What Was the Leading Indicator?
A rally in April. A freeze in July. And the harder signal that never shows up on anyone's steps.
On April 16, 2026, a crowd gathered on the steps of City Hall in New York. Parents. Teachers. Students. A petition went into the mayor’s office asking for a two year moratorium on generative AI in the city’s public schools.
The coalition behind it was not one angry group. It included the Alliance for Quality Education, Class Size Matters, and Climate Families NYC. Organizations that do not ordinarily share a stage. Students spoke that day, and the framing was not about technology. It was about children.
To most of the education world, if it registered at all, it read as a protest.
It was not a protest. It was a forecast
.
Three months, three moves
Here is what followed, without commentary.
The district had a draft AI policy in hand, written in March. It permitted teachers to use AI for lesson planning and prohibited it for grading, assessment, and discipline decisions. Reasonable on its face. The kind of policy most cabinets would have been relieved to have.
In June, that policy stalled. Officials announced the delay at a City Council hearing, citing nearly 6,500 public comments on the draft. By then, more than half the City Council had signed a letter urging a pause on AI in schools.
On July 7, an email from the chancellor asked principals across the nation’s largest school district to hold new educational software purchases until the revised AI guidance is finished.
Not quite three months. From a rally on the steps to a freeze across the system.
Nobody in that room was careless
This part has to be said plainly, and it has to be sincere.
The people making those calls are serious professionals operating at a scale most of us will never touch. Roughly a million students. More than 1,600 schools. A political environment that would exhaust anyone. The decision to pause purchasing was defensible, and it was made by people trying to protect children while they got the policy right.
Any superintendent who has faced an organized community knows how fast the ground can move under you. One month you are managing a concern. The next month it is a coalition, and it has a name, and it has a press list.
So the useful question is not what New York got wrong.
The question is what the leading indicator was, and whether any of us would have recognized it in April.
How to tell a coalition from a complaint
Every district has complaints. Every superintendent’s inbox holds a dozen right now, and most of them will never become anything. That is the honest starting point, because a rule that flags everything is not a rule.
Three things separated this one.
Duration. The concern did not appear at a single board meeting. It built across months, which means someone was organizing between the meetings, in the space where leaders rarely look.
Breadth. Organizations that do not usually align were aligning. A class size advocacy group and a climate families group share no natural agenda. When they show up together, something larger has pulled them there.
Who is speaking. Students were at the microphone, and teachers stood with them. When students speak, the story stops being about policy and becomes about children, and the shape of the public conversation changes permanently.
None of the three, by itself, tells you much. Together, they are worth a cabinet’s attention long before they are worth a press release.
The signal nobody looks for
Here is where the advice usually stops, and here is where it gets dangerous.
Tell leaders to watch for organized concern and you build a system that responds to whoever can organize.
Organizing costs something. It takes time away from a shift. It takes childcare and transportation. It takes knowing which meeting matters, when it happens, and who in the building will actually listen. It takes enough experience with public officials to believe one of them might take your call.
None of that is a comment on the families. It is a description of the door we built.
The families who can get through that door are not usually the families whose children carry the most risk from a bad decision.
Which means the coalition is the easy signal.
The hard one is the silence. The school community that has every reason to be alarmed and has not said a word. The parents who assume the decision is already made. The families who have learned, over years, that showing up changes nothing, so they stop showing up.
A leader who only reads coalitions will hear from the organized and govern for them. That is not neutrality. That is a choice, made by default, and it compounds.
So the leading indicator question has a second half, and it is the harder half. Who is affected by this and has said nothing? And what does their silence actually mean?
A moment from my own superintendency stays with me here.
Our district spent seven months building a graduate profile. Community partners, administrators, teachers, parents, faith leaders, district representatives. Seven months of good faith work by people who cared deeply. Mainly adults.
As we prepared the rollout, one of my principals raised her hand in a planning conversation. Mr. Almendarez, can I ask you a question? You are asking us to create a system for kids, and there are no kids in the room.
Nobody had organized against us. No petition was coming. The students were not silent because they had given up. They were silent because we never asked. Seven months of work, and it took one principal’s question to make the absence visible.
We paused the rollout that year and spent the time in student listening sessions instead. What we heard changed the trajectory of the work. Not the edges of it. The trajectory.
Nobody was ever going to rally on our steps over a graduate profile. Without that question, the absence would have quietly become the design.
The pause guarded the front door
Something else about the freeze troubles me, and it has nothing to do with New York.
A purchase pause stops contracts. It does not stop AI.
AI arrives now through platforms a district already licenses, in features released inside products already under contract. The learning management system adds one. The assessment platform adds one. The student information system adds one.
Most districts do have review processes for this. Data privacy agreements. Vendor addenda. Approval workflows that look sound on paper.
The problem is what triggers them. Those reviews fire on a procurement event. A new contract. A renewal. A purchase order. They do not fire when a vendor ships a feature on a Tuesday.
The freeze governs procurement. Procurement is no longer where AI enters a school system.
Which means a district can halt every purchase, satisfy every demand for a pause, and still have AI moving through classrooms by the first week of September, arriving through the side door, unexamined and unapproved.
That is not a criticism of anyone’s judgment. It is the condition all of us are working in.
Taking the moratorium seriously
The coalition asked for a two year stop. Not for better governance. For a stop.
Reading their demands as a governance wish list, and quietly setting the moratorium aside, would be the convenient move. It would also be dishonest, and they would notice.
So here is a straight answer. A pause is sometimes right.
A pause is right when a district cannot say who is accountable for a tool that is already in classrooms. It is right when no one can produce evidence that a product does what the vendor claims. It is right when families were never consulted and the district cannot honestly say what it would do with their objection. Under those conditions, stopping is not weakness. Stopping is the only responsible move available.
What a pause cannot do is govern. A stop with nothing behind it buys time without buying trust, and it expires with the same questions unanswered.
The demand for a moratorium and the push from a principal who wants to move faster are asking the same question from opposite directions. Who is accountable, and what has to be true before this touches a child? One asks it in fear. The other asks it in ambition. Answer the question and both are served. Refuse it and neither is.
What an instrument looks like
At Creative Futures Collaborative, the work we do with cabinets on this is called the District AI Decision Architecture. The name matters less than the questions it forces.
Who decides. What must be true before a tool touches a student. What evidence is required, and who reviews it. Where families enter, and how early. What happens when a platform you already own turns on a feature nobody approved.
None of those require a crisis to answer. All of them get harder once a crisis arrives, because by then every answer is a public one.
One thing to do in August
Frameworks are slow. School starts in a few weeks. So here is a move that costs nothing and can be done before students arrive.
Take your three largest platform contracts. The LMS, the assessment system, the SIS. Ask one question of each vendor, in writing: what AI features have you released into our environment in the last twelve months, and which are on by default?
Then take the answer to your cabinet and ask a harder question. Who approved these, and how would we have known?
Most districts will not like the answer. That discomfort is the point. It is also the cheapest diagnostic available, and it will tell a cabinet more about its own decision architecture than any consultant could in a day.
The close
Somewhere in your community, families are already talking about this. Some of them will organize. Most of them will not.
The ones who organize will find you. The ones who do not are the ones this work is really for.
The question is not whether they show up. The question is whether you are looking for them in April or hearing from them in July.
Go find out what they are saying, and go find out who is not saying anything, while there is still runway to build something with both.
If any of this connects to something you are navigating, reply here or find me on LinkedIn. Happy to hear about it.
Jerry Almendarez Founder, CFC