In this article
- The Buyers Who Haven’t Changed With AI Are at a Disadvantage
- Why Dropping a Data Room Into Claude Isn’t the Answer
- Step 1: Uploading Your Documents
- Step 2: The First Read and What Kudra Surfaces Immediately
- Step 3: Asking Questions and Seeing The AI Agent at Work
- What This Changes About How Due Diligence Feels
The Buyers Who Haven’t Changed With AI Are at a Disadvantage
Private acquisition has always been information-intensive. But something shifted in the last two years that most search funders haven’t fully reckoned with: AI has entered both sides of the transaction. Not just the buyer’s side. Both sides.
Sellers are already using AI to present their businesses in the best possible light. CIMs are more polished. Financial narratives are more coherent. Add-back schedules are more defensible. The documents that land in your data room have been reviewed, refined, and in some cases restructured with AI assistance before you ever open them.
That’s not nefarious. It’s the natural adaptation of a market to available tools. But it creates a specific problem for buyers who haven’t adapted in the same direction. A seller using AI to present their business, against a buyer doing due diligence with a spreadsheet and a checklist, is not a level playing field.
Why Dropping a Data Room Into Claude Isn’t the Answer
The natural first instinct for a search funder who accepts that AI should be part of their due diligence process is to reach for the most accessible tool available. Claude. ChatGPT. A general-purpose large language model that can read documents and answer questions.
It’s understandable. These tools are good at many things. But using a general-purpose AI as a due diligence copilot is a specific case where the gap between what the tool feels like it can do and what it actually does reliably is large enough to be genuinely dangerous. In a context where the quality of your analysis determines whether you overpay for a business, miss a dealbreaker, or walk into a management meeting without knowing what to probe, that gap matters.
Here are four specific, documented reasons why general-purpose AI falls short in a due diligence context.
- 1. Unreliable document extraction in complex formats: Private company data rooms often contain scanned PDFs, multi-column financial statements, spreadsheet exports, and annotated documents that general-purpose AI struggles to interpret accurately, particularly when meaning depends on layout and table structure rather than text alone. (Source: DataStudios — “Claude AI PDF Reading Capabilities, Text Extraction Accuracy and Layout Support”, January 2026)
- 2. Overstated completion of work: Multiple users have reported cases where Claude stated that an analysis was completed, only for follow-up questioning to reveal that portions of the work had been skipped, left unfinished, or inferred without actually being reviewed. (Source: ArsTurn — “Is Claude’s Coding Ability Declining? User Complaints”, August 2025)
- 3. Skipping instructions and truncating complex tasks: Anthropic publicly acknowledged performance issues in 2026 that reduced the quality of complex, multi-step reasoning, following user complaints that Claude was selectively addressing prompts, stopping mid-task, and failing to fully execute instructions. (Sources: Fortune — “Anthropic explains Claude Code’s recent performance decline after weeks of user backlash”, April 2026; Inc. — “Users Say Anthropic’s Claude Is Getting Worse”, April 2026)
- 4. Lack of domain-specific acquisition expertise: While general AI understands financial concepts such as EBITDA and working capital, it lacks the specialised pattern recognition developed through reviewing hundreds of acquisitions, making it less reliable at identifying deal-specific risks, aggressive add-backs, or unusual financial trends. (Source: Glean — “ChatGPT, Claude, or Gemini: Which AI Model Excels in Document Understanding?”, November 2025)
Step 1: Uploading Your Documents
The starting point is the data room. In a typical private acquisition, this arrives in stages, an initial package of financials and a CIM early in the process, followed by contracts, operational documents, employee records, and management accounts as the deal progresses. Kudra is designed to work with documents as they arrive, not only once the data room is complete.
The upload process is straightforward. You drag and drop your documents directly into Kudra, PDFs, Word documents, Excel spreadsheets, management accounts in whatever format the seller’s accountant produced them. Kudra accepts all of it. You don’t need to reformat anything or clean up the data before it goes in. Private company financials are rarely neat. Kudra is built for the messy reality of how information actually comes out of a private business.
As each document lands, Kudra extracts the text and structure not just reading the words on the page but understanding what it contains, and how it relates to the other documents already in the workspace. A P&L gets read as a P&L. A customer contract gets read as a customer contract. An employment agreement gets read as an employment agreement. It figures out what it’s looking at.
Step 2: The First Read and What Kudra Surfaces Immediately
Once the initial document set is processed, Kudra produces its first pass analysis without you having to ask for it. This isn’t a summary of each document, it’s a cross-document read that starts connecting what the financials say to what the contracts say to what the operational documents reveal.
At this stage, Kudra surfaces three things: an initial view of the financial picture, the early red flags that warrant closer attention, and the questions the document set raises that aren’t yet answered by anything in the data room.
Step 3: Asking Questions and Seeing The AI Agent at Work
This is where Kudra works most differently from any other tool in a due diligence process. Once your documents are loaded, you can ask Kudra anything about the business and it answers based on what’s actually in your data room, not from general knowledge.
The questions you ask are the same ones you’d ask an analyst who had read everything. Except the analyst read all of it simultaneously, never got tired, and can cross-reference across fifty documents in the time it takes you to formulate the next question.
It doesn’t make the judgement call about whether the dependency is a dealbreaker. That call is yours. But it makes sure you’re making it with a complete picture, not the partial one that a data room reviewed under time pressure tends to produce.
Every answer Kudra gives is grounded in the documents you uploaded and traceable. It doesn’t speculate. It doesn’t draw on general knowledge about other businesses. If the answer isn’t in your data room, it tells you that and flags it as something to request from the seller.
What This Changes About How Due Diligence Feels
The practical difference Kudra makes is not just analytical. It changes the experience of doing due diligence as a solo search funder in ways that are harder to quantify but just as real.
Without an AI copilot, due diligence on a private business feels like trying to hold too many things in your head at once. You’re reading a document while remembering something from a document you read three days ago, wondering whether those two things are consistent, knowing you should go back and check but not having the time. The anxiety of what you might be missing sits alongside everything else you’re trying to do.
With Kudra running alongside you, that anxiety is substantially reduced, not because the risks are gone, but because you know that the cross-referencing is happening systematically, that nothing is being missed because you were tired or distracted, and that what you’re bringing to the management meeting reflects a complete read of the data room rather than the parts you managed to get to.
References
- https://www.mckinsey.com/capabilities/strategy-and-corporate-finance/our-insights/artificial-intelligence-in-ma-and-private-equity
- https://www.bain.com/insights/how-ai-is-transforming-private-equity/
- https://www.pwc.com/gx/en/issues/data-and-analytics/artificial-intelligence.html
- https://www2.deloitte.com/us/en/insights/focus/cognitive-technologies.html
- https://hbr.org/2023/07/how-generative-ai-changes-work
- https://www.ibm.com/topics/artificial-intelligence
- https://www.investopedia.com/terms/d/due-diligence.asp
