Fictional teaching case; no real organization or study is described.

Two teams must approve every AI-drafted grant report before it is sent, and either team can refuse. A year later, both still hold exactly that veto.

• Team A uses the assistant to draft and spends part of the time saved checking a sample of claims against the cited sources, keeping a short log of what it finds.
• Team B has stopped opening the sources: the drafts usually look fine, so approval has become a quick read of the summary.

Their decision authority is identical; their checking practices now differ. That difference in practice is all the case establishes. It does not show which team would catch a real defect, how either team's abilities have changed, or that Team A's checks are effective: Team B may still understand the subject well, and Team A's log may miss what matters. Finding out would require observing how each team actually checks relevant cases, and inferring a change would need comparable evidence over time. The assistant did not determine either practice: the time it saved gave Team A an opportunity to check more, while Team B used the same tool in a way that dropped a routine check. A veto alone reveals neither team's practice nor its capability.

Questions to consider (ungraded): What is observed here, and what is only assumed? What evidence would show whether either team can spot a real defect? Could Team B rebuild a checking routine without giving up the assistant? Which disagreements are about facts, and which are about how much checking is worth its cost?