Beyond the Shortcut. What It Really Means to Learn With AI
The LEARNERS series
L is for Learn: AI as Tutor
If you know me, you know technology is my favourite teaching tool.
For years, though, I quietly wondered if I should have learnt coding. As technology reshaped the world around me, I sometimes asked myself whether I had missed a language that mattered — whether I was always going to be the educator who loved the tools but didn't quite speak their dialect.
Then one day, a student typed something into an AI chatbot that stopped me in my tracks:
"Don't give me the answer. Ask me questions so I can improve my English."
That moment stayed with me. Not because of the technology. Because of what it revealed about learning. AI wasn't replacing thinking , it was helping a learner practise language, test ideas, and make thinking visible. The student was in charge. The AI was in service.
That reflection led me to develop the LEARNERS Framework — a learner-centred model that describes seven distinct roles AI can play in supporting language learning and deeper thinking. It builds on the work of Ethan Mollick and Lilach Mollick of the Wharton School, University of Pennsylvania, whose research proposes seven ways AI can be meaningfully assigned in education. Their aim, as they put it, is "to help students learn with AI and to help them learn about AI." The LEARNERS Framework takes that foundation and grows it into something designed for real classrooms, real teachers, and real learners right here.
Over the next seven weeks, I'll unpack each role — one letter, one idea, one cup at a time.
Because the question is no longer whether students will use AI. It's whether we design learning so they think better with it.
We start where all good learning starts: with a tutor.
The Student Who Came Back Disappointed
Last week, one of my students bounced into class with the unmistakable energy of someone who had just discovered a life hack.
"Cher! I used AI to do my homework!"
I smiled. I waited.
Twenty minutes later, she was back at my desk — worksheet in hand, three questions wrong, forehead slightly furrowed. She slid the paper towards me without a word.
"Cher... the AI gave me the wrong answer."
She looked genuinely betrayed. Like she'd trusted someone and they'd let her down. And I had to stop myself from smiling again — not because I was pleased she got things wrong, but because of what happened next. She picked up her pencil. She looked at the questions again. She started thinking.
That moment — that small, slightly deflating moment — was one of the best learning experiences she had all week. Because she used a tool, questioned its output, and had to think for herself. That's not failure. That's exactly how it's supposed to work.
The greatest misconception of our time is that AI is a shortcut to knowing things. What it actually is — at its best — is a pathway to understanding them. There's a difference. A big one. And that difference is what this article is about.
The Anxiety Belongs to Us, Not to Them
Here's something worth admitting: we adults are far more anxious about AI than the children are.
Senior Minister of State for Education Janil Puthucheary said exactly this in a recent podcast on AI in Singapore classrooms — and it landed. Children approach new tools with curiosity. They poke, they try, they fail, they try again. It's the grown-ups who freeze. Who catastrophise. Who read a headline about AI writing essays and immediately picture a generation of students who can no longer think.
But consider how many times we've been here before. When Google arrived, people feared students would stop memorising, stop thinking, stop trying. Education didn't collapse. It adapted. And as SMS Janil observed, a good education is not so different in the face of disruption — what changes is the landscape, not the fundamentals. AI is moving fast and it can now create — but that doesn't eliminate what educators bring to the table. The humanistic, the inventive, the deeply relational dimensions of teaching? Those don't get automated.
The question isn't whether AI will change education. It already has. The question is whether we meet that change with wisdom — or with panic.
The Best Tutor Doesn't Give You the Answer
Raghav Gupta, Head of Education for India and Asia-Pacific at OpenAI, put it plainly in The Straits Times: students must treat AI as a "personal tutor" rather than a shortcut to grades. And OpenAI, he noted, is deliberately introducing features designed to preserve the "friction" essential for meaningful learning.
That word friction is worth sitting with.
The best tutors have never been the ones who handed you the answer. They're the ones who asked the question, waited, and asked again. Who let you sit in the discomfort of not-knowing just long enough for understanding to click. AI, designed well, can do exactly this. It can ask why do you think that? It can say let's try that again from a different angle. It can explain the same concept seventeen different ways without sighing.
The Mollicks' research frames this well — they challenge students to "remain the human in the loop," maintaining that students are responsible not just for their own work, but for actively overseeing the AI's output, checking with reliable sources, and thinking critically about what they receive. The AI explains. The student decides. That division of labour matters enormously.
The research backs this up. A 2025 randomised controlled trial published in Scientific Reports found that a well-designed AI tutor outperformed traditional in-class active learning — students learned more, in less time, and felt more engaged and motivated. The Brookings Institution, reviewing multiple global studies, found that AI tutoring platforms delivered substantial learning gains, greater knowledge transfer, and improved motivation — providing the kind of personalised, bespoke learning that was once available only to the privileged few.
In Singapore, where households spent $1.8 billion on private tuition in 2023, that last point is not a small one.
AI Tutoring Works Differently at Different Ages — and That's the Point
Not every child is at the same stage. And one of the sharpest observations SMS Janil made in the podcast is that AI's role in learning should evolve with the learner — not be applied uniformly across all ages.
In the early primary years (P1–3), the priority isn't AI literacy. It's socialisation. Can a child take turns? Persist when things are hard? Listen, collaborate, ask for help? These are the foundational skills that no app can build.
By P4–6, something shifts. Students begin to ask — perhaps for the first time — what am I good at? What actually interests me? How do I learn best? This is where the seeds of self-directed learning are planted. And this is where AI, used thoughtfully, can start to serve a real purpose: helping students practise language skills, explore ideas, and begin to understand how to engage with a powerful tool without being passive consumers of it.
For teachers, AI can function as a genuine adjunct — a co-pilot that extends reach, personalises practice, and frees up human energy for what matters most: the conversation, the connection, the teachable moment that no algorithm can anticipate.
How, Exactly?
This is the question I get asked most. Not should we use AI. But how.
Here's what AI as Tutor actually looks like in practice — and how you can try it today.
Ask it to explain, not to answer. Instead of typing “What is the definition of reluctant?” Type "Explain the word ‘reluctant’ to me like I’m 8, then explain it again with a more precise definition and use it in a sentence. Notice what changes. Notice what you understand better in each version. The AI isn't doing the learning — it's creating the conditions for it.
Ask it to quiz you, not to tell you. A prompt as simple as "I’m practising my comprehension skills. Don’t explain the passage to me, just ask me five questions and tell me if my answers are accurate." transforms AI from a search engine into a study partner. This is precisely the Socratic method, available on demand.
Ask it to slow down your thinking. Try: "I’m going to explain why the character made this decision. Tell me if my understanding is correct, and point out anything I’ve missed." This is what my student stumbled upon instinctively — teaching the AI in order to clarify her own thinking. It's one of the most powerful learning moves available, and it costs nothing.
Use it to fill gaps, not to skip steps. Struggling with a specific part of a topic? AI can target exactly that gap, without making you sit through everything you already know. “I know how to write a composition, but I’m not sure how to make my introduction more engaging. Can you show me a few ways to start a story about a rainy day?”
Always verify. Always. This is non-negotiable. AI gets things wrong. Confidently, fluently, sometimes convincingly wrong. The habit of checking AI output against reliable sources isn't a limitation of the technology — it's the most important critical thinking skill we can build in our students right now. My student who came back with the wrong answers? She's already practising it.
The Mollicks put it simply: the goal is for AI to serve as "a supportive tool rather than a replacement" for thinking. Every prompt above is designed with that in mind. The student still does the intellectual work. The AI just makes that work more targeted, more immediate, and more responsive than a textbook ever could.
This Isn't New. Good Teaching Has Always Been Personal.
Here's the quiet truth underneath all the noise: AI as tutor isn't a revolutionary idea. It's an ancient one, made newly scalable.
The Socratic method — ask, probe, wait, redirect — is thousands of years old. The one-to-one tutoring relationship, where a knowledgeable guide adapts entirely to the learner in front of them, is what good education has always aspired to. The problem was never the idea. It was the logistics. Thirty students. One teacher. Fifty minutes. The maths never worked.
What AI changes is the ratio.
Researchers at NIE NTU Singapore observed that AI tutoring companions can pose guiding questions, assist with language practice, and help students work through multi-step problems and all while tracking progress and providing timely, encouraging feedback. They're not advocating for replacing human mentors. They're imagining extending them, making the human teacher's impact go further, reach more students, and reduce the gap between those with access to private tutoring and those without.
The best classroom of 2026 isn't the one with the most devices. It's the one where AI handles the repeatable, the explainable, and the scalable so the teacher can focus on the irreplaceable.
The "L" in LEARNERS — and Why This Is Just the Beginning
L stands for Learn with AI as Tutor — where AI explains concepts and ideas, patiently, repeatedly, differently, until understanding arrives.
But here's what the LEARNERS Framework recognises that a single letter cannot: AI is not one thing. Depending on what a learner needs in a given moment, AI can be a teammate that challenges assumptions, a tool that helps organise and structure thinking, a student that the learner teaches to clarify their own ideas, a mentor that gives feedback on writing, a coach that prompts reflection, or a simulator that generates debates and discussions.
Each role is a different relationship. Each requires a different kind of engagement. And each , when used with thought and purpose , leads somewhere deeper than a shortcut ever could.
My student who came back disappointed last week? She's already further along than she realises. She used a tool. She evaluated its output. She identified an error. She came back to think it through.
That's not a student who was failed by AI. That's a student who is learning to learn.
And that, in the age of AI, is everything.
Next week: E is for Examine — what happens when AI stops agreeing with you.
This article is part of the LEARNERS AI Framework series by Rosvinder Kaur (2026). The LEARNERS Framework describes seven ways students can use Artificial Intelligence to support visible thinking, language development, and deeper learning. Adapted from: Mollick, E., & Mollick, L. (2023). Assigning AI: Seven approaches for students, with prompts. Wharton School, University of Pennsylvania.
References:
Mollick, E., & Mollick, L. (2023). Assigning AI: Seven approaches for students, with prompts. Wharton School, University of Pennsylvania.
Raghav Gupta, The Straits Times / ST-SMU Education Series (March 2026)
Senior Minister of State Janil Puthucheary, Podcast on AI in Singapore Education (2026)
Kestin et al., Scientific Reports (June 2025)
Brookings Institution, What the research shows about generative AI in tutoring (February 2026)
Looi Chee Kit & Wong Lung Hsiang, NIE NTU / The Straits Times (April 2025)
LEARNERS Framework, Rosvinder Kaur (2026)
Linkedin Post and comments here: https://www.linkedin.com/pulse/beyond-shortcut-what-really-means-learn-ai-rosvinder-kaur-qunff/
