your categorised issues list]"
Step 4: Structure the full report (Claude)
"Using the category summaries and executive summary, create a table of contents and section structure for the full due diligence report. For each section, include: the scope of review, the key issues found (Red and Amber), and the recommended actions. Format this as a structured outline I can use to build the final report."
The results
Before: 8 to 15 hours to write a due diligence report once the review was complete.
After: 3 to 5 hours, with a better-structured report and a clearer executive summary.
The biggest time saving comes from the synthesis step. Turning 200 rows of a review log into coherent, readable prose by category is mechanically slow when done manually. NotebookLM and Claude can produce a first draft of each category section in minutes, leaving you to apply your legal judgement to the output rather than spending time on the drafting itself.
The executive summary improvement is also significant. Clients read the executive summary and make decisions based on it. A clear, risk-stratified summary with specific recommended actions is far more useful than a list of issues without context.