Main idea

A prospect sends a security questionnaire in which one box is already filled in: the vendor's ISO 9001 certification supposedly dates from 2017. The sentence is precise, written, placed next to the right question. The correct date is 2018.

An old answer has all the outward properties of knowledge: it asserts, it is well written, it answers the question asked. What it lacks is what would make it trustworthy — who wrote it, what it rests on, when it was true, whether it still describes the product. It may come from a document template, from another vendor, from an exchange nobody remembers; it may have been right the day it was sent and wrong today.

A client's questionnaire keeps a value: it says what the market wants to know, in its own words and with its own priorities. But it proves nothing. The request comes from the client; the evidence can only come from the vendor.

Why it matters

It shifts the question you ask of a store of past answers: not "have we answered this before?" but "what still allows us to believe it?". A reusable answer is one that keeps its origin, its date and its limits; without them, reusing it amounts to signing a statement whose source you don't know.

The risk grows with an AI assistant tasked with filling in the document: an answer already written is ideal material for it — assertive, well placed, consistent with the question — and it has no reason to be wary of it. It speeds up the spread of an error nobody has seen.

Nuances and limits

An old answer is not wrong by principle: it can point to where to look, and it is often right. What is missing is not correctness but the means to establish it; throwing it away unread would mean losing a lead.

A one-year gap on a certification date is harmless in itself. Its value is as a signal: it reveals that nothing in the answer would have made it possible to detect it.

Open questions

• Beyond what age must an answer with a source and a date be verified again before it is sent, and who sets that limit?