I have attended enough edtech conferences to recognize the difference between a trending topic and a paradigm shift. At ISTE 2023, held June 25-28 in Philadelphia, artificial intelligence was not a trending topic. It was the atmosphere. It permeated every keynote, every vendor booth, every hallway conversation, and every late-night discussion at the hotel bar where educators process what they have just experienced. In more than two decades of attending professional conferences, I have never seen a single technology so thoroughly dominate the discourse.

I have spent the weeks since Philadelphia processing my notes, revisiting session recordings, and synthesizing what I observed into something actionable. Here are my takeaways.

The Tone Has Shifted from Fear to Strategy

When ChatGPT launched in November 2022, the dominant emotion in education was panic. Schools banned it. Op-eds predicted the death of the essay. Teachers described feeling professionally threatened for the first time in their careers.

Seven months later, at ISTE, the panic was gone. It had been replaced by something more productive - a sense of strategic urgency. The question was no longer "Should we allow AI in schools?" but "How do we implement AI in schools in ways that are equitable, effective, and aligned with sound pedagogy?"

This shift matters enormously. Panic produces bad policy. Strategy produces workable frameworks. The educators at ISTE 2023 were overwhelmingly in strategy mode, and the quality of conversations reflected that maturity.

Session after session, I watched presenters move past the "AI is coming" framing and into the practical work of curriculum design, assessment reform, and policy development. The most valuable sessions were not the ones explaining what ChatGPT is - by June, most educators had figured that out - but the ones offering tested frameworks for integrating AI into specific subject areas, grade levels, and instructional contexts.

Three Sessions That Changed My Thinking

"AI-Resilient Assessment Design" by a team from Stanford's Graduate School of Education. This session presented a taxonomy of assessment types ranked by their vulnerability to AI circumvention. At the top (most vulnerable): formulaic essays, factual recall, and standardized problem sets. At the bottom (least vulnerable): metacognitive reflections, collaborative problem-solving with documented process, and assessments that require integration of in-class discussion. The taxonomy was immediately useful - I have already begun mapping my own assessments onto it.

A panel on AI equity that included rural, urban, and Title I perspectives. The panelists made a point I had been wrestling with all semester: AI does not just raise questions about academic integrity - it raises questions about access and justice. Students in well-resourced schools are learning to use AI as a cognitive tool. Students in under-resourced schools are being told to avoid it. If this pattern holds, AI will become another mechanism by which educational inequality compounds. The panel proposed several interventions, including school-provided AI access with scaffolded instruction, which struck me as both necessary and politically challenging.

A workshop on prompt engineering as a teachable skill. This session reframed prompt crafting not as a trick for getting AI to do your homework, but as a form of computational thinking - a structured, iterative process of communicating with a machine to achieve a desired outcome. The presenter demonstrated how teaching prompt engineering develops the same cognitive skills we value in traditional education: precision of language, logical sequencing, critical evaluation of outputs, and iterative refinement. I left that workshop convinced that prompt engineering belongs in the curriculum, not on the banned list.

The Vendor Floor Told a Story

Every major edtech vendor at ISTE 2023 had an AI angle. Every single one. Learning management systems advertised AI-powered analytics. Assessment platforms promoted AI-generated question banks. Content providers showcased AI-driven personalization engines. Accessibility tools demonstrated AI transcription and translation. The vendor floor looked like what happens when an entire industry pivots simultaneously.

I want to be measured about this observation. Vendor enthusiasm does not equal pedagogical value. Many of the AI features I saw demonstrated were incremental improvements dressed up in revolutionary language. An LMS that auto-suggests due dates based on student performance patterns is useful but not transformative. A writing tool that provides AI-generated feedback is interesting but raises immediate questions about the quality and bias of that feedback.

The vendors worth watching are the ones solving real problems that teachers have identified, not the ones manufacturing AI features to chase a trend. I left the vendor floor with a shorter list of genuinely promising tools than I expected, but the tools on that list were impressive.

What ISTE Got Wrong

The conference was not perfect. Three gaps stood out.

Insufficient attention to teacher workload. Many sessions proposed AI integration strategies that would add work to already overwhelmed teachers - new assessment designs, new literacy units, new policies to write and enforce. Almost none addressed how AI could reduce teacher workload. AI-assisted grading, AI-generated lesson plan drafts, AI-powered administrative task automation - these practical applications received far less attention than they deserved.

Limited representation from classroom teachers. Too many sessions were led by researchers, administrators, and vendors. The voices of practicing K-12 teachers - the people who will actually implement these ideas under real constraints - were underrepresented. The best sessions were invariably the ones where classroom teachers shared their semester-in-the-trenches experience.

No consensus on policy frameworks. Despite the strategic tone, ISTE 2023 did not produce anything resembling a shared framework for AI policy in schools. Different presenters offered contradictory recommendations. Some advocated for comprehensive acceptable use policies. Others argued that policies would be obsolete before the ink dried and advocated for principle-based guidelines instead. The field needs more convergence here than it currently has.

My Action Items

I returned from Philadelphia with a concrete list.

First, I am redesigning my fall assessments using the Stanford vulnerability taxonomy as a guide. Every major assessment will be evaluated for AI resilience, and the ones that score as highly vulnerable will be restructured or replaced.

Second, I am building an AI literacy module for the first two weeks of the fall semester. Students will learn what large language models are, how they work, what they do well, and where they fail. This is no longer optional preparation - it is prerequisite knowledge.

Third, I am advocating within my school for a formal AI policy that is educational rather than prohibitive. The NYC ban reversal proved that prohibition does not work. Schools need policies that set clear expectations while preserving the opportunity for guided learning.

Fourth, I am connecting with the educators I met in Philadelphia to form an informal professional learning community focused on AI in K-12. The hallway conversations at ISTE were more valuable than most formal sessions, and I intend to keep those conversations going.

The Verdict

ISTE 2023 confirmed what I suspected after my first semester with AI: this is not a fad, it is not a passing disruption, and it is not going away. The educators who treat AI as a temporary inconvenience will find themselves increasingly disconnected from the reality of their students' lives and the trajectory of their profession.

The work ahead is substantial, but Philadelphia showed me that the community of educators tackling this work is large, committed, and growing. That gives me more hope than any single technology ever could.