Your AI Tutor Knows You're Struggling — But What Does It Do Next?
- Marielyn Wong
- Aug 13
- 4 min read
There's a version of AI in education that most people are familiar with: the student types a question, the AI answers it. A smarter search engine. A patient explainer that's available at 2am.
That's already useful. But it's not what's arriving in 2026.
The new generation of AI learning tools doesn't wait to be asked. It watches. It notices. And then it acts before the student even realises they need help.
What "agentic" actually means in a learning context
The word "agentic" is appearing everywhere in ed-tech right now. Strip back the jargon and it means this: the AI has goals, and it takes steps to achieve them without being prompted at every turn.
In a learning context, that means an AI tutor that doesn't just respond — it plans. It observes how a student is engaging with the material. It notices when a student keeps getting a particular concept wrong, or skips ahead too quickly, or loses momentum partway through a session. And then it does something about it.
This might look like generating a new practice problem targeting the exact gap it detected. Or slowing down and re-explaining a concept from a different angle. Or flagging the pattern to the educator so they can follow up.
The student doesn't need to ask "can you explain this differently?" The system already knew to try.
What changed in 2026
This isn't theoretical anymore. Canvas launched an AI teaching agent in March 2026 (Inside Higher Ed, 2026). Microsoft embedded Copilot into LMS environments as core product infrastructure earlier this year (Microsoft Tech Community, 2026). The shift from AI as a chatbot add-on to AI as an active participant in the learning process is already underway.
What's driving it isn't just better technology. It's a clearer understanding of what actually helps students learn. Passive support — available when called upon — turns out to be less effective than proactive guidance. Students who struggle often don't know they're struggling, or don't ask for help even when they do.
An agentic tutor closes that gap.
What it feels like from the learner's side
Done well, a student using an agentic AI tutor doesn't experience it as surveillance or automation. They experience it as a tutor that seems to understand where they are.
The questions get harder in the right places. The explanations come from angles that make sense for how they've been engaging. When they get stuck, the system doesn't just give them the answer — it asks them a question back, trying to find the specific point where their understanding broke down.
This is what Walter, Noodle Factory's AI tutor platform, is built around. Walter doesn't wait for a student to ask a question — it observes where each student is, remembers what they've covered, and guides them proactively based on that picture. It knows what a student has already learned in previous sessions, where they got stuck, and what kind of support helped them move forward. The result is support that feels less like a tool and more like a tutor who's been paying attention.
What educators need to think about before adopting one
Agentic AI tutors are powerful, but the design choices behind them matter enormously. A few questions worth asking before your institution rolls one out:
What content does it draw on? An AI tutor that pulls from the open internet will give different answers than one grounded in your course materials. The latter is considerably more useful — and more accurate to what you're actually teaching. Who sets the learning sequence? The best agentic systems follow the educator's logic — topics, sequence, and learning outcomes defined by the teacher, not the AI. If the AI is deciding what to teach and when, that's a meaningful transfer of curriculum control worth examining.
What visibility do educators have? An AI tutor that acts autonomously but gives educators no window into what it's doing or how students are progressing isn't really supporting the teaching relationship — it's replacing it. Full dashboard visibility matters.
Does it remember? A system that starts fresh every session can't be agentic in any meaningful sense. Tracking progress across sessions is foundational to genuinely adaptive learning.
The shift to expect
The AI tutors arriving in the next 12–18 months will be significantly more proactive than anything most educators have used. That's largely a good thing — students get better support, educators get a clearer picture of where their cohort actually is, and the gap between "students who ask for help" and "students who need it" starts to close.
The institutions that will get the most from this shift are the ones who choose platforms carefully, keep educators in the loop, and make sure the AI is working within their content and outcomes — not around them.
Sources
Noodle Factory builds AI tutoring tools for universities, polytechnics, and K-12 schools across Southeast Asia. Learn more at noodlefactory.ai.
