Classrooms are filling again. This year, teachers are walking back in with a new question weighing on their mind.
A great deal of what they read all summer framed artificial intelligence as something arriving to do their job.
Damian Creamer, founder and CEO of Primavera Online School and StrongMind, thinks that framing is wrong, and wrong in a way that produces bad products.
“Every serious version of this technology makes the teacher more important, not less,” Creamer says. “The parts of the job a machine can absorb are the parts that were never teaching. What is left when you remove them is the actual work, and right now there is nowhere near enough time for it.”
Start by Counting the Week
His case begins with an inventory most educators recognize on sight.
“Ask a teacher to account for last week honestly,” Creamer says. “Grading forty near-identical assignments. Producing three reading-level versions of the same worksheet. Moving a lesson plan into whatever template was adopted this year. Chasing attendance. Writing the same feedback sentence over and over. Pulling numbers out of systems that do not speak to each other. Almost none of that is teaching, and a fair amount of it is why people leave the profession.”
None of those tasks are unimportant. His point is that they are the tasks a machine has been capable of absorbing for years.
“We have built something strange,” he says. “The scarcest, most highly trained resource in the building spends a large share of its week on clerical work. Then we express surprise that there was no time for the student who needed twenty focused minutes on Tuesday.”
That, in his view, is where the technology belongs. Pointed at the load, not at the child.
The Part That Does Not Automate
The other half of the inventory is what Creamer describes as structurally out of reach, and he is specific about why rather than leaving it as sentiment.
“A student who has concluded he is not smart does not need a better explanation,” he says. “He needs one adult who visibly does not believe that about him and who keeps turning up anyway. That is not a capability gap in the software. The thing being asked for is a relationship, and a relationship requires somebody on the other end who can be let down and stay.”
He extends it to professional judgment. “A teacher reads a room in about four seconds. Who did not sleep. Who is nodding without following. Who is one bad interaction from giving up and needs catching before the bell. That is years of accumulated human context producing a decision about one specific child in one specific moment. I have no interest in automating it. I want to hand the teacher the time and the information to do it more often.”
Which produces the condition Creamer puts on the entire argument. Recovered time is not a saving to be banked.
“If a district uses this to cut administrative load and then raises class sizes with what it freed up, nothing has improved,” he says. “That is the same system run more cheaply. The hours have to go back into human contact or the exercise was pointless, and everyone in the building will know it was pointless by October.”
Amplification Is a Design Choice
Creamer is direct that not every product being sold to schools this fall clears that bar, and he offers a test rather than a complaint.
“Ask who the tool is aimed at,” he says. “Aimed at the teacher’s workload, it amplifies the teacher. Aimed at the student in place of the teacher, it becomes a substitute, whatever the feature is called in the deck”. Those are different products built on different assumptions about what school is, and a district should know which one it is buying before the second year of the contract.”
At StrongMind, the version he is building is deliberately unremarkable from the outside.
“The value to a teacher is mostly informational,” Creamer says. “Knowing in the second week that a student’s grip on a foundational concept is slipping, instead of discovering it in November after three units were stacked on top of it. Walking into a small group that was already grouped correctly. Not spending Sunday evening hand-building four versions of one assignment.”
In every one of those, he notes, the teacher still makes the decision. The decision is made earlier, with better information and more time than was previously available. Multiply that across a school year and you have not replaced anyone. You have given a school considerably more of the one thing it cannot go out and purchase.”
Why the Relationship Is Load-Bearing
Underneath the practical argument sits his conviction about what makes learning possible at all.
“Autonomy, competence, connection,” Creamer says. “A well-built system can support the first two directly. The third one is a person, and in a school it is almost always a teacher. There has never been a functioning version of this where that person turned out to be optional.”
Which is why he treats designing teachers out as a technical error and not only an ethical one. “You would be removing the load-bearing element and then expressing confusion when the structure does not hold.”
What It Looks Like When It Works
Creamer says the thing outsiders most often misread about a well-run online school is how human it has to be.
“People assume the online school is the automated one,” he says. “It is the reverse. Ours works because a person always catches it. The software’s entire contribution is making sure someone catches it early enough to matter, and the day we forget that ordering is the day the model stops working.”
The Better Question This Fall
His closing argument is about what the industry keeps asking.
“The question everyone is asking is what AI can do instead of a teacher,” Creamer says. “It is the wrong question and it produces products nobody in a school actually wants. The right question is what a good teacher could accomplish if you handed them back a third of their week. Build for that answer and the ceiling is far higher, and the people doing the work will meet you halfway instead of bracing for you.”
