A free demographic tool for enrollment directors
Your Next Five Kindergarten Classes Have Already Been Born
Enrollment directors: the new season is starting, and I want to share a tool I built that shows you what is happening to the potential market right around you. I also want to tell you how it came to exist, because that second part is really a story about where AI belongs in this kind of work.
Start with the tool.
The market in your own backyard
This year, while I have been learning AI, experimenting, testing, breaking things, enrollment data has been my muse. Every time I wanted to try something new, that is the data I reached for.
Two forces act on your market at the same time. How many children are being born in your county, and whether people are moving into it or out of it. Looking at them side by side gives you a much better picture of what is happening to the size of your market.
So I built a tool that shows you both for your own county. Type it in, and you will have the answer in about ten seconds.
Your next five kindergarten classes have already been born. Those children are counted, county by county, so you will get five real numbers and the shape they make to bring to your next budget meeting. Before you look up your own county, look at these two.
TX: Williamson County
District of Columbia: District of Columbia
Two markets that start within 1,400 children of each other. Four years later, they are 1,167 apart in the opposite direction. One is building a class; the other is losing one.
We hear a lot about national trends. But we all know that what is happening in your own backyard is what actually matters.
Then look up your county.
The tool also shows the second force: whether people are arriving or leaving, and what households in your market earn. Some of you will find both pointing the same way. Some will find them fighting each other.
What AI had to do with it
Here is why AI enables me to create this tool.
All year I've been building a much larger body of data work. The county lookup is one small slice of it — a simple subset, pulled out and made easy to use. Type in your county, and you have your answer in about ten seconds. The work behind it is not so simple. Pulling and reconciling the full body of data takes time and effort I'm happy to hand to AI. Some of those runs go ten to twelve hours, in the background, overnight — tens of thousands of rows of federal data pulled apart, cross-checked, and reconciled county by county.
But even if I had been willing to do all of that by hand, there were three things AI did that I could not have done alone. It synthesized at a scale my brain can't hold. The scale I mean is scope: how this data relates to everything else I am holding at the same time. It brainstormed where the data might carry real value, and just as usefully where it wouldn't, across far more possibilities than I would ever work through on my own. It validated. It checked the figures against independent sources and told me where they disagreed, which is how I learned that the federal data had quietly revised the numbers I had been working from. I also toggled between ChatGPT and Claude, asking each to audit the other's work.
It found the problems. This is the work AI was made for. When something didn't reconcile, it was dogged about figuring out why. The county whose figures don't add up. The state that abolished its counties in 2022. The pattern that simply doesn't make sense. It brought the root issue back to me and asked me to make the judgment call.
That last part is the key to the arrangement. My job is the bridge between the data and what it can validly say. Which outliers are real signals and which are artifacts. What to do with a pattern nobody can explain. Where a number would mislead a board rather than inform it. That judgment comes from knowing this world, and it isn't something I can hand off.
One of those calls is worth naming: there is no AI inside the finished tool. In my literacy and fluency work, I constantly talk about deterministic versus probabilistic tools, and this is what the distinction looks like in practice. AI did the work to find the insights, then built me something that returns the same answer every time you ask it the same question. That is a key part of the desired outcome. The number has to survive the trustee on your board who decides to check it.
The number is yours to use
So the two halves fit together like this. The tool gives you five real numbers for your own county, and they will be the same five numbers tomorrow and when a trustee looks them up. The work that produced them was done with AI, at a scale I could never reach and with a commitment to checking every number that I could not sustain. The judgment about what those numbers can and cannot say was mine, and it stays mine. I am still heavily the human in this, and I intend to stay there. But I am not the one doing the ten-hour data pulls.
There are more insights coming from me about this project. For now, , see which way the two forces are pointing, and bring the shape they make to your next budget meeting.
