For five months I measured my weeks by what I added. On Friday I counted what I stopped instead, and the number was 33. No new product, no new app. Just one decision: the tools and routines I had built to run my business had gotten bigger than the business itself. So I made a list of what to stop.
Field Note: the student data check that comes before any AI pilot
There is a box in my 30/60/90 AI Implementation Plan that is not part of any phase. It sits above all three of them, and if you cannot clear it, the clock does not start.
It says: before the pilot begins, check whether the tool will touch student information. If it touches anything that identifies a student, it runs only in a district-approved tool, with written sign-off from someone who has the authority to give it.
Three sentences, and they are the most skipped part of the plan. Not because people disagree. Because across a whole district, the question has no clear answer. "Who signs off on student data in this tool" goes through curriculum, technology, student services, and legal, and a question with four owners has none. So the pilot starts anyway, somebody uploads a class list to see if the thing works, and nobody ever actually signs off.
This is why the first of my three rules is start at department scale, not district scale. One team, one building. At that size the sign-off question has a name attached to it, and you can go get the name in an afternoon.
Two of the three pieces below turn on exactly this. A Maryland board voted to stop student AI use from pre-K through twelfth grade and wrote in an accommodation for students with disabilities. Florida grew its state rule to eighteen requirements and put the line plainly: AI may assist, a person must decide. Both are that same check, written by someone else, after the fact.
So here is a version you can run, and it takes one meeting. Before your pilot opens a tool, write down three things. The decision the tool touches. The person who owns that decision. Who gets told when the tool is involved. If you lead special education, run it for eligibility, referral, and goal drafting specifically, because those are the three where a tool's suggestion can start to look like a decision someone already made.
Then put the paper where the pilot team can see it. A check nobody can point to is not a check.
Worth Your Time
Frederick County School Board votes to ban AI use for pre-K through 12th grade students by Shirin Rajaee, FOX 5 DC
A Maryland district voted 6 to 1 to stop student AI use across every grade, weeks after piloting Gemini in its middle and high schools. The part I want you to notice is the carve-outs: lessons about AI itself, and accommodations for students with disabilities. That second one is doing quiet work, so if your district writes a restriction this year, put the special education exception in the policy text instead of leaving it to case-by-case calls.
Orange County schools face new requirements for classroom AI use by Tylisa Hampton, WKMG News 6
Florida's state board grew its AI rule to eighteen requirements, and districts have until July 1, 2027 to adopt policy. The line I would put in front of every team is from the district's own slide: AI may assist, a person must decide. That covers grading, discipline, threat assessment, and exceptional student education referrals, so if you lead special education, go find out who signs off when a tool flags a kid.
An AI Policy Isn't a Playbook: Educators Need Agency, Not Just New Rules by Beck Tench, Emily Weinstein, and Carrie James, Center for Digital Thriving at Harvard Graduate School of Education
Most districts I talk to have a policy and no plan for the conversation that policy requires. This one gives you five short structured activities for talking with students about AI use, pressure, and the gray areas, and the activities are free under a Creative Commons license. Download them and run one with a single department before your next big assignment goes out.
Try This: the park-or-retire pass
Thirty minutes, once a month. I ran my first one on Friday.
Open a blank page and list every initiative your team has open. Not tasks. Initiatives: the pilot, the committee, the new tool, the thing you keep saying you will get back to. Write them from memory first, because the ones you forget are already answering the question.
Sort every line into one of three piles. You must pick one.
Finish. It is close and it matters. Name the smallest next step and who does it. If that step is longer than one sitting, it goes in the next pile instead.
Park. It matters later. Here is the rule that makes this pass work: a parked item gets a written restart trigger, in the form "restart this when X happens." Not a date. A condition. If you cannot write the trigger, the item is not parked. It is retired. Move it.
Retire. It was a good idea and it is not this year's idea. Write one sentence about why, then delete it from every list it lives on.
Then do the part people skip: go delete it. From the agenda, the shared drive, the standing meeting, the dashboard. An initiative you decided to retire and left on a slide will come back as a question in February.
Here is the prompt I used. The last instruction is the one doing the work:
Here is every open initiative on my team, pasted below. For each one, ask me two questions: what changed because of it in the last 60 days, and what breaks if I stop. Then sort it into finish, park, or retire. For anything you put in park, make me write a restart condition in the form "restart this when X happens." If I cannot write one, move it to retire and tell me you moved it.
The rule underneath all of it: parking something without a trigger is not a decision. It is a way of not deciding, and it costs the same attention as the work would have.
That's it for this one. If something here was useful, share it with someone who'd get something out of it. If it wasn't, that's part of the deal too. You can find more of what I'm working on at evalveconsulting.com, or book a call if you want to talk through what you're dealing with.
Talk soon,
Chris