A diligence finding is only as good as your ability to check it. That was true before AI, and it matters more now that anyone can generate a fluent, confident summary of a company in seconds.
Fluent is not the same as right
Language models are good at producing text that sounds correct. They can also combine facts about two companies with similar names, repeat outdated information, or state a number that appears nowhere in the underlying record. A summary without sources gives you no way to tell which parts to trust.
What a citation does
- It makes the finding checkable. Anyone on the deal team, the investment committee or outside counsel can open the source and confirm it.
- It separates fact from inference. "The registry lists this subsidiary as dissolved in 2023" is a fact with a source. "The company may have governance issues" is an inference and should be labeled as one.
- It makes disagreement productive. If a claim in the management presentation does not match the record, the citation turns an argument into a specific question.
An example
Suppose a confidential information memorandum says a company holds 40% of its regional market. A useful finding does not just say the claim looks high. It says what public data shows (for example, that carrier filings put the company nearer 18% and third in its region) and links to the data used. This example is illustrative.
Every line in a diligence brief should carry the filing, registry, dataset or list behind it. If it cannot, it belongs in a list of questions, not a list of findings.