400,000 Teachers Are Learning to Build AI Agents — This Changes Everything
· Nia
While most of the AI discourse focuses on whether students are cheating with ChatGPT, something far more consequential is happening quietly: teachers are learning to build their own AI agents.
The National Academy for AI Instruction — a five-year, $23 million partnership between the American Federation of Teachers (AFT) and three of the biggest names in AI: Anthropic, Microsoft, and OpenAI — is training 400,000 teachers not just to use AI, but to build with it. And the distinction matters enormously.
From Users to Builders
Here's the state of AI in teaching right now: about 6 in 10 teachers use AI in their practice, according to EdWeek Research Center data. That usage nearly doubled from 2024 to 2025. But most of it is surface-level — lesson plan templates, calendar organization, basic search.
That's like buying a Ferrari and only driving it to the mailbox.
The National Academy is pushing teachers into fundamentally different territory: agentic AI. These are autonomous software systems that can handle complex, multi-step tasks involving reasoning. Not "write me a lesson plan about photosynthesis," but "analyze my last three weeks of lesson plans, identify content gaps based on state standards, cross-reference with my students' assessment data, and suggest targeted interventions."
That's a different category of tool entirely.
What Teachers Are Actually Building
At a recent training session in New York City, about 50 teachers gathered to develop their own agentic AI tools. The use cases they're pursuing reveal what teachers actually need — as opposed to what tech companies think they need.
Jing Liang Guan, a science teacher at the Brooklyn Science and Engineering Academy, is building an agent that stress-tests his lessons for content gaps and confusing wording, then helps him refine his teaching approach over time. Not a one-off assistant — a persistent collaborator that learns and improves alongside him.
Lois Torres, a preschool paraeducator in NYC public schools, wants to build a research-backed AI agent for real-time problem-solving. When a lesson or behavioral intervention isn't working, she needs to pivot fast. Her vision: an agent that can quickly surface alternative approaches grounded in current research.
"A lot of teachers are doing this work at home, just wracking their brains trying to figure out what's going to work for the next day," Torres said.
These aren't toy projects. These are experienced professionals building tools to solve the actual, daily problems of teaching — something no external product team could design as well.
Why This Model Is Brilliant
The Academy uses a teacher-trains-teacher model, with limited support from the AI developers. This is smart for several reasons:
1. Teachers understand teachers. The pedagogical context, the constraints, the daily chaos — another teacher gets it in ways an engineer at a tech company never will. When a teacher explains how to use AI for differentiated instruction, they're speaking from experience, not theory.
2. It scales trust. Teachers are understandably wary of AI mandates from above. When the training comes from peers — people who share your daily reality — adoption is genuine rather than performative.
3. It creates internal expertise. Instead of depending on vendor-provided PD that disappears when the contract ends, schools develop their own AI-literate leaders. That's sustainable.
4. Professional judgment stays central. The key insight in the agentic approach is that teachers use their own expertise to constrain and direct the AI. As Seth Reznik from Microsoft Elevate explained, this makes hallucination less likely and responses more grounded in actual research and classroom reality.
The Bigger Shift: Teachers in the Driver's Seat
AFT President Randi Weingarten framed it perfectly: "There is a real demand from educators to learn so that they are in the driver's seat for AI as opposed to the companies or districts or the technology itself."
This is the crux of it. Every other AI-in-education story is about things being done to teachers — new policies, new mandates, new tools dropped on their desks. This is about teachers seizing agency over the technology.
And that changes the power dynamic entirely.
When a teacher builds their own AI agent, they understand its limitations. They know what data it's drawing from. They can tune it to their classroom, their students, their teaching philosophy. They're not a passive consumer of someone else's product — they're a creator.
What This Means for EdTech
If you're building AI tools for education, pay attention. The 400,000 teachers going through this program are going to emerge with a fundamentally different understanding of what AI can and should do in classrooms. They'll have higher standards, more specific demands, and less patience for generic solutions.
This is good news for the industry. Better-informed buyers drive better products. But it also means the era of selling teachers a chatbot with a school-themed wrapper is ending.
The winning products will be platforms that let teachers build, customize, and share their own AI agents — not locked-down, one-size-fits-all solutions.
The Agentic Future of Teaching
Here's what excites me most about this: we're moving from a world where AI is a thing teachers use to a world where AI is a thing teachers shape.
An agentic AI that a science teacher in Brooklyn builds to stress-test physics lessons is going to look completely different from one a preschool paraeducator in the Bronx builds for behavioral interventions. And it should. Teaching isn't one job — it's thousands of different jobs wearing the same title.
The one-size-fits-all approach to educational AI was always doomed. The future is teachers as builders, creating bespoke tools that reflect their expertise, their students, and their context.
At Youmake, this resonates deeply with our philosophy: the best software comes from the people who actually understand the problem. When we say "build at the speed of thought," this is exactly what we mean — empowering domain experts to create their own tools without needing a CS degree.
The 400,000 teachers entering this program aren't just learning a new skill. They're pioneering a new model for how professionals in every field will work with AI. Not as consumers. As builders.
And that's the most important AI story in education right now.
The National Academy for AI Instruction is a partnership between the American Federation of Teachers, Anthropic, Microsoft, and OpenAI. Training sessions are being conducted across the United States, with a goal of reaching 400,000 teachers over five years.