Differentiated instruction with AI: three tiers of practice from one lesson, without tripling your prep

Every teacher knows the feeling of setting one worksheet for a class of thirty-odd and watching it land three different ways. Some finish in ten minutes and drift. Some are lost by question two. A handful are exactly right. Differentiation is the answer everyone agrees with and almost nobody has time for, because preparing three versions of anything triples the work. This is one of the places where AI can honestly help, provided the teacher stays in charge of what each version is for.

There's some evidence for the time saving, though it's worth reading carefully. A randomised trial run by the National Foundation for Educational Research, ChatGPT in lesson preparation: a Teacher Choices trial, involved 259 Year 7 and 8 science teachers in 68 English schools over ten weeks in summer 2024. Teachers using ChatGPT with a guide spent around 56.2 minutes a week preparing lessons and resources, compared with 81.5 minutes in the comparison group, a saving of 25.3 minutes a week, roughly 31 percent. Expert reviewers, who didn't know which resources were generated with AI, found no noticeable difference in quality. Teachers mostly used it for creating questions and quizzes and finding activity ideas, and only for one or two activities rather than across whole lessons.

Be honest about the limits of that finding. It was a UK trial in one subject and two year groups, it measured preparation time and not learning outcomes, and teachers used the tool on only part of their timetable. It can't tell you what will happen in a Pakistani school. What it does suggest is that the realistic gain comes from targeted use: pick the tasks where a draft saves the most effort, and check the output yourself. Preparing tiered practice material is a strong candidate, because it's repetitive, structured, and easy to review.

The most important principle comes before any tool: differentiate the route, not the outcome. If your lesson is built around a learning outcome, for example "students will be able to identify the parts of a plant and state their functions," every tier should aim at that same outcome. What changes is the support, the language load, and the amount of independence. This keeps the class on one syllabus and one set of expectations, and it keeps the marks comparable. A plan built on SLOs, as we describe in our guide to SLO-based lesson plans, makes this much easier, because the target is explicit before you write the first task.

Here is a workable three-tier pattern. Tier one, the supported task, uses shorter sentences, a word bank, matching or labelling, and a worked example first. Tier two, the core task, is the standard version: a mix of short-answer and fill-in-the-blank items at the level of the textbook. Tier three, the stretch task, asks students to explain why, compare two things, or apply the idea to a new situation. All three tasks should be recognisably the same lesson, so no group feels it's been given "the easy sheet." Some teachers label the tiers with neutral names or colours, and rotate students between them as understanding changes.

In practice, the workflow is short. Generate the core worksheet for the chapter. Then ask for a simpler version of the same content, with less text per question and more visual or matching items, and a harder version with two or three higher-order questions. Read all three yourself: check the facts, adjust the Urdu phrasing where it sounds stiff, and replace any example that doesn't fit your students' world. Then decide which students start where, based on what you already know from class, and revisit that decision after each task.

Muallim fits this workflow in specific, limited ways. Muallim is built by DIGIT Pakistan for teachers and school admins, and the Worksheet Builder generates worksheets at easy, medium, or hard difficulty for a chosen chapter, with mixed formats including MCQs, fill-in-the-blanks, true/false, matching, and short and long questions, in English, Urdu, or Roman Urdu, with the answer key generated together. That maps naturally onto tiers: the same chapter at three difficulty levels. Because generation is scoped to your board, the three sheets stay on the same syllabus. Every worksheet opens in an editor, so you can tune a tier before export. It doesn't decide which student belongs in which tier, and it doesn't know your class. That judgment remains yours.

Reuse is where the real saving compounds. Once you've built a set of tiered materials for a chapter, they should live somewhere you can find them next term. Content History & Reuse saves generated worksheets and lets you start a new one from an existing topic, so you aren't retyping the class, subject, and chapter each time. Over a year, a teacher who differentiates a handful of key chapters ends up with a reusable bank, which is the only realistic way differentiation survives a heavy timetable.

Two cautions before you start. First, don't let tiers become permanent labels. A student who is in tier one for fractions might be in tier three for geometry, and reading level is not the same as understanding. Second, don't put a student's name, marks, or learning profile into the tool to get a "personalised" worksheet. Ask for a version at a level, never for a named child. And if you try only one thing this week, try this: take your next planned worksheet and produce a supported and a stretch version alongside it. Measure how long it takes, and how the class responds.