Skip to content
Joseph Gitonga
<-All notes

The same calculator, different grades: a lesson for using AI

2026-10-30 / 5 min / ai / learning / judgment / work

Everyone in a maths exam can have the same calculator and still get different grades. AI makes the same distinction visible: a powerful tool helps, but understanding, judgment and checking the work still matter.


Imagine a maths exam where every student is allowed to use the same calculator.

Same paper. Same time. Same tool.

Some students score As. Others get Bs, Cs, Ds and even Es.

The calculator helped all of them with the arithmetic. It did not give all of them the same understanding of the question, the same preparation or the same ability to notice a mistake.

That is the comparison I keep coming back to when people talk about AI replacing the need to learn a skill.

Access to the tool matters. What you can do with it still depends on how you think about the work.

The calculation can be right and the answer can still be wrong

Take a rectangular garden, six metres long and four metres wide.

You want to buy fencing. You multiply six by four and get 24. The calculator has done exactly what you asked.

But you need 20 metres of fencing around the boundary. The 24 would be useful if you were buying grass to cover 24 square metres.

The mistake happened before the buttons were pressed. You chose an area calculation for a perimeter problem.

That is why knowing which operation to use matters as much as being able to perform it. You also need to notice what the units are telling you. Metres and square metres describe different things.

AI makes this problem more interesting because it can help choose the method as well as produce an answer. It can explain, compare, draft and suggest things you had not considered. It can also misunderstand the task or give a convincing explanation of a mistake.

A calculator does not usually write a confident paragraph defending the wrong fence measurement. An AI assistant might. That is where the analogy has a limit, and where checking becomes even more valuable.

The same pattern shows up outside the classroom

A student asks AI to work through a difficult question. Reading the answer feels like understanding it. A useful check is closing the answer and trying a similar question alone. If they get stuck, they can ask for a hint, an explanation or another example. The tool becomes a way to learn the method rather than a place to collect finished homework.

A teacher asks for a lesson plan. AI can suggest a structure and exercises. The teacher knows which ideas the class struggled with yesterday, which examples will make sense to those learners and whether the suggested work fits the time available. A well-formatted plan still needs that knowledge of the room.

A shop owner asks AI to explain why sales are down. An answer based on sales totals alone might sound reasonable while missing that the most popular item was out of stock for a week. The owner can bring in the stock records, compare the same periods and test the explanation before spending money on a promotion.

Someone applying for a job asks AI to improve their CV. It can make their experience easier to read. They still need to check that every claim is true and that they could explain the work in an interview. A polished description of experience they never had creates a problem the tool will not attend the interview to solve.

A designer or writer uses AI to explore several directions quickly. Their work includes choosing what fits the audience, what sounds like the client and what deserves to be left out. Having ten options is useful. Being able to explain why one is right is part of the skill.

A software engineer asks AI to implement a feature. The generated code can be a useful start. Someone still has to establish what the feature should do, test it and decide whether it is safe to run. I have written about the work between an AI demo and a real product, because a visible result is only one part of that job.

Even at home, a weekly meal plan is only useful if it fits the people eating, the money available and what is already in the cupboard. The tool needs the circumstances of the person who will use the answer.

You can use the tool to build understanding

You do not have to become an expert before using AI. One of its useful roles is helping you work through something you do not yet understand.

Ask why a method was chosen. Ask what information is missing. Ask for a simpler example. Try changing one condition and predict what should happen before looking at the new answer.

Then check the explanation against something outside the conversation: the exercise, the source material, your records, a test or someone who knows the subject.

Asking the same assistant whether its own answer is correct can help uncover a mistake. It is not independent evidence that the answer is right.

The distinction is whether you are becoming more capable as you use it. Can you explain the result? Can you spot where it stops fitting your situation? Can you handle a nearby problem without starting from nothing again?

Some work can be handed over

AI can take over parts of a task. Repeated formatting, a routine first draft or a calculation you have already specified may need very little human involvement once the process has been checked.

It is useful to let those parts go. Keeping responsibility for an outcome does not mean personally typing every sentence or doing every sum.

The amount of review should match the consequence of a mistake. A rough list of birthday ideas and a report that determines what someone gets paid deserve different levels of checking.

Using AI well includes knowing what can run with light review, what needs a careful check and what needs another person’s expertise.

Keep the part that makes the work yours

Before accepting an answer, I want to be able to say what problem I am solving, what information the answer depends on and how I will tell whether it worked.

That applies to a student, a teacher, a business owner and an engineer. The details change. The habit carries across.

Use AI to get further than you could alone. Use it to practise, explore and remove work that no longer needs your time. Keep learning enough to question the result and take responsibility for using it.

Everyone in the exam can have a calculator. You still have to know whether the question is asking for the fence or the grass.


Read next