400,000 Teachers Are Learning to Think With AI — Not Just Use It

The Most Disruptive Force in Education Isn’t AI. It’s Untrained Teachers Using It.

Here’s a number that should stop every administrator in the country: 61% of teachers are now using artificial intelligence in their professional practice.¹ That’s nearly double what it was just two years ago.

Now, you’re probably thinking, “Great — teachers are embracing AI.” But here’s what the headline misses. The vast majority of those teachers are using AI as a glorified search engine. Lesson templates. Calendar management. Administrative busywork. The kind of surface-level interaction that treats the most powerful educational technology in a generation like a slightly faster Google.

The real question isn’t whether teachers are using AI. It’s whether they’re using it in ways that actually change how students learn. And for most educators, the honest answer is: not yet.

That gap — between AI adoption and AI fluency — is what I call the Implementation Ceiling. And this week, the American Federation of Teachers launched the most ambitious effort yet to break through it.

From Template Generators to Thought Partners

On March 18, about 50 teachers gathered at the United Federation of Teachers headquarters in New York City for something that looked nothing like a typical professional development session. They weren’t learning how to write prompts. They weren’t being shown how to generate a quiz.²

They were building agentic AI tools — autonomous software systems capable of handling complex, multi-step tasks that require reasoning. Instead of asking AI to “make me a lesson plan,” these teachers were teaching AI to think alongside them about instruction.

That distinction matters more than any statistic. A teacher who uses AI to generate a worksheet has saved 15 minutes. A teacher who uses AI to stress-test a lesson for content gaps, identify where students will struggle, and suggest real-time adaptations? That teacher has fundamentally changed the quality of instruction.

Jing Liang Guan, a science teacher at the Brooklyn Science and Engineering Academy, captured the shift perfectly. He’s not interested in a tool that writes his lessons. He wants one that helps him interrogate his lessons — finding weak spots, confusing language, and missed opportunities before his students do.²

That’s not AI replacing a teacher. That’s AI amplifying professional judgment. And it’s the difference between AI as a convenience and AI as a professional competency.

The $23 Million Bet on Teacher-Led AI

The training Guan attended is part of something much larger. The National Academy for AI Instruction is a five-year, $23 million partnership between the AFT and three of the biggest names in artificial intelligence: Microsoft ($12.5 million), OpenAI ($10 million), and Anthropic ($500,000 in year one).³

The goal: train 400,000 teachers — roughly one in every ten educators in the United States — to use AI not as a shortcut, but as a professional amplifier.

Now, you might be thinking: “We’ve seen big EdTech promises before. What makes this different?”

Two things. First, the training is designed by teachers, for teachers. The academy uses educators as trainers, with limited support from developers. That means the curriculum reflects how real classrooms actually work — not how a Silicon Valley product team imagines they work.

Second, it’s free. All 1.8 million AFT members have access to virtual training through Share My Lesson, the union’s digital platform. The Manhattan facility provides hands-on, in-person workshops. No additional cost. No district budget approval required.⁴

AFT President Randi Weingarten framed the stakes clearly: “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.”²

What Sophisticated AI Use Actually Looks Like

The word “agentic” gets thrown around a lot in AI circles. But here’s what it means when a teacher builds an agentic tool instead of just typing a prompt.

Lois Torres is a preschool paraeducator in New York City. She’s developing a research-backed AI agent that helps her and her co-teacher brainstorm alternative approaches in real time — when a lesson isn’t working, when a behavioral intervention falls flat, when a child needs something different right now, not after a two-hour planning period tonight.²

“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. She’s not asking AI to do her job. She’s asking it to be a thought partner during the hardest parts of it.

Yasheema Cook teaches 12th graders in self-contained special education classes in New York City — students with autism, intellectual disabilities, and other complex needs. She’s building agents to create and monitor individualized education programs and differentiate lessons using actual daily student progress data.²

These aren’t hypothetical use cases. They’re teachers solving real problems with AI — problems that no lesson template generator was ever going to touch. And the key insight is this: the AI works better because these teachers bring professional expertise to the interaction. Their judgment narrows the AI’s scope, reduces hallucinations, and produces outputs that are genuinely useful.

That’s the paradox nobody talks about. The more skilled the human, the more powerful the AI becomes.

The Training Gap Is an Equity Gap

There’s an uncomfortable truth buried in the data. According to RAND Corporation research, the number of districts training teachers on AI doubled from 23% in 2023 to 48% in 2024.⁵ Yet the growth is deeply uneven.

Nearly all low-poverty districts are projected to have trained their teachers on AI by fall 2025. But only six in ten high-poverty districts have done the same.⁵

Read that again. The communities that could benefit most from AI-enhanced instruction are the least likely to have teachers trained to deliver it. The Implementation Ceiling isn’t just a professional development problem — it’s an equity problem hiding in plain sight.

This is exactly why the AFT academy’s free, union-wide access model matters. When training is free and available to every member — not just those in well-funded districts — it starts to close the gap instead of widening it.

The Seed: What This Means for Every Educator

Here’s the takeaway that doesn’t require a $23 million partnership or a trip to Manhattan.

The shift from basic AI use to professional AI fluency isn’t about learning new software. It’s about changing the question you ask. Instead of “What can AI create for me?” try “What can AI help me think through?”

When you stop using AI as a production tool and start using it as a thinking partner — to pressure-test your ideas, challenge your assumptions, brainstorm alternatives you haven’t considered — the technology transforms from a time-saver into a professional amplifier.

The AFT academy is formalizing that shift for 400,000 educators. But every teacher can start making it today. Ask your AI one question that isn’t about creating something. Ask it to analyze something you’ve already created. Ask it to find the weakness in your best lesson. Ask it to suggest what you might be missing.

That’s the first step through the Implementation Ceiling. And it starts right now.

Building AI fluency is a daily practice, not a one-time workshop. Get practical strategies delivered daily in the free Harvest Kernel community.

Sources

1. EdWeek Research Center, “More Teachers Are Using AI in Their Classrooms. Here’s Why,” January 2026. edweek.org
2. Sparks, Sarah D., “Teachers Move Beyond AI Basics to More Sophisticated Instructional Uses,” Education Week, March 20, 2026. edweek.org
3. CBS News, “Teachers union partners with Anthropic, Microsoft and OpenAI to launch AI-training academy,” July 2025. cbsnews.com
4. American Federation of Teachers, “AFT to Launch National Academy for AI Instruction,” July 2025. aft.org
5. RAND Corporation, “More Teachers Than Ever Before Are Trained on AI,” Education Week, April 2025. edweek.org