Finally, AI That Can ‘Walk The Talk’: Agentic AI In Education

What Makes Agentic AI Different?

With a typical large language model (LLM) like ChatGPT, you can ask for the weather, and it’ll tell you if it’s raining. An AI agent, on the other hand, would take the next step. It could even order you an umbrella. This proactive behavior is what shifts AI from “knowing” to “doing.” In edtech, that means creating systems that can actively handle tasks like automating grading and building assessments on the fly.

A Firsthand Look at the Education Crisis

Outside my role at Magic EdTech, I volunteer with Junior Achievement in underserved schools around New York. I’ve seen teachers spending so much time on administrative tasks. They’re prepping coursework, outlining lessons, and grading assignments. This takes away from face-to-face teaching time and personalized attention to students.

At Magic EdTech, I’ve seen how AI (and soon, more agentic AI) can:

• Automate curriculum generation: We can create entire lesson plans or assessments aligned with a school’s objectives.
• Localize content: We can adapt media and text in 200+ languages, preserving the essence and tailoring it to different grade levels.

The goal is to give back educators the time to spend with students, individualizing instruction, and providing mentorship.

Agentic AI: A Game-Changer for Publishers

Publishers stand to benefit hugely from the shift. You can, essentially, generate an entire curriculum for a subject in mere minutes and then have a subject matter expert do a quick spot check instead of a full rewrite. That’s a big efficiency boost! It could open up new revenue streams like personalized e-learning products.

Of course, this isn’t going to happen overnight. We’ll likely see different publishers run smaller pilot projects first. Maybe publishers could try implementing an AI agent to handle a single step in content creation or distribution. Over time, as they see success, they’ll scale up. It’s a journey, one that involves some trial and error, but that’s how real innovation happens.

How to Start Building Agentic AI

I often get asked: “How do I get going with something that seems so massive?” My advice is simple:

1. Define Clear Objectives

Know exactly what problem you want to solve with agentic AI before you dive in.

2. Run Pilot Programs

Experiment on a small scale to validate if your approach works. At Magic, we do R&D with prototypes, testing both internal use cases and partner projects.

3. Focus on Infrastructure

If a pilot succeeds, make sure your IT systems can handle the new AI capabilities, whether you build them
in-house or work with a cloud provider.

4. Train Your Team

AI is only as good as the people guiding it. Make sure your developers, educators, and domain experts are communicating and have the skills to use and refine AI tools.

5. Keep Ethics Front and Center

Always involve subject matter experts. Make sure data privacy, security, and compliance are built in from the start.

6. Scale Iteratively

Roll out AI solutions step-by-step, learning and adjusting along the way.

Looking Ahead: Responsible, Proactive AI

We’re already seeing AI move beyond just churning out text. For instance, Anthropic’s Claude can control your computer to perform tasks, and ChatGPT is introducing similar functions. As AI gains more autonomy, ethical and responsible development becomes even more critical. Europe is leading with strong regulations, and I expect the rest of the world will follow suit.

At Magic EdTech, we’re building AI solutions that don’t just deliver on performance but do so in a human-centric and compliant way. We often work within a client’s own infrastructure or create isolated cloud environments to keep data and intellectual property secure. This sort of careful approach is crucial, especially as we collaborate with publishers and educational institutions.

The Real Question: Not “If,” but “How”

AI is undoubtedly reshaping education faster than many of us expected. The real question now is how we steer this transformation responsibly. If we do this right, teachers will have more time to connect with students, publishers will innovate at scale, and learners everywhere will enjoy an engaging experience.

I’m excited about the possibilities ahead. We’re moving beyond incremental improvements in AI and finally letting the technology “cook the meal” for us. And I believe that by starting small, scaling carefully, and prioritizing ethics, we can make sure agentic AI delivers on its promise.