A colleague asked me last week what I thought AI would look like by December. Not in some keynote-speech sense — not “what’s the future of education” — but practically. What’s actually going to land on our desks in the next six months?
It’s a fair question. The pace has been relentless. In the last year alone, we’ve seen AI models gain million-token context windows, autonomous agents that can execute multi-step workflows, and tools that adapt to individual students in real time. Teachers who were just getting comfortable with ChatGPT are now fielding questions about Claude, Gemini, and a dozen specialized education platforms they’ve never heard of.
So here’s my honest read on what the next six months look like — not from a tech conference stage, but from a classroom in Shanghai.
Agents Are Coming, But They’re Not Ready
The biggest shift happening right now is the move from AI as a tool you prompt to AI as an agent that acts. OpenAI, Anthropic, and Google are all racing to build systems that don’t just answer questions but complete tasks — booking meetings, writing and debugging code, managing workflows across multiple apps.
For schools, this matters more than most people realize. Within six months, students will have access to AI agents that can research a topic, draft an essay, find and cite sources, format the document, and submit it — all from a single instruction. That’s not speculation. The architecture already exists. What’s coming is the consumer-friendly version.
I talked to the head of IT at my school last week. He confirmed that teachers here will have agentic AI integration in their workflows within six months. Then he paused and added what I already suspected: most teachers have no idea how these systems actually work. They’re not opposed to using them. They just don’t have a mental model for what “an agent acting on your behalf” even means yet.
And the hype cycle misses something important: these agents are still unreliable. Microsoft’s own researchers found that current models lose content and corrupt outputs during long task chains. Gartner is predicting that AI agents will hit their “trough of disillusionment” this year. The technology works impressively in demos and breaks in unpredictable ways in practice.
For teachers, the takeaway isn’t “don’t worry about it.” It’s “prepare for a tool that’s powerful and inconsistent at the same time.” That combination is harder to design around than something that’s simply good or simply bad.
The Policy Scramble Is Real
While the tech accelerates, legislators are trying to catch up — and the results so far are a mess. In the first quarter of 2026 alone, 25 state legislatures introduced 52 AI education bills, and they’re often contradictory. New York wants to restrict AI in classrooms below ninth grade. Texas and Connecticut are considering banning AI from delivering instruction entirely. Ohio became the first state to require every K-12 district to adopt formal AI policies, with a July 2026 deadline. South Carolina is proposing some of the strictest guardrails in the country: written parental opt-in, annual public disclosure of AI tools, and a flat prohibition on AI replacing licensed teachers in core instruction or grading.
Meanwhile, on the other side of the world, China has mandated AI literacy as part of its national curriculum starting this fall. The gap between how different systems are approaching this is widening, not narrowing.
For teachers in international schools — which is most of my world — this creates a strange limbo. Your host country may have one set of rules, your accreditation body another, and your parent community a third set of expectations entirely. If your school doesn’t have a clear AI policy by now, the next six months are going to force one.
The Classroom Divide Nobody Talks About
Here’s the number that sticks with me: 61% of teachers reported using AI tools in 2025, nearly double the rate from two years earlier. That sounds like adoption is working. But dig into the data and the picture gets more complicated.
Seventy percent of teachers say AI weakens critical thinking and research skills. More than half of students say using AI in class makes them feel less connected to their teachers. And the digital divide — which AI was supposed to narrow — appears to be widening instead.
What I see in my own building tracks with this. Teachers who already had strong pedagogical instincts are using AI well — as a planning accelerant, a differentiation tool, a way to give faster feedback. Teachers who were struggling before AI arrived are now struggling with an additional layer of complexity. The technology doesn’t fix weak instruction. It amplifies whatever was already there.
The next six months will sort schools into two categories: those that invested in helping teachers think about AI pedagogically, and those that just handed out licenses and hoped for the best. The second group is going to have a rough fall.
What’s Actually Worth Your Attention
If I had to tell a teacher to watch three things between now and November, it would be these.
First, watch the EU AI Act’s August deadline. Starting August 2, AI-generated content must be machine-readable and clearly disclosed as synthetic. That’s going to change how AI tools present themselves to students — and it will probably surface in your classroom as a conversation about trust and verification sooner than you expect.
Second, watch what happens with student data. California just prohibited using student data to train AI models. Other states are following. This matters because many of the “free” AI education tools your school adopted last year were built on business models that depend on exactly that kind of data use. Some of those tools will disappear. Others will start charging. Either way, the free-lunch era is ending.
Third, watch the fraud problem. Experian is forecasting an explosion in AI-powered scams in 2026, and deepfake technology is now good enough that one in four job candidate profiles could be fake by 2028, according to Gartner. Your students are going to encounter AI-generated misinformation that’s far more sophisticated than anything they’ve seen before. Media literacy isn’t a nice-to-have anymore. It’s survival curriculum.
Takeaway for Teachers
Pick one of the three items above — AI agents, data privacy, or media literacy — and build a single 20-minute classroom activity around it before the end of this school year. Not a unit. Not a curriculum overhaul. One activity. Show students an AI agent completing a task and ask them where it might go wrong. Pull up your school’s AI tool list and ask who has access to their data. Show them a deepfake and ask how they’d verify it. Twenty minutes. The goal isn’t to cover everything. It’s to start the conversation before the fall semester arrives and these questions stop being theoretical.
The next six months are going to be noisy. New models, new policies, new fears, new promises. Most of that noise won’t matter to your Monday morning. But some of it will, and the teachers who are paying attention now — not panicking, just paying attention — are the ones who’ll be ready when it does.
David Jacobson is a high school history teacher. He writes about AI, education, and the messy intersection of the two at shouldiuse.ai.
