Two findings sit next to each other in a due diligence report. One says the model hits 94 percent precision. The other says the team has a strong retraining process. They look equally authoritative on the page. But the first might be a number we verified against production logs, and the second might be something a founder said on a call. If the report does not tell you which is which, it has hidden the most important thing about each finding: how much you should trust it.
Our team grades every finding by its evidence level. There are three.
Reported
The lowest level. Someone told us. A founder, a CTO, a slide. Reported findings are not worthless. They are context, and often the starting point for a deeper look. But a reported finding carries no independent confirmation, and a report that presents reported claims with the same weight as verified ones is misleading the reader, usually without meaning to.
Corroborated
The middle level. The claim is supported by more than one source, or by a document, even if we have not independently reproduced it. A metric that appears in an internal report and matches what two team members describe is corroborated. It is more reliable than a single verbal claim and less reliable than something we tested ourselves.
Verified
The highest level. We independently confirmed it. We ran the evaluation, inspected the production logs, reproduced the result, or examined the actual system output. A verified finding is one we will stand behind, because we did not take anyone's word for it.
Why grading changes the decision
When findings are graded, the decision-maker can act intelligently. A verified red flag is a hard stop. A reported strength is a reason to look closer before relying on it. Collapse the grades into one undifferentiated list and the reader loses the ability to weigh anything. They are left with a document that looks rigorous and tells them nothing about confidence.
There is a second benefit, and it is one good founders appreciate. Grading protects them too. A founder making an honest claim they cannot yet prove would rather see it marked "reported" than have it inflated into a verified fact that later does not hold. Evidence grading is not about catching people out. It is about being precise regarding what is known, what is supported, and what is still just a claim, so that everyone is making the decision on the same honest footing.
If you are weighing an AI investment, acquisition, vendor selection, or training programme, our team is happy to start with a conversation about scope and approach.
The views and findings in this article are shared for general information only. They are high-level perspectives, not legal, financial, regulatory, or other professional advice, and should not be relied upon for any specific decision or circumstance. For guidance tailored to your situation, please consult a qualified adviser.