Can AI Help a Teacher Plan a Better Lesson?
AI in Business and Education — Philip M Russell Ltd
AI can write a lesson plan in seconds. That does not mean it can teach the lesson.
Imagine sitting down on a Sunday evening with several lessons to prepare for the following week.
One student needs help understanding algebra. Another is struggling with A-level Physics. A third is revising Biology but can remember facts much more easily than they can apply them in examination questions.
Each lesson needs something different. There may be demonstrations to organise, worksheets to prepare, practical equipment to check and homework to produce.
Now imagine asking artificial intelligence to help.
Within seconds, AI can produce a lesson outline, suggest starter questions, explain difficult concepts, devise practical demonstrations, create differentiated worksheets and generate homework.
That sounds rather attractive, particularly to a busy teacher.
But there is a fundamental difference between producing a lesson plan and delivering an effective lesson.
The question is not whether AI can plan a lesson. It is whether AI can help an experienced teacher plan a better one.
And I believe the answer is yes — provided we understand what AI does well, recognise its limitations and remain firmly in control of the teaching.
1. A Lesson Plan Is More Than a List of Activities
When I started teaching, lesson preparation involved textbooks, handwritten notes, worksheets, reference books and considerable time spent thinking about how best to explain a subject.
Today, we have an extraordinary range of digital resources. We can use simulations, interactive whiteboards, data loggers, videos, online examination questions and, increasingly, AI.
Yet the underlying challenge has not changed.
A good lesson must answer several important questions:
What should the student understand by the end?
What does the student already know?
What misconceptions might prevent progress?
What is the best way to explain the concept?
How will we know whether genuine understanding has developed?
What should happen if the student finds the work too easy or too difficult?
AI can suggest answers to these questions.
However, a teacher must evaluate those suggestions against the needs of the student.
In my private tuition work at Hemel Private Tuition, lessons are frequently individual rather than delivered to a classroom of thirty students.
That provides an important advantage: I can alter the direction of a lesson almost immediately.
If a student has difficulty interpreting a graph, we can spend more time on graphs. If they understand a concept quickly, we can move to more demanding questions.
The lesson plan provides a direction. The student's progress determines the journey.
2. What Happens When We Ask AI to Plan a Lesson?
Let us take a typical example.
Suppose I want to prepare a sixty-minute GCSE Physics lesson on electrical resistance.
I could give AI a very simple instruction:
The first AI prompt
CopyCreate a 60-minute GCSE Physics lesson on electrical resistance.
The resulting plan might look something like this.
Time | Activity |
|---|---|
0–10 minutes | Introduction to resistance |
10–20 minutes | Explain Ohm's law |
20–35 minutes | Work through calculations |
35–50 minutes | Complete a worksheet |
50–60 minutes | Review and homework |
There is nothing particularly wrong with this structure.
It is orderly, covers relevant material and provides a reasonable starting point.
But it is also rather uninspiring.
Where is the experiment? How do we know whether the student understands voltage and current? What happens if they struggle with rearranging equations? Is the student studying Foundation or Higher Tier? What do we do if they finish the worksheet in five minutes?
Most importantly, where is the opportunity for the student to discover something rather than simply be told about it?
This is where the teacher must start directing the AI.
3. The Secret Is Giving AI Better Instructions
AI is highly sensitive to the quality of the information it receives.
A vague instruction normally produces a generic response. A detailed instruction can produce something considerably more useful.
Suppose I improve my original request:
An improved lesson-planning prompt
CopyPlan a 60-minute one-to-one GCSE Higher Physics lesson on electrical resistance. The student knows the difference between current and voltage but struggles with rearranging equations. Include an engaging starter, a practical investigation using a variable DC power supply, ammeter and voltmeter, questions that uncover misconceptions, a graduated set of calculations, extension work and homework. Identify the intended learning outcomes and approximate timings. Include safety precautions. Do not invent examination-board requirements.
Now we have given the AI information about the student, the equipment, the challenges and the teaching approach.
The resulting lesson can be very different.
An example of a more useful 60-minute lesson
0–5 minutes — Starter: Show two components or wires of different resistance. Ask the student to predict how they might affect current
5–10 minutes — Prior knowledge: Check the meanings of voltage, current and resistance, and how meters are connected
10–30 minutes — Practical: Assemble a circuit, vary the potential difference and record current and voltage readings
30–40 minutes — Analysis: Calculate resistance and plot a current–voltage graph
40–50 minutes — Application: Work through calculations involving V = I × R, including equation rearrangement
50–55 minutes — Extension: Consider why a filament lamp does not behave in exactly the same way as a fixed resistor
55–60 minutes — Review: Test understanding, identify uncertainties and set differentiated homework
This is a much stronger starting point.
A genuine investigation also introduces a range of teaching opportunities: measurement uncertainty, graph interpretation, circuit construction, safe working and the difference between an ideal scientific model and real experimental results.
For a school class, practical group sizes and equipment availability would affect the timings. For individual tuition in a well-equipped home laboratory, the experiment can be the centre of the lesson.
We must still ensure the selected components, currents and voltages are appropriate, follow suitable practical risk assessments and avoid overheating components.
AI can propose the structure, but it cannot see the apparatus on the bench or judge whether the student is connecting it correctly.
That remains the teacher's responsibility.
4. AI Can Create Questions — But Can They Reveal Understanding?
One of the most useful applications of AI is generating questions.
A good question is not simply one that has a correct answer. It should reveal what a student understands and, sometimes more importantly, what they misunderstand.
Consider a student who has just been introduced to resistance.
AI might produce:
A resistor has a resistance of 5 ohms and carries a current of 2 amps. Calculate the potential difference.
The correct answer is:
V = I × R
V = 2 × 5 = 10 V
That checks whether the student can use a formula.
But does it establish whether they understand resistance?
Not necessarily.
I might ask AI to produce conceptual questions instead.
For example:
Question 1: If the potential difference across an ohmic resistor doubles while its temperature remains approximately constant, what happens to the current?
Question 2: Two different resistors have the same potential difference across them. One allows twice as much current to flow. Which has the greater resistance?
Question 3: A student says, "The current gets used up as it passes through a resistor." Explain why this statement is incorrect.
Question 4: Why might a filament lamp produce a curved current–voltage graph rather than a straight line?
These questions reveal much more about the student's understanding.
They also allow me to explore misconceptions that calculations alone might fail to uncover.
The important point is that AI can generate a large question bank, but the teacher decides which questions are worth asking and how to interpret the answers.
5. A Demonstration Can Be Worth a Thousand Words
One of the aspects of teaching I particularly enjoy is practical science.
There is an enormous difference between explaining what will happen in an experiment and allowing a student to investigate it.
In our home laboratory, we can use apparatus ranging from conventional laboratory equipment to digital sensors, data loggers, microscopes and high-speed cameras.
AI can help identify suitable demonstrations and suggest ways of making them educationally useful.
For example, imagine planning an A-level Physics lesson on simple harmonic motion.
A conventional lesson might involve diagrams, equations and several calculations.
AI could suggest using a mass on a spring, recording its movement and plotting displacement against time.
That is a useful idea. But we could take it further.
We might use a motion sensor to capture the oscillations, examine the graph in PASCO Capstone and compare the experimental motion with the theoretical model.
We could investigate how changing the mass affects the period and explore why real oscillations gradually decrease in amplitude.
Then we can ask the question that really matters:
Does the student's experimental evidence support the model?
AI might help suggest that sequence, but it is the teacher who must decide whether the apparatus is suitable, whether the model is being applied correctly and whether the student understands the relationship between the observations and the mathematics.
Practical experience also develops skills that AI cannot provide merely by writing instructions: handling equipment, identifying faults, recognising anomalous results and adapting methods when an experiment does not behave as expected.
And, of course, real experiments do not always go according to plan.
That can make them even more educational.
6. Pacing — The Skill That Comes with Experience
AI generally assumes that a lesson will proceed in a fairly predictable way.
Ten minutes for the introduction. Fifteen minutes for an experiment. Twenty minutes for questions.
Real lessons rarely behave so neatly.
Sometimes a student understands a concept immediately. Sometimes an apparently simple idea requires most of the lesson.
Consider a student working on GCSE algebra.
The plan might involve simplifying expressions, expanding brackets and solving equations.
Within five minutes, however, it becomes clear that the student's difficulty is not really algebra.
They are struggling with negative numbers.
An inexperienced approach might be to continue with the planned material because it is on the schedule.
An experienced teacher recognises that there is little point building more advanced work on an insecure foundation.
We might spend twenty minutes revisiting negative numbers, use a number line or visual examples and then return to algebra.
The original lesson plan has changed considerably.
But the student's learning may have improved enormously.
A successful lesson is not necessarily one in which every planned activity has been completed. It is one in which meaningful learning has taken place.
AI can help us prepare alternative explanations and supplementary material.
It cannot independently recognise the hesitation in a student's voice or the look of uncertainty when they say they understand something but clearly do not.
That is where human interaction matters.
7. Personalising Lessons for Individual Students
Perhaps the greatest potential for AI-assisted planning lies in differentiation.
Two students preparing for the same examination may need completely different approaches.
One might be highly capable but lacking confidence. Another might remember facts but struggle to interpret the language of examination questions. A third might learn most effectively through practical activities.
An AI-assisted planning process can reflect these differences, provided we give it suitable information.
For a student who struggles with lengthy questions, we might ask AI to generate a sequence that begins with short instructions, introduces key vocabulary, builds towards more complex questions and gradually removes support.
For a confident student, we might request more demanding application questions, unfamiliar contexts and extensions beyond the examination specification.
For a student who responds particularly well to demonstrations, we could request a practical-first lesson.
These are different routes towards the same underlying learning objective.
However, there is an important safeguard.
AI should not be given unnecessary identifiable information about students, particularly sensitive educational or health records. We can describe the relevant teaching needs without supplying names, addresses, diagnoses or confidential reports.
It is also important not to assume that AI can diagnose a learning difficulty or determine the most suitable educational provision.
It can suggest strategies.
The teacher must make the professional decisions.
8. Why the First AI Answer Is Rarely the Best One
This is probably the most useful lesson I have learnt from working with AI.
The first answer is a starting point, not a finished product.
The real improvement comes through discussion, criticism and refinement.
Let us imagine preparing a GCSE Biology lesson on photosynthesis.
The original AI plan begins with a fifteen-minute explanation, followed by a worksheet.
That might be acceptable, but perhaps we want something more engaging.
We can tell the AI:
Feedback to improve the plan
CopyThis lesson contains too much teacher explanation. Redesign it so the student begins with a practical observation or a prediction. Include a way of investigating photosynthesis, explain the controls and limitations, and reserve the final ten minutes for application questions. Keep the work accessible at GCSE level.
The AI now has a more precise understanding of what we are trying to achieve.
Perhaps it suggests investigating oxygen production by an aquatic plant under different light conditions.
We could improve the request again:
"Include a comparison between counting bubbles and measuring the volume of gas produced. Ask the student to explain why the two methods may give different levels of reliability."
Suddenly, the lesson involves not only photosynthesis but also experimental design, measurement, reliability and evaluation.
We could then ask for:
Five diagnostic starter questions.
A student worksheet and a separate teacher answer sheet.
Three progressively more demanding examination-style questions.
An extension activity for a student who finishes early.
Homework that tests understanding rather than simply repeating the classwork.
Each iteration can improve the plan.
But I would still check the scientific accuracy, the level of difficulty, the relevance to the examination specification and the practical feasibility.
AI can make mistakes in science, mathematics and mark schemes. It may suggest apparatus that is unsuitable or tasks that cannot realistically fit within the lesson.
Giving better feedback to AI improves the material it produces. Evaluating that material remains our responsibility.
9. Creating Homework That Builds on the Lesson
Homework is another area where AI can save significant preparation time.
One of the limitations of generic worksheets is that they may not address the particular difficulties revealed during a lesson.
Suppose a GCSE Maths student has been working on standard form.
During the lesson, the student demonstrates that they can convert numbers into standard form but struggles when multiplying and dividing them.
There is little value in setting twenty more basic conversion questions.
Instead, I can ask AI to construct homework that concentrates on the specific weaknesses:
A few straightforward questions to reinforce confidence.
Multiplication and division involving powers of ten.
Problems requiring careful handling of coefficients.
Examination-style questions requiring more than one step.
A final challenge question that combines several skills.
I can also ask for fully worked solutions and an explanation of the most likely incorrect answers.
That last request is particularly valuable.
If AI can show the common errors students might make, I can design feedback that addresses the underlying misconceptions.
However, I must check every worked solution. An answer sheet containing a convincing but incorrect calculation can be worse than having no answer sheet at all.
Good homework should also be proportionate. A sixty-minute lesson does not automatically justify another hour of homework, especially when a student has several other subjects to study.
The objective is useful practice, not simply producing more pages.
10. Can AI Help Plan Lessons Beyond the Syllabus?
This is an especially interesting possibility.
Many students benefit enormously from exploring ideas that are not strictly required by their examination.
A mathematical pattern, an unusual experiment or an unexpected connection between two subjects can generate genuine curiosity.
For example, imagine teaching GCSE Maths and introducing exponential growth.
The examination specification might require students to work with percentages and repeated multiplication.
We could extend that into an investigation of how quickly a small population grows under idealised conditions, why real populations do not grow indefinitely or how compound interest works.
AI can suggest these connections.
In Physics, a lesson on resonance might lead into musical instruments, buildings, bridges and mechanical vibration.
In Biology, an investigation of enzymes might lead into food production, biotechnology or medicine.
In Computer Science, a simple programming exercise might lead into cellular automata or Conway's Game of Life.
These extensions can help students see that their school subjects are connected to a much larger world.
But it is important to distinguish between essential examination preparation and enrichment.
A student approaching their GCSE examination may need focused support on core topics. Another with strong understanding and time available may benefit from a more exploratory approach.
Again, the teacher decides when an extension is educationally valuable.
11. What Happens When the AI Plan Meets a Real Student?
This is the ultimate test.
A lesson plan can look excellent on a screen. It can have clear objectives, beautifully structured activities and differentiated homework.
But what happens when the student arrives?
Perhaps the student has had a difficult day at school.
Perhaps they have forgotten something they understood last week.
Perhaps a demonstration generates an unexpected question that is more interesting than anything in the original plan.
I might decide that a planned ten-minute activity should become the focus of the whole lesson.
Or I might abandon a particular approach altogether.
This is not evidence of poor planning. It is evidence of responsive teaching.
One of the benefits of one-to-one tuition is the freedom to make these adjustments.
I can observe how a student approaches a question, ask why they chose a particular method and encourage them to explain their reasoning.
When they make a mistake, I can decide whether to intervene immediately or allow them to discover the error through further questioning.
I can also recognise progress that may not be reflected in a test score.
A student who previously refused to attempt a difficult problem but now confidently makes a start has achieved something significant.
That improvement may not appear in an AI-generated lesson objective.
But it matters.
12. Closing the Loop — Using AI After the Lesson
There is another opportunity that is easy to overlook.
AI can be just as useful after a lesson as before it.
Once the session is complete, I can review what happened using anonymised notes.
For example:
A post-lesson review prompt
Review this anonymised lesson summary. The student completed basic resistance calculations correctly but struggled to distinguish current from potential difference when interpreting circuit diagrams. Suggest a five-minute recap for the next lesson, three diagnostic questions and a short homework task. Avoid repeating questions the student already answered confidently.
This can help connect one lesson to the next.
Over time, it could provide a more coherent sequence of instruction than preparing each session independently.
The teaching process becomes a cycle:
This is a much more powerful way of using AI than simply asking it to generate a new worksheet every week.
It becomes a tool for structured reflection.
13. Five Mistakes to Avoid When Using AI for Lesson Planning
AI has enormous potential, but there are several traps.
Mistake 1 — Accepting the first answer
An AI-generated lesson plan should be treated as a draft. Challenge it, refine it and look for weaknesses.
Mistake 2 — Assuming every answer is accurate
Always check scientific explanations, mathematical solutions, curriculum references, examination-style questions and mark schemes. AI sometimes produces very confident errors.
Mistake 3 — Overloading the lesson
AI is excellent at suggesting activities. It can easily produce enough material for three lessons when asked to plan one. The teacher must be realistic about timings and cognitive load.
Mistake 4 — Confusing entertainment with learning
A spectacular demonstration can capture attention, but the educational value comes from the questions, predictions, observations and explanations that accompany it.
Mistake 5 — Forgetting the student
The most beautifully formatted plan is of little value if it does not respond to the individual sitting in front of us.
14. Will AI Replace Teachers?
It is a question that regularly appears whenever AI and education are discussed.
My view is that AI will change some aspects of teaching considerably.
Routine preparation may become faster. Teachers may have easier access to differentiated resources. Students may be able to receive additional explanations and practice outside lessons.
AI may also help identify patterns in student performance and suggest areas requiring attention.
But effective teaching involves considerably more than the delivery of information.
It involves motivation, trust, encouragement, professional knowledge, communication and knowing when to change direction.
An AI system does not take responsibility for a student handling electrical equipment in the laboratory. It does not reliably interpret every emotional or behavioural signal, nor can it independently establish what an individual student truly understands.
And although AI can provide useful feedback, there is still something particularly powerful about a teacher recognising the moment when a difficult idea suddenly makes sense.
I have spent around forty years teaching, and one lesson remains particularly important:
The method that works beautifully for one student may be completely unsuitable for another.
That is why teaching remains a profession requiring knowledge, experience and judgement.
AI can be a very capable assistant.
It should not be mistaken for an experienced teacher.
15. Could AI Actually Make Teaching More Human?
This may sound rather contradictory.
We introduce more technology into education, yet hope that teaching becomes more personal.
However, consider where teachers spend their time.
If AI reduces the time required to produce routine worksheets, generate alternative questions, prepare draft resources or organise teaching materials, that time might be redirected towards something more valuable.
It could provide more opportunities to think about individual students, design meaningful investigations, analyse misunderstandings or simply spend time helping someone who is struggling.
Of course, this benefit is not automatic. Teachers must still check AI-generated material, and poorly implemented technology can create additional work.
But used intelligently, AI has the potential to reduce some of the administrative burden associated with teaching.
The objective should not be to generate more material, more quickly, simply because we can.
The objective should be to improve the quality of learning.
Conclusion — The Best Lesson Is Not Written by AI
There is little doubt that AI can help teachers plan lessons.
It can generate ideas, prepare starter activities, suggest demonstrations, create differentiated questions, develop homework and provide alternative explanations.
It can respond to feedback and help refine a lesson until it becomes considerably more useful than the original draft.
But a good lesson requires more than a carefully organised set of activities.
It requires understanding the student, recognising misconceptions, adjusting the pace and making professional decisions throughout the session.
In my work at Hemel Private Tuition, I see AI as another useful addition to the teaching toolkit — alongside laboratory equipment, data loggers, textbooks, interactive displays and decades of teaching experience.
The important thing is to use it intelligently.
I would much rather spend less time producing routine worksheets and more time helping a student understand why an experiment works, discover an elegant mathematical method or overcome a difficult concept.
That, after all, is why we teach.
AI can write a lesson plan in seconds. A good teacher turns that plan into learning.
Philip M Russell Ltd — Hemel Private Tuition, Hemel Hempstead, Hertfordshire.
Individual GCSE and A-level tuition, combining experienced teaching, practical demonstrations, examination preparation and modern educational technology.
