Provisional Order on Generated Email Reread
A counterfactual account of what office correspondence might look like had AI drafting remained a disclosed practice.
What Would Have Happened If Generated Email Reread Had Remained Disclosed
In the counterfactual, disclosure norms kept pace with adoption. The following is a reconstruction.
An office worker drafts a reply. The draft is inadequate. They open an AI tool and enter the context. The tool returns a response. Under the disclosed version of this scenario, they would paste it, add a line noting the assist, and send.
Instead: someone rereads an AI drafted email several times to make sure it still sounds like them. This check is not a disclosure. It is a reversal. The question is not whether the tool helped. The question is whether anyone can tell.
In the counterfactual, teams developed brief conventions. A parenthetical. A footer. Something. Under those conventions, generated email reread would have become a quality step rather than a concealment step.
Instead, two things happened. Generated email reread adds one private manual check to every send. That check became part of the normal workflow. And then the second thing: assisted work becomes routine while disclosure stays exceptional. The check expanded. The disclosure did not.
The governing rule holds in both versions of events. Under the governing rule, generated email reread becomes procedure once repetition feels safer than reconsidering it. In the actual timeline, the procedure is private. In the counterfactual, it is visible. The mechanical result is the same. The help is useful enough to use and awkward enough to disguise, except in the counterfactual, it was not disguised.
The counterfactual does not assign fault. It identifies the fork. The fork was early 2023. The behavior had not yet calcified. It calcified shortly after.