You arrive in school with ideals, energy, and a stack of lesson plans. Then the bell goes. In that moment teaching stops being an idea and becomes a job with outcomes. Pupils must leave knowing more or able to do more. That is the test.

Some call teaching an art. Some call it a science. A better image is architecture. Sculpture can move people and still do nothing. Architecture must stand up, keep the rain out, and serve its users. Teaching should be beautiful when it can be, but it must work. The measure is learning.

If teaching must work, we need ways to know whether it does. In medicine no one prescribes a treatment because it feels promising. It must beat the current best option. Education is messier than medicine, but the question still stands: what works, for whom, under what conditions? That takes you past taste and habit and into evidence.

Evidence is not a guru’s thread or a glossy pitch. Strong studies show their methods and limits and invite challenge. Sample sizes matter. Conflicts of interest matter. A single study tells you little; patterns across many studies tell you more. You will meet effect sizes, control groups, and meta-analyses. Learn enough of this language to weigh claims and to defend your own choices.

Across that body of work some themes recur. Feedback helps. So does practice that is spaced and cumulative. Teaching pupils how to plan, monitor, and review their learning helps them teach themselves. Peer explanation can shift understanding. None of this is new. The craft is to make these ideas routine in your room, with your pupils, in your subject.

Where does technology fit? We have always taught with tools. Slates, textbooks, boards, projectors, laptops, and phones are all attempts to widen access, make thinking visible, and save time. The web changed who can publish and who can learn. The rise of AI changes who can get help on demand. Yet the principle holds: technology is a means, not the point. Start with learning. Ask, “Which problem am I solving?” Then choose the tool.

Used with care, tech can strengthen the core moves of teaching. It can make explanations clearer with images, models, and step-throughs. It can give pupils structured practice and immediate feedback. It can surface misconceptions at the start of a lesson. It can track patterns across a class so you can adapt tomorrow. Marking multiple-choice checks takes minutes online and can free time for richer tasks.

But tech is not a substitute for teaching. During the pandemic we learned that replacing rich interaction with screens harms equity and progress. Keep whole-class instruction, modelling, guided practice, and talk at the centre. Use tech to extend and reinforce, not to displace.

AI deserves its own note. As a teacher, AI can draft a scheme, suggest examples, vary question sets, and propose feedback stems. Treat it like a keen but flaky assistant. Check claims. Edit for tone and accuracy. As a learner support, AI can explain, question, and scaffold. That can be powerful for revision and catch-up. The risk is that pupils outsource the thinking that builds memory and skill. Set clear ground rules. Distinguish seeking an explanation from asking for the finished product. Teach pupils to use AI to plan, to check, and to practise, not to plagiarise.

What does this mean for you this term?

Plan for learning, not performance. Decide what you want pupils to know and do, and how you will see it. Build sequences that revisit key ideas and increase the level of challenge. Keep success criteria concrete. Show models and non-examples. Narrate the thinking.

Check for understanding. Use quick questions, exit tickets, or short online quizzes. Look for patterns, not marks. If half the class chose the same wrong option, reteach that step. If a few pupils are stuck, intervene while others practise.

Make feedback bite. Whole-class feedback can target common errors and save time. Digital tools can speed collection and sorting of work, but you decide what pupils do with the feedback. Give them time to act on it. Build improvement into the lesson.

Teach pupils how to learn. Model planning, self-questioning, and checking. Ask them to predict before revealing an answer. Get them to explain a choice to a partner. Invite them to set a next step. These routines build metacognition and do not need software, though software can prompt them.

Choose a small toolkit and master it. Pick one platform for quizzing, one for sharing resources, and one for collecting work. Learn the shortcuts. Set up templates. Create banks of questions that link to your curriculum. Depth beats novelty.

Mind cost and access. Not every pupil has a device or a quiet space. Offer low-tech routes to the same goal. If you set online practice, provide printed versions or time in school. If your tool needs logins, plan for the first five minutes of “I can’t get in.”

Attend to ethics and privacy. If a product tracks behaviour, ask how data is used, who sees it, and for how long. If a system nudges pupils through rewards and penalties, decide whether that aligns with your values and your school’s policy.

Stay part of a community of practice. Share what you tried and what happened. Borrow ideas, but also borrow methods of evaluation. A simple pre- and post-check can tell you if a sequence moved the dial. Over time, you build your own evidence base.

Most of all, keep the human heart of the work. You bring purpose, care, humour, and high expectations. Technology cannot see a pupil’s face fall and change tack. Evidence cannot tell you which story will land with this group on this morning. Craft bridges the gap. Craft is what you build through repeated, reflective attempts to help pupils learn hard things.

You do not need to choose art or science. You need both. Aim for lessons that stand up like good buildings and lift like good art. Let evidence guide you, let technology serve you, and let your judgement grow through use. That is a path to teaching that works.

Notes for trainees, based on the a whole cohort professional studies lecture