Working

How to use AI for revision without skipping the learning

Try the question first, ask for the smallest useful hint, then attempt a new example without the answer in view. Use AI to explain a step or respond to your reasoning. Check its output against your notes, textbook, an official mark scheme where available, or your teacher. Follow the AI rules for your particular assessment.

Working is the AI study product publishing this guide. The workflow below is a suggested way to use a study assistant; it is not a claim that Working has been shown to improve grades. The examples are original practice, not official exam questions.

What should I tell an AI tutor before I start?

Give the subject, study stage and the part you already understand. For school exams, include the exam board and tier when relevant. For a university module, name the topic and assumptions; a degree label on its own is not a syllabus.

I’m studying A-level maths. I can differentiate powers, but I’m unsure when to use the product rule. Ask me one question to check the gap before explaining.
I’m taking an undergraduate linear algebra module. I can compute a determinant, but I don’t understand why eigenvectors keep their direction. Use a 2 × 2 example.

How do I ask for help without getting the whole answer?

Show your attempt and request one step. A useful request identifies the decision you are struggling with: “Which rule applies here, and why?” is more specific than “Explain everything.” If a complete solution appears, cover it and reconstruct the reasoning yourself.

Worked school example: choose and check a method

Try differentiating y = x² sin x. The mistake to watch for is differentiating each factor and multiplying the results. A first hint is to identify u = x² and v = sin x.

(uv)′ = u′v + uv′

y′ = 2x sin x + x² cos x

Explain why there are two terms before moving on. Then try y = x³ cos x without the worked answer in view. Your check is 3x² cos x − x³ sin x. A correct final expression is useful evidence, but being able to justify each term is a stronger check of the method.

Worked university example: interpret, then calculate

Take the matrix A = [[2, 1], [1, 2]]. Before computing a polynomial, apply it to the vectors (1, 1) and (1, −1). The results are (3, 3) and (1, −1), so those directions are preserved and the corresponding eigenvalues are 3 and 1.

Now connect that observation to det(A − λI) = (2 − λ)² − 1 = 0. Ask the tutor to explain the connection, rather than merely expand the determinant. For a new attempt, replace A with [[4, 1], [1, 4]]: the same two directions have eigenvalues 5 and 3.

How can I tell whether an AI explanation is wrong?

Test a step using a different representation: substitute a simple value, differentiate an antiderivative, check dimensions or units, or compare with a definition from your course. For an essay or factual claim, inspect the cited source itself. A confident tone and a plausible-looking citation do not prove accuracy.

Can I use AI on a past paper?

Released material can be useful practice when you have permission to use it. Attempt the question first and compare feedback with the correct mark scheme for that exact paper. If no matching mark scheme is provided, treat AI marking as provisional. Do not upload secure, unreleased or live assessment material.

Is this allowed for my course?

That depends on the institution and the individual assessment. Oxford’s guidance on responsible GenAI use emphasises checking outputs and following assessment-specific instructions. Its rules are an example, not a policy for every school or university. Ask your own tutor when unsure, and acknowledge permitted AI use in the way your course requires.

How does this work in Working?

Use Quick Revision for a specific question, Learn for guided questions and feedback, or the beta Tutor beside a document. Whatever the interface, the useful test is what you can do on the next problem yourself.

Written by the Working product team. These worked examples explain the intended workflow; they are not evidence from a user study or a promise of exam results. About Working and its operator.