AI and Teaching: Why I’m Not Losing Sleep
AI will change how we teach, but it won't change why we're needed.
 |
| 1961 airborne television broadcast (Rockefeller Archive Center) |
This post is a collection of my current thinking on why the replacement framing gets teaching wrong, what it misses about the fundamental nature of the profession, and why I think new teachers should have confidence that there will be plenty of work in their future. I’m not a tech futurist, but I am a mathematics educator with focus on the integration of technology. I'm also a person who has spent my career in and around schools, and is passionately committed to teacher education and supporting teachers.
A Profession Already Tested by Disruption
I have no doubt that generative AI, agentic AI, and whatever comes next will change the work of teaching just as other technologies have changed the work of teaching. For example, consider how the internet, email, and learning management systems have changed what teachers do over the last 30 years. A friend of mine who recently retired remarked one day to me how much time these different technologies take and how different this was from when he started teaching in the 90s and was using a paper grade book. It's true, the work of teaching has changed, but not the need for teachers.
A short look back at history shows that this is not the first time technology has been expected to fundamentally transform (or even replace) teachers. For example, in the 1950s and 60s, the baby boom created a need for teachers and educational television was positioned as a solution to the postwar teacher shortage. Proponents believed the medium could deliver expert instruction at scale. The image above, from the Midwest Program on Airborne Television Instruction, captures this notion—a classroom of students oriented toward a screen rather than a person. The American Federation of Teachers pushed back, resolving that television was no substitute for professional classroom practice. They were right that it did not happen. But not because television failed as a medium. I argue that it failed because the core of teaching was never primarily about content delivery. Television became a useful tool. It did not replace teachers. The pattern repeated with MOOCs in the early 2010s, which promised to render the traditional classroom obsolete and quietly became a supplement instead.
It is true that AI is genuinely different from all of these predecessors. Television was passive. MOOCs were recorded lectures with discussion forums. AI can converse, adapt in real time, and sustain something resembling a persistent relationship with a learner. The capabilities are categorically more sophisticated. But the question is not whether AI is more powerful than a television set—it obviously is. The question is whether that power crosses into the territory that actually constitutes teaching. And there the historical pattern still holds: what teachers do that matters most has never been content delivery.
The Relational Core
Teaching is a relationship profession that uses content as its medium. Those relationships extend far beyond the teacher-student connection—to peers, to families, to the broader school community. That web is not incidental to the work. It is the work.
A teacher's influence over a student's development comes from knowing that student: their hopes, their interests, their prior knowledge, their specific confusions, the moment when they're ready to be challenged versus when they need support. That diagnostic, relational judgment is built through sustained human contact. What makes it categorically different from an adaptive AI tutor is not just current capability but something harder to quantify: accountability, moral authority, and the motivation that comes from a student knowing a teacher chose to invest in them.
For anyone planning a long career right now, the question to ask is not "Will AI affect my field?" It will affect every field. A better question is: What is at the center of my work, and is this center changing as AI develops? So what's at the core of teaching? Developing relationships—work that is not easily exposed to AI methods. The most exposed work is content development and delivery, tasks that AI can assist with but that were never the heart of the job.
Compare this to software engineering, which five years ago was considered one of the safest long-term career bets. Today, entry-level developers are among the most exposed to AI displacement. Teaching's exposure runs differently: the parts that can be automated were never the core, and the core is exactly what makes the profession structurally resistant to replacement.
I view teaching as a trade. Like any trade, you get better by doing. You are a carpenter building chairs—someday you'll make wonderful ones, but first you need to make some that aren't so great. Every experienced teacher has a mental and emotional library of failed lessons, misread students, and collapsed discussions. Those failures are not incidental to expertise. They are the mechanism by which expertise develops: the ability to connect content to student interest, to anticipate how a lesson will unfold, to understand what a student needs now in light of where they have been and where they are going.
The useful distinction for novice teachers is between AI as a library and AI as a decision-maker. AI is genuinely useful for surfacing materials and brainstorming possibilities. But the fitting work—shaping general materials to specific students at a specific moment in a course sequence—is precisely where expertise gets built. The carpenter who uses a machine to build the perfect chair on the first attempt has a chair. They do not have the judgment that comes from building fifty imperfect ones.
What AI Might Change—and What It Doesn't Solve
I think there are many potential gains on the table for our profession if we thoughtfully integrate AI. The administrative load consumes teacher time without contributing to student learning. AI could reduce that friction, and experienced teachers should use it without guilt (once the very real concerns around data privacy are addressed). Planning support, differentiation scaffolding, resource curation: these are genuine productivity gains for educators who already know what good teaching looks like.
But AI does not resolve the profession's deeper structural questions. What are schools actually for? The tension between workforce training, civic education, and the cultivation of a well-rounded person is not a technical problem. Neither is the fact that many school structures prioritize compliance over actual learning. AI can assist in redesigning those structures. It cannot tell us what we are redesigning toward.
AI also does not solve equity problems. The productivity gains are most accessible to schools that already have infrastructure, training time, and institutional support. Without deliberate intervention, AI integration risks widening existing gaps. And students now arrive with AI access teachers cannot control, which changes what assessment means and what instruction needs to do.
The Bottom Line
AI is being integrated into teaching, and the pace will only accelerate. What it will not do is remove the need for teachers. I believe that the uniqueness of our profession—the teaching lifestyle that I’ve written about before—gives it a degree of insulation that other fields simply do not have. It won't stop the flood of AI integration, but I think the levee will hold. If anything, it seems to me that AI adoption should make the human-centric work of teaching even more vital and attractive to folks experiencing disruption, much like any other skilled trade.
For teachers planning 20- or 30-year careers, I think you should be reasonably confident that your place in the classroom is secure. As we figure out how to integrate these new tools, the work of centering relationships will become more critical to our role and our success.
First Published April 2, 2026
Comments
Post a Comment