Writing · November 14, 2026

Plan for the dip

Every rollout gets worse before it gets better, and the plans that survive are the ones that said so out loud before it happened.

Somewhere around week six of almost every rollout, someone pulls up the usage dashboard and the line is going the wrong way. Logins are down from the first week. A few educators have gone back to the old way of doing things. Someone asks, carefully, whether this was the right tool. Nothing has actually gone wrong. The plan just never told anyone this part was coming.

I took a course called Risk Management Process and Practice in early 2024. Most of it was what you would expect. Make a list of what could go wrong. Rate each risk by how likely it is and how much it would hurt. Decide ahead of time what you will do if it happens. Then keep watching, because the list changes. Frameworks with acronyms. Matrices with red, yellow, and green squares.

What my own risk list got wrong

One of my papers was a risk plan for bringing technology into schools. I listed the risks I thought mattered. Funding that rises and falls with federal budgets. Privacy and accessibility rules. Products that age out fast. Bugs, outages, support tickets. Near the end, almost as an add-on, I wrote a section about people.

Reading it now, I had the order backwards. Every risk on that list is real, and none of them is the one that stalls most rollouts. If you run the course’s own test on it, the dip wins easily. How likely is it? Close to certain. How much does it hurt? A lot, if nobody saw it coming. By the course’s own logic, it belongs at the top of the list.

The shape of it

Almost every change I have been part of follows the same shape. Excitement first, then real life and disappointment, and then the growth.

The excitement is real. The kickoff goes well, the tool looks good, and the first week’s numbers jump. Then real life happens. An educator who has taught a unit the same way for eight years is fast at it. The first few times through with a new platform, they are slower. Their class runs a little rougher. They are doing something hard in front of twenty-five students, and the cost of a bad day lands on them, not on the plan. That is the disappointment, and it is where most rollouts get judged.

If people get through it, the growth comes. The new way gets faster than the old one. Educators start doing things with the tool that nobody put in the training. But the growth only shows up for the people who were still there when the line turned.

A risk that has been named is a risk you can manage. A dip that was promised in August reads as progress in October. A dip nobody mentioned reads as failure, and failure gets rolled back. I have seen good tools pulled in the exact week they were about to work.

Where AI changes the math

The dip is mostly a time problem. During the slow stretch, educators are carrying the new thing on top of everything they already do. That is where AI helps in a way it couldn’t three years ago. A lot of the routine work that fills an educator’s week, a first draft of a parent email, a rubric, a quiz built from a reading, is now work an AI tool can take a real share of. Every hour it takes back is an hour that can go to getting through the slow part of something new. You can’t remove the dip. You can make it shallower and shorter by clearing room before it starts.

The practical version is plain. Before launch, tell the district leaders and the educators that the numbers will soften somewhere around weeks four to eight, and say what you will do when they do. Choose two or three measures ahead of time and read them against where you started, not against the first exciting week. Put the extra coaching in the middle of the year, where the dip is, instead of all of it in August.

In that paper I put people at the end of the risk list. I would put them first now, and I would write the dip in as a line item.

Next week: Research Theory and Planning, the course where the question behind the dissertation started to take shape.