On September 2, 2026, two very different organizations reached the same conclusion within twenty-four hours of each other. New York City announced a one-year moratorium on student-facing generative AI in grades K through 8, covering roughly 600,000 students, and Mayor Zohran Mamdani framed it plainly: “Children need teachers and human connection in order to learn and grow.” That same season, the Burning Glass Institute and aiEDU released a national labor-market analysis of thousands of job descriptions, finding that as AI automates routine work, employers are placing higher value on communication, collaboration, critical thinking, and creativity — the exact capacities schools have always been slowest to name and quickest to lose funding for. Policy and the labor market rarely agree on much, and the fact that they landed in the same place this week, from opposite directions, is worth sitting with. What AI can’t replace in education turns out to be what the work has always been, and educators have been carrying that knowledge quietly in the practice for a long time.
What the fast-adoption case gets right
The default in ed-tech conversations is that AI is inevitable, that resistance is protectionism, and that schools should adopt fast or fall behind — a reflex with real substance behind it. The tools can genuinely accelerate lesson design, differentiate instruction, and shave hours off administrative work that no one went into teaching to do. In a University of West Alabama poll released in May, teachers who felt poorly prepared for AI reported moderate-to-severe burnout at 72%, compared with 43% among the ones who felt prepared, and that preparedness gap is a serious workload story. The equity picture matters too: RAND’s late-2025 panel found roughly 54% of teachers were using AI for school work, and high-poverty districts were less than half as likely as low-poverty districts to have provided training (39% versus 67%). Moving slowly costs something, and the costs land unevenly. If the case ended there, faster adoption would be the answer, and the two signals from this week are the reason it doesn’t.
The rule every school technology has been judged by
Every generation of school technology raises the same question: what does the tool free teachers to do that they couldn’t before? The printing press freed them from copying passages by hand, the calculator freed them from checking arithmetic, and the learning management system freed them from managing paper. Each rung got answered by the same principle — the tool is judged by whether it protects the part of the work only a person can do. AI belongs in that lineage, and the question this week is whether we’re going to apply the principle to it or set it aside because the tool looks smarter than the ones that came before.
What actually sticks in a student’s memory
Cognitive science has been clear on this for three decades. Robert Bjork’s foundational work on “desirable difficulties” showed that conditions which feel harder during learning — spacing, retrieval practice, interleaving — produce stronger long-term memory, because reduced retrieval ease forces deeper encoding and effortful learning outperforms easy learning in retention. That research base is thirty years old and it holds across contexts.
The connection to AI is more recent. Bellwether Education Partners’ 2025 synthesis, Productive Struggle, argued that generative AI, by reducing the cognitive effort a task requires, risks eliminating the exact struggle that moves learning from short-term into long-term memory. The National Education Policy Center issued a critical review of that synthesis in October 2025, so the field is still working through the implications, but the core observation is straightforward and it’s why New York City schools chancellor Kamar Samuels framed the moratorium around protecting students’ “productive struggle” rather than around any specific tool’s flaws.
Cognitive intelligence, in the Human Intelligence framework we teach at Breathe for Change, includes critical thinking, problem-solving, and focus — competencies that don’t develop under conditions of frictionless answer retrieval. They develop through the friction itself.
The same pattern in four places at once
In K through 8 classrooms, the NYC policy protects the age band where productive struggle builds foundational literacy and numeracy, and RAND’s data shows AI adoption in schools accelerating faster than the guidance layer can catch up. The moratorium is a bet that a year of clarity is worth more than a year of undirected adoption.
In high school, Khushali Narechania, Director of Learning at aiEDU and a Breathe for Change faculty member, describes an assessment shift already underway — essays written at home, oral arguments delivered in class the next day. As she puts it, students “could have done whatever you did and had AI write everything for you. But did you actually learn it? Can you explain the why behind it?” The assessment moves toward what only a human can prove they know, and that’s the return of a much older pedagogy.
In higher education, the same shift is emerging: written work loses signal when the tool that produced it is unaccountable, and the signal comes back in real-time argument, in the ability to defend a claim under pressure.
In the workforce, the Burning Glass and aiEDU analysis found that the skills rising in value are the ones students only develop through repeated practice with other people — judgment, communication, and the ability to direct a technology well, which as Narechania framed it, requires that “you have to understand what you’re directing.” For a superintendent or department chair allocating budget, that’s the ROI case for investing in teacher development rather than only in software licenses. When policy and market and research and practice all point the same way, that’s the pattern to build around.
Where this leaves the practice
- Protect the struggle. Identify the parts of a lesson where the cognitive effort is the point and hold the line there; AI can prepare and debrief, but it doesn’t do the learning.
- Move the assessment. Ask students to defend, explain, and adapt what they’ve done in real time — that’s what separates the learner from the tool’s output.
- Teach the human skills as content. Communication, collaboration, and critical thinking are the primary asset in the labor market these students are entering, and the curriculum should treat them that way.
- Build tool literacy alongside. Refusing to teach students how AI works doesn’t protect them from it, and the choice to be empowered rather than replaced is a practice educators can grow into over time.
The course this moment requires
None of this argues for slowing down — it argues for knowing what the work is. The William Jewell College Master’s of Science in Education (M.S.Ed.) in Transformative Teaching & Learning, in partnership with Breathe for Change, includes an AI Teaching & Learning course built with Khushali Narechania that treats these capacities as daily practice for teachers rather than a theory chapter. The Cognitive layer of the Human Intelligence framework names critical thinking, focus, and problem-solving as competencies to develop deliberately in educators first, so they can develop them in students. With 20,000+ certified educators and 20M+ students impacted, the through-line has been the same the whole time: the work is human before it’s anything else. The M.S.Ed. is enrolling now.
What both signals mean
A mayor’s office arrived at it through the question of what children need to learn, and a labor-market study arrived at it through the question of what employers now pay for, and the conclusion in both cases was the same. The human side of teaching is the point, and educators have been carrying that quietly, in the practice, long before this week caught up.
“AI does have the capacity to unlock more space to do all of the human things that are so critically important that AI just can’t and will never replace.”
Narechania has been clear on this all along; she came to the field as a skeptic and stayed one long enough to build the alternative.
Want to go deeper? Listen to the full conversation between Dr. Ilana Nankin, co-founder and co-CEO of Breathe for Change, and Khushali Narechania on A Work of Heart, the Breathe for Change podcast. They cover the three meta skills of teaching with AI, what productive struggle looks like in practice, and why school as a place of convening becomes more important, not less.
