Keeping Dollars in the Classroom
Where AI Belongs in a District That Has Already Cut
"A bad system will beat a good person every time."
Deming spent most of his career trying to land a single inconvenient point. When something goes wrong inside an organization, the instinct is to blame the person closest to it. He argued, with decades of evidence from manufacturing, healthcare, and public administration, that the instinct was almost always wrong. Ninety-four percent of problems were systems. Six percent were people.
People showed up. Systems failed them.
That distinction sits underneath a conversation public education is starting to have, awkwardly, about AI. Districts have cut. Jobs are gone. Budgets are tight. And the AI conversation has arrived in the same breath as displacement fears, water and energy concerns, and well-founded worry about machine-made judgment creeping into work that should stay human. In a district that has already cut, though, the question changes. It is no longer whether AI will take a job. It is whether the people still doing the work will be left to absorb the failure of systems no one has the time or capacity to address.
A different question
A dominant framing of AI in the workplace is job replacement.
It is the right framing where staffing is stable and budgets are intact. It does not describe most districts I talk to.
I have been on calls with several superintendents in recent months where I bring up, carefully, that our AI and our services are designed to empower their team, not replace anyone. It usually gets a chuckle. Sometimes a flat-out disbelieving look. The response I have heard, in some version, more than once: "Dave, my team is lean and there is no shortage of work. I wish I had more people. You're not replacing anyone. What you're doing is helping us in a massive way when we most need it."
When an organization is already running lean and every dollar matters, effective AI paired with experienced people who have worked alongside districts across the country is an asset, not a threat. The threat is somewhere else.
In districts that have already cut, where positions sit vacant, where every available dollar is rightly going to the classroom so learning can keep happening, the question changes. The dollars are in the right place. The classrooms are doing the urgent work. What nobody has time to do is the systematic work behind the symptoms - the root-cause analysis underneath rising sub days, sliding attendance, and shifting enrollment - that would let the frontline get the right support for the right problem.
This is not an argument for AI as a fix. It is an argument for being honest about what is and isn't getting done.
The honest concerns
Large models use water and energy. Displacement is a real risk where workforces are intact. Outsourcing judgment to a model is a real temptation, and a dangerous one inside any institution whose legitimacy depends on the public being able to see how decisions are made.
Honesty cuts both ways.
The children in school this week cannot wait for the debate to resolve. Attendance is sliding. Substitute coverage is stretching schools that were already stretched. Root causes go unexamined because the people who could examine them are doing five other jobs.
The spiral doesn't pause while leaders deliberate.
The right priority, the leftover work
When budgets shrink, districts make the right call. Dollars get prioritized into the classroom. Learning is the work. Students are in the building today and need teachers in front of them today. Sub coverage gets paid for. Attendance support gets resourced. The most urgent symptoms get triaged.
What gets left undone is the systematic work behind those symptoms.
Why is sub usage rising in three middle schools and not the others? Why is attendance sliding in one cohort and holding steady in another? Why is enrollment moving the way it is, and which of those moves are correctable? This is the analytical work that use to take a research team months in a stable district. In a stretched one, it does not happen at all — not because anyone decided it didn't matter, but because the people who could do it are doing five other jobs.
So the right priority gets honored. And the root causes keep compounding.
That gap is where AI, used carefully, has something real to offer. Not by reaching into classrooms. By doing the systematic work that doesn't get done when every dollar and every person is rightly focused on the kids in front of them.
The right supports for the right problems
Deming's percentages do work here.
If most of the problems in a system come from structure rather than from effort, the right question is not how to push frontline staff harder. It is how to find the structural causes those staff are absorbing every day.
In a stretched district, that analysis rarely gets done - not because leaders don't care, but because the capacity to do it has been redirected, correctly, to keeping classrooms running. Districts without a research team can't reconstruct why attendance is falling in three schools while it holds steady in twelve others. They can't trace which substitute patterns line up with which kinds of absence. They can't test which interventions stuck and which faded.
Without that capacity, frontline staff get generic supports for specific problems. Attendance campaigns that don't match the local cause. Wellness initiatives that don't touch the structural strain. Engagement plays that respond to symptoms instead of sources.
These efforts can look like care. Often they are. But care aimed at the wrong problem rarely moves it.AI at the analytical layer can begin to close that gap. Not by deciding what to do, but by helping leaders see what is actually happening before they act. Patterns that would take a research team months can surface in hours. Causes that would otherwise stay hidden become visible early enough for the support to be designed around them.
That is what the right supports for the right problems actually means.
The spiral
When the root causes stay unaddressed, the symptoms compound.
Sub usage rises. Continuity thins. Students notice. Attendance erodes. Engagement weakens. Outcomes follow. Funding tightens. More positions get cut. The next year, even more dollars rightly go to the classroom, and even less capacity remains for the systematic work.
The spiral doesn't announce itself. It accumulates.
The intervention point is rarely at the bottom. It is earlier, at the layer where the root causes are generating the symptoms the frontline is absorbing day after day. Doing that work doesn't take dollars out of classrooms. It surfaces what the classroom can't surface on its own.
AI in this context is not a workforce play. It is a stewardship decision.
It belongs in the analytical and systematic work that doesn't get done because every available person is rightly focused on classrooms. It belongs in surfacing the root causes the frontline shouldn't have to investigate while they're teaching. It doesn't belong where children need adults, where families need to be recognized, or where decisions carry the kind of moral weight only people can carry.
Deming's point, in this moment, is clarifying. The people inside the system are not the problem. The people in classrooms are doing the urgent work, exactly as they should. The system around them has been generating problems — rising sub days, slipping attendance, leaking enrollment — whose root causes no one has the capacity to chase.
The work isn't to defend AI. It is to do the systematic work that keeps getting deferred, so the dollars stay in the classroom and the people in front of students get the right support for the right problem.
That starts with seeing the system clearly.
