Nothing says collaboration quite like a 14-page AI policy arriving after the software is already in everyone’s toolbar.
That may be a little unfair. Policies matter. Organizations need boundaries around what tools people can use, what data those tools can touch, and what risks are unacceptable.
But a policy is not fluency.
A policy tells people where the guardrails are. Fluency is what happens when people understand how to work inside those guardrails with judgment, accuracy, and responsibility. Too often, leaders confuse the two. They make the decision, write the rules, explain the benefits, and then wonder why the people closest to the work treat the whole thing like another dictated adoption.
That is the part we should sit with.
The Associated Press reported on a legally binding agreement between Microsoft and the American Federation of Teachers governing AI in schools. The agreement includes restrictions on student and educator data, third-party audits, plain-language transparency, and school control over when and how covered tools are used. That is important. It is also an education agreement, not evidence about workplace outcomes.
And even there, the central question does not disappear. A critic quoted by AP warned that strong privacy protections do not, by themselves, answer whether an AI product belongs in a classroom or when it should be introduced.
That distinction matters at work too.
You can have a responsible policy and still avoid the harder conversation. Does this tool improve the work? Does it create a new burden? Does it change someone’s role? Does it make the human decision better, or just faster and harder to question?
Take meeting assistants. I am in plenty of meetings where tools are present. The sensitive issue is not always that a bot exists. In some organizations, people may be comfortable with recording, as long as the transcript stays inside the organization or inside the meeting context. In other moments, the right move may be explicit: we are stopping the recording now because this part of the conversation should not be captured.
That is not just a privacy setting. It is a leadership practice.
It also raises a useful tension. AI can help preserve historical records of what was said, decided, missed, or left unresolved across meetings. That can be valuable. But the people in those meetings know where the real line is. They know when accuracy matters, when context matters, and when a transcript is not the same thing as understanding.
So before introducing one AI tool, ask the team four things:
What should this never do?
Will you actually use it? If so, what would that look like?
How will we know whether it is useful or not useful?
Where must a human stay in the loop to keep the work accurate and up to standard?
Then show people what changed because they answered.
That last part matters. If people give input and nothing changes, you did not create collaboration. You created theater with a survey link.
Leaders still have to decide. But if AI is going to change the work, the people doing the work should not be introduced to the change after the decision has already been laminated.
Take this free AI Fluency test: www.checkaifluency.com
Sources
Microsoft commits to sweeping AI privacy rules for students. Will other tech giants follow? — Associated Press
https://apnews.com/article/ai-school-student-data-privacy-microsoft-google-89b040631d635e0d8673f1e9aa9a6e28
AFT, UFT and Microsoft announce National AI Safety & Privacy Standard for schools — Microsoft Source
https://news.microsoft.com/source/2026/09/09/aft-uft-and-microsoft-announce-national-ai-safety-privacy-standard-for-schools-to-protect-students-families-and-educators/