What the EdTech?
What the EdTech?
Building My AI Research Team
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
This week I built an AI system to help me be a better CIO — and there are lessons here for how we approach technology in our districts.
Topics Covered:
- Setting up automated daily research briefings
- COPPA compliance deadline: April 22, 2026
- Shadow IT in education and approved tool strategy
- Matching AI tools to tasks (cost optimization)
- Agentic AI trends in K-12
- Leadership filter: Does it help teachers teach or kids learn?
Key Dates:
- April 22, 2026 — COPPA compliance deadline
- March 9-12, 2026 — SXSWedu, Austin TX
- April 13-15, 2026 — CoSN, Chicago IL
Hey, it's Rob. Welcome to What the EdTech, my weekly notes to self on the intersection of technology and education. Let's get into it. So this week I did something that felt a little meta. I built an AI system to help me be a better CIO. And before you roll your eyes, hear me out. Because there's actually a lesson here about how we approach technology in our districts. And here's the setup. I've got four research jobs running automatically every morning before I wake up. At 4:30, there's conference prep. I've got Sexics Wedu in March and QSN in April. So it's pulling the latest on AI and education, what other districts are presenting, the conversations happening in our space. By 5 a.m., it's doing a broader tech industry scan. What's moving in the AI world, major acquisitions, trends that might hit K-12 in the next year or two. And 6 a.m. is the K-12 CIO intelligence brief, cybersecurity threats, data privacy updates, e-rate deadlines. And 6:30 is my ed tech briefing with a Kansas and Midwest focus. By the time I pour my coffee, I've got four reports waiting. Summaries hit my phone, full documents in my workspace. Now here's the thing, and this is the note to self. The AI isn't making me smarter. It's making me faster. There's a difference. I still have to read these reports. I still have to decide what matters. I still have to connect the dots between a COPA compliance deadline and what that means for our vendor contracts. The AI doesn't know my district. It doesn't know that we just rolled out a new SIS or that my team is stretched thin on cybersecurity audits. That context, that's my job. But what used to take me two hours of scanning news, checking LinkedIn, reading newsletters, now takes 20 minutes of focused reading. And that's the unlock. Speaking of deadlines, here's one that should be on every K-12 Tech Leader's radar April 22nd, 2026. That's when the new COPA rules kick in. And these aren't minor tweaks. We're talking major revisions to how we handle student data, especially around third-party vendors. If you haven't started auditing your vendor agreements, now's the time. Not next month, now, because 83 days sounds like a lot until you realize how many tools your teachers are using that you might not even have visibility into. And that connects to something else I've been thinking about, the shadow IT problem in education. Teachers are resourceful, they find tools that work. But every unapproved app is a potential compliance risk. So how do we balance teacher autonomy with data protection? I don't have a clean answer yet. But I think it starts with making the approved tools actually good. If we give teachers friction-free options that solve their problems, they won't need to go rogue. Yeah, that's the goal anyway. All right, pivot. Let's talk about AI costs for a second, because this matters for districts too. So I run my personal AI assistant on Claude Opus. It's expensive, like really expensive. But it's also the best at complex reasoning and understanding context. For my daily research, though, I'm running a local model on my Mac, a fine-tuned Quen model that costs me literally zero dollars per query. And for web search, I'm using perplexities API. You know, the lesson here isn't about which model is best. It's about matching the tool to the task. You know, for creative work, complex problem solving, uh, strategic thinking, yeah, I'll pay for the premium model. But for routine uh research summarization first drafts, uh a good local model handles that fine. And that's exactly how we should be thinking about AI in our districts. Not every use case needs GPT-5. Uh sometimes a simpler uh model running locally with better data privacy is the right call. The question isn't uh what's the most powerful AI? It's what's the right AI for this specific job. Conference prep is ramping up. SXS Wadoo is uh March 9th through 12th in Austin. CoSN is April 13th through 15th in Chicago. And I'm doing some edtech digest panels at CoesN, the F3 future focus forums. What I'm most interested in this year is the conversation around a genic AI. Not just chatbots that uh answer questions, but AI that can do things, take actions, run workflows. The stats I've been seeing are wild. Over 70% of enterprises are deploying generative AI in some form, but only 6% have achieved what they call full agentic implementation. That gap is where the interesting work is happening. For K-12, I think the agentic applications are going to be in operations first, instruction second, automating IT help desk tickets, um, streamlining enrollment workflows, uh, generating compliance reports, um, the stuff that eats administrative time. Um, teaching and learning applications will come, but they'll come slower, you know, as they should. We need to get uh the safety and efficacy questions uh right before we put AI directly in front of students. Uh the operations side is where we can experiment, learn, and build institutional knowledge without as much risk. Uh, one more note before I wrap up. This one's personal. I've been thinking about what it means to lead in an era where the tools are changing faster than our policies can keep up. Every week there's a new model, a new capability, a new company claiming to revolutionize education. Um my job, our job as technology leaders in schools, isn't to chase every shiny thing. It's to be translators to take this flood of innovation and filter it through the lens of does this help a teacher teach? Does this help a kid learn? Does this keep our community's data safe? Yeah, if the answer isn't yes to at least one of those, it can wait. That's my razor, that's my filter. And honestly, most things can wait. The best uh technology decision uh I made uh this week uh wasn't adopting something new. It was building a system to help me uh think about what to adopt, uh, but to be more informed, more intentional, more strategic. That's the edge. Not moving fast, um moving right. Um, that's the episode. Thanks for listening to What the EdTech. I'm Rob Dixon, and I'll talk to you next week.
