What does good teaching look like and how should AI support it?

AI is increasingly seen as an efficiency tool. But saving time and improving learning outcomes aren't the same thing.

Atomi's recent Tech In Schools 2025 survey backs this up: 60% of educators believe that AI edtech tools support their understanding of student progress, while 72% believe that AI is changing teaching methodologies.

In schools across the country, our best teachers are working nights and weekends trying to keep up with marking, grading and feedback. The research also shows that timely, personalised feedback is one of the most powerful ways to boost student outcomes. AI makes it possible to genuinely personalise education for every student at scale.

The challenge is ensuring a teacher's professional judgement isn't bypassed in the process. In conversation with Sarah-Eleni Zaferis (Teacher and School Enablement Leader), we explore the distinction between AI that supports teaching and AI that bypasses it, and unpack what good AI-supported learning looks like.

Watch Sarah's full interview on YouTube

What does effective teaching look like?

Good teaching is responsive to the diverse needs of learners in the classroom.

Rather than sticking to a fixed lesson plan, the best teachers understand that evidence-based teaching strategies are most effective when they're adapted to meet the student, the moment, and the energy in the room. Tweaking pacing and quick checks for understanding are what make lessons clearer, calmer, and more effective.

The skill here is noticing subtle, often unspoken, signs from learners. Educators need to spot when a student looks like they've grasped a concept, but are silently struggling, or when to push an advanced student with more challenging material.

A big part of effective teaching is relational. Students need to feel safe enough to get things wrong, because mistakes and constructive feedback are where the real learning happens. Building trust and a safe space to make mistakes takes time, and requires the skill and judgement of a human.

Good teaching isn't one fixed method. It's responsive. It's noticing when a student looks like they've understood something but actually hasn't. It's knowing when to push someone further, and when to step back and support them.
—Sarah-Eleni Zaferis

Good teaching understands how to check for understanding. Slowing the pace and extending wait times after questions gives students time to retrieve information, work through ideas and challenge themselves.

These moments also give educators a chance to see who's hesitating, who's powering ahead and who might need more support. The ability to read the room in real time is a deeply human skill, and something that can't be handed over to a tool.

Where does AI go wrong in the classroom?

There are many scenarios where AI can be used to remove the productive struggle and helpful friction that makes new knowledge stick.

• For students, using AI to generate complete answers to assessments can bypass the thinking and knowledge-retrieval process entirely. While their answers might appear polished, the student might have missed important learning milestones that will be crucial when applying new thinking in future assessments.
• For teachers, AI-generated lesson plans used without adaptation risk producing generic material that doesn't connect deeply with learners. Students aren't likely to stay engaged with material that doesn't feel tailored to them.

Students submit something polished, but they haven't engaged with the material, and as a teacher, you lose visibility into what they do and don't understand.
—Sarah-Eleni Zaferis

Consolidating knowledge is key to the teaching and learning process. While some students might grasp new information quickly, others might need differentiated instruction, additional revision materials, or extra support to help it stick.

At its best, AI can be used in these situations to spot who's starting to slip and step in sooner. Using purpose-built AI to mark responses, build curriculum-aligned quizzes and deliver immediate feedback can help tailor support to each student, making differentiation at scale a reality.

What's the difference between AI that supports teachers and AI that bypasses them?

The key distinction between AI that supports teaching and AI that bypasses it lies in understanding who is doing the thinking and decision-making.

AI that supports teachers:

• Pinpoints where students are struggling so teachers can decide how to step in and respond
• Generates resources and feedback as a starting point for teachers to review and refine
• Offers immediate help to students when they're stuck, while still asking them to do the thinking
• Keeps teachers visible and in the loop at every step, ensuring they're shaping and building on every output
• Extends what a teacher can do, without replacing the human touch or professional judgement

AI that bypasses teachers:

• Makes decisions without teacher input, ignoring the context of the class and students
• Generates complete answers for students, removing the productive struggle that builds understanding
• Produces lesson plans or material that go straight to students, without a teacher in the loop
• Removes teacher visibility; output is shared without oversight or the ability to course-correct
• Replaces a teacher's ability to connect, step in and relate to students to build rapport and trust

AI is most effective in the classroom when it enhances good teaching, creating the conditions for teaching that's more personal, more responsive, and more human.

AI that supports a teacher's judgment gives you options. AI that bypasses it makes those decisions for you, and the risk is it ignores all of the context. What these specific students need, where they're at, and how they learn best.
—Sarah-Eleni Zaferis

What does good AI-supported learning look like?

When AI is used well, students feel more supported. The immediacy and accessibility of help is a big win. Learners can get hints, personalised feedback, or targeted revision resources exactly when they need it, rather than getting stuck and waiting days to receive support.

But good AI-supported learning still needs students to work, even when the material feels sticky and challenging. It should nudge rather than spit out complete answers. It should break problems down, ask guiding questions, and prompt reflection.

The teacher still plays a central role, but they're freed up to focus on what only they can do: address misconceptions, spark discussions and facilitate moments that shift understanding.

For a student, good AI-supported learning should feel like more support, not less thinking. Students can get help when they're stuck, feedback that's actually relevant to where they are. But they're still the ones working through it.
—Sarah-Eleni Zaferis

Leveraging AI to help teachers do their best work

Good teaching is built on relationships: knowing what students are ready for, where they're stuck, and what support they need. For a long time, delivering that at scale felt like an impossible task.

That's the gap that AI (when built with intention) can begin to close. AI can work alongside educators to help them focus on the moments only they can navigate.

Ultimately, AI that's worth using in a classroom makes teaching more human, not less.

Good teaching is personal. What I hope AI can do — when it's built thoughtfully — is close that gap. More students getting the support they need, and educators with more space to do the human parts of teaching that only they can do.
—Sarah-Eleni Zaferis

More practical AI tips for teachers

If you're thinking about where to start, the practical AI Guide for Educators covers prompts and use cases that keep teachers in the loop.