Three Years of Trailblazers: What 300 Teachers Taught Us About AI Readiness
They came expecting to learn AI tools. They left realizing the most effective way to teach AI Readiness often required no tools at all.
Alfred Gonzalez planned for fifteen minutes.
A sixth-grade math and science teacher in rural Washington, Gonzalez picked an AI Snapshot, one of aiEDU's short classroom discussion prompts, on AI companions: what do you think about people turning to AI for advice, for therapy, for friendship? He expected a quick warm-up before the day's real lesson.
An hour later, he had to cut his students off. "The kids could have just kept on talking and sharing and discussing with each other," he said. "That blew me away." His sixth graders were unequivocal: on relationships, emotional attachment, and especially therapy, advice and support needed to come from a human. The longer they talked, the firmer they got. Several described, from their own experience, how apps flatter and cling. "It's not your friend," is how Gonzalez sums up what he wants them to carry into middle school. "[AI]’s whole job is to keep you on that app."
AI showed up in classrooms faster than educators could plan for it. Teachers know students are using AI and that they need to do something about it without abandoning their lesson plans. But what does an adequate response look like? Banning AI? Teaching students to prompt? Treating it as a computer science problem? Teaching students fluency on today’s tools has a shelf-life problem as the tools change faster than any curriculum can, and fluency tied to what the tools look like this year may be obsolete by graduation.
Now, three years into running aiEDU’s Trailblazer Fellowship with sixteen cohorts and more than 300 K-12 teachers from the South Bronx to rural South Carolina to Kauai, we have a different answer, and it surprised many of our own fellows. Building AI Readiness is about building future-ready students, and it often has very little to do with students using AI tools at all.
Instead, it has much more to do with structured classroom conversations: how AI works, what it's for, where it doesn't belong, and what only humans can do. That’s what we mean by AI Readiness — understanding what AI is, judging when to use it and when not to, and developing the durable skills such as communication, collaboration, and judgment. When teachers led those conversations, with curriculum built for that purpose, in classrooms they know, alongside their peers, teachers and students succeeded.
The fellowship was built to test exactly that. It's a paid, 12-week virtual program with five live sessions and a closing showcase. In between, fellows implement two to three aiEDU classroom activities, recording and reflecting on them with peers. Underlying the entire program is the theory that the people best-positioned to get a school ready for AI are already in the building, and the most impactful work we can undertake is building their capacity to lead. Teachers seem to agree: this spring's cohort, our largest season yet, followed a record application cycle, with more than 600 educators applying for fewer than 175 spots. Following our Spring 2026 cohort and ahead of our next cohort launching in Fall, we wanted to share more about what we’ve learned so far and what we’re carrying into these future sessions.
Conversations first, not tools
Most fellows applied expecting AI tool training, learning which platforms to adopt, how to write better prompts, what to put in front of students. It's a reasonable expectation as it's what is currently being sold to schools as "AI professional development."
Cassie Owens Moore, a school librarian in Seneca, South Carolina, assumed the fellowship would be exactly that — "use Magic School AI, use this, use that." Leah Aiwohi, who just finished her 36th year teaching on the island of Kauai, expected the same: "Honestly, I thought it was going to be about learning more tools." Sangeetha Janakiraman, a business teacher in Michigan, came because her students were already using AI — not always well — and she wanted a say in how her district responded, plus a sense of where her own knowledge stood.
They left saying the conversation-first approach is what made the lessons work. "In no lesson are you having students on a device," Owens Moore says of the AI Snapshots and Elementary Explorations she taught. It's now her opening argument with colleagues who are skeptical about engaging with AI at all: "The pushback I get from a lot of educators is, 'I don't want to use AI in my classroom.' And I'm like, well, then you need to look at this curriculum."
Her feelings mirror the broader cohort’s. Across 123 post-program responses, 91% of fellows said the fellowship considerably or significantly increased their capacity to support AI Readiness, and 84% reported feeling confident teaching it. Savanna Dempsey, a recent fellow, explained why the conversation-first design mattered: "It removed the pressure to be a software expert and instead allowed me to be a facilitator of critical thinking."
Every lesson comes as a Google Slide or Google Doc, low-prep enough that Gonzalez says he could pull one up minutes before class, and editable so teachers can customize for their classrooms. Fellows did exactly that: only one in five taught the lessons exactly as provided; most tailored them to their students or wove them into coursework they were already teaching.
Students had more to say than expected
Once teachers led short, structured discussions about real AI dilemmas, students across every grade level and community surprised them with the depth and specificity of what they wanted to talk about. Gonzalez planned for fifteen minutes and it lasted an hour, and that turned out to be the norm: 45% of fellows reported lessons taking more time than recommended. Just 2% finished early.
Janakiraman ran an experiment with her web design students. After twelve weeks learning Bootstrap and JavaScript, she had them hand a final challenge to AI chat agents — which produced, in one class period, what would have taken them weeks. Then she asked them to change the color scheme. They couldn't, because they hadn't written the code — AI had.
Around the same time, she polled the class on a set of AI dilemmas (e.g., AI songs topping the charts, AI friends giving dating advice, etc.) and found her students answering more conservatively than she would have: AI shouldn't write songs, shouldn't take the credit, shouldn't be a friend.
So she brought both interactions to her students, side by side: you say AI shouldn't do this, and you just used it to earn credit — and now you can't fix what it made. In the discussion that followed, her students landed the principle themselves: it's okay to ask AI for help, as long as you don't outsource the learning. "When they realized it on their own, I know that it'll stick," she says, "because at least I didn't say so."
In the South Bronx, Shilpa Agrawal took an AI Snapshot on dynamic grocery pricing and built it into her AP Computer Science course: students debated the ethics, then coded the pricing model itself, calculating how it would change prices and profits at a neighborhood store. "It's rare that they have these types of discussions in my class," she said. The debate made the code feel less abstract and more relevant to their day-to-day lives. Her bar for AI in the classroom: "It needs to be a thought partner, not a thought replacer."
In Owens Moore's library, the class was a self-contained special education group with students three or four grade levels behind. She'd asked their teacher beforehand what they cared about (i.e., cars), picked the lesson on self-driving cars, and came ready to prompt them through the discussion. She didn't need to. Unprompted, one student laid out both sides: self-driving cars could be safer for a woman getting home from the airport who doesn't want to ride with a stranger, but on the other hand, that same technology could take work away from people who drive for a living.
His teacher, caught off guard, turned to Owens Moore: "Is that right?" Yes, Owens Moore told her because the lesson doesn't ask students to pick the correct answer; it asks them to weigh benefits and risks, and he had, on his own. It's why Owens Moore argues the curriculum works with nearly any student because it asks for reasoning about their own lives, not performance at grade level, so every student can participate.
Those conversations built measurable AI Readiness. We surveyed students in all five cohorts before and after the fellowship which included nine items across the three domains of aiEDU's AI Readiness framework: knowing the basics of AI, thinking critically about it, and leading with the human advantage and every cohort improved in all three. The strongest gains align with what these discussions practice: AI knowledge (+0.55 on a 5-point scale), understanding bias and fairness (+0.49), and deciding when to use AI (+0.48). Judgment, in other words — not tool use.
You don't need a device-rich classroom, or students who arrived already fluent, to build AI Readiness, just a teacher willing to lead the conversation. Alongside elementary and secondary schools, the fellowship runs a dedicated rural cohort and an Indigitize cohort — Indigenous educators working with AI Snapshots the Indigitize team adapted to reflect cultural examples and materials that are relatable to their communities. Students in those two cohorts entered with the least exposure to AI and the lowest baseline scores. If AI Readiness tracked with tool access, they should have gained the least. Yet they gained the most: +0.57 for Indigitize and +0.53 for rural, improving across every domain.
Recognizable good teaching, but not all of it
It would be fair to look at all this and say that teaching AI Readiness is just…good teaching. A lot of it is, and that's welcome news. But we found that AI Readiness work is asking teachers to stretch themselves in new ways, and our fellows are candid about them.
The first is a willingness to learn alongside students. Teachers are used to being the expert in the room, "we're not always teachable," Owens Moore says of her own profession and this work doesn't allow it, because with AI, everyone is still learning. Aiwohi, 36 years into her career, handles it by saying so out loud: "I always start out by telling students I'm learning all of this with them. It's my challenge to just keep one step ahead."
The second is knowing when AI belongs in a lesson and when it doesn't. The pressure is on both sides (e.g., ban it entirely, or embrace it everywhere) and the answer is a judgment call teachers need to make lesson by lesson. Gonzalez, for one, aims for balance, he won’t put AI tools in front of his sixth graders daily but still wants to ensure his students are prepared to engage with AI critically outside the classroom.
The third is talking honestly with students about what they're already doing outside the classroom. It's easier to police AI use than to discuss it, but the discussion is where students realize they have a say in when and how they use AI. Janakiraman put her students' stated principles next to their actual homework and asked them to account for the discrepancy in their own thinking. That conversation ended up being the one where they worked out for themselves what not to outsource. And nothing in aiEDU’s curricular resources requires settling, once and for all, whether AI is good.
Teachers don't work any of this out alone, either. Fellows learn alongside peers facing the same situations and in the post-program survey, two-thirds ranked peer collaboration among the most valuable part of the fellowship. "Being part of a community of educators who are actively thinking about how to bring AI into their classrooms made the experience feel less isolated," wrote Phillip Cox, a fellow teaching in a small Montana town. Aiwohi, teaching on Kauai, found herself comparing notes with teachers as far away as the Philippines. And once the community completes the twelve weeks, fellows join an alumni network of Trailblazers across three years of cohorts. Owens Moore says questions she posts there get answered within a day or two.
The fellowship's larger bet is that none of this stops at the classroom door and that fellows become the people who implement AI Readiness through their schools and communities. There’s evidence that’s already happening across participants. For instance, Owens Moore presented the curriculum at her state librarian conference, is earning an AI micro-credential from Clemson, and was invited to lead an AI session at a local church as she has become the person in her community who explains and leads this work. Aiwohi is sharing subject-specific Snapshots with math and social studies colleagues at her school. Janakiraman helped write her district's AI policies and plans to lead professional development when school reopens. Gonzalez has become the teacher other teachers come to and his advice starts with AI Readiness, not tools.
"Whether you're going to use AI yourself or not, whether you're going to put AI in front of your students or not, that's for you and your school to work out," Gonzalez says. "What we cannot skip is AI literacy. It has to be done, in every classroom, with every teacher. We didn't help so much with social media. We can help with AI. And I think we have to."
aiEDU’s philosophy is that the people who can get a school ready for AI are already in the building. Over three years, more than 300 Trailblazers have proved that right, one conversation at a time.
The Trailblazer Fellowship is a paid opportunity for K–12 educators to explore, test, and lead in the future of teaching with AI, made possible in part by the generous support of Google.org, Cognizant, the W.K. Kellogg Foundation, and Booz Allen Hamilton. Applications for the next cohort open August 10 — sign up here to be notified when applications are live.
