AI Tools for Teachers: Master the Prompt, Then Cut Your Planning Time

Introduction: Reclaiming Teacher Time Through AI

Ask a teacher what they lack most, and the answer is rarely ideas. It is time.

You need time to plan the lesson properly. Time to read the chapter again and spot where students will get stuck. Time to build a version of the worksheet for the student who is three steps behind, and another for the one who finished early and is now distracting everyone else. Most weeks, that time does not exist. So the planning happens late at night, or it gets squeezed into a free period that was never really free.

Manish Pandey, Teacher Career Mentor and Associate Director at Suraasa, starts from exactly this problem in his webinar on AI in teaching. He does not pitch artificial intelligence as a replacement for teachers. He frames it as a teaching assistant. A capable one, but an assistant that still needs a teacher to tell it what good looks like.

That framing changes how you approach AI tools for teachers. Most guides list apps and features. This one focuses on something more durable: prompting literacy. The ability to give an AI clear, structured, pedagogically sound instructions is what turns a generic chatbot answer into a resource you would actually use in class.

In this guide, you will learn:

• The Who, What, Feel, and Do formula for everyday prompts
• The CO-STAR framework for complex, multi-layered teaching materials
• How to ground AI in your own textbook with Google NotebookLM
• How to turn one chapter into tiered quizzes, analogies, podcasts, songs, and visual diagrams
• Why the prompt, not the tool, is the skill worth building for the rest of your career

If you are new to the wider debate about AI in schools, our guide on how artificial intelligence is changing teaching gives useful background. Here, we stay practical.

The Foundational Prompt Formula: Who, What, Feel, and Do

In the full session below, Manish Pandey demonstrates every framework and tool in this article live with a Grade 8 science textbook, so you can watch each prompt turn into a finished classroom resource.

Most weak AI output comes from weak instructions. Type "make a quiz on photosynthesis" and you will get something. It will probably be too easy or too hard, pitched at no particular age, and written in a tone that suits nobody. The AI is not failing. It is guessing, because you gave it nothing to work with.

Manish Pandey's answer is a four-part structure that mirrors how good teachers already think about a lesson. He is direct about its importance:

"This formula you have to remember always. This is the only AI formula you will need for the rest of your teaching career. Everything we do today builds on this one."
Manish Pandey, Suraasa

Here is what each part asks you to define.

Notice that this is not really an AI formula. It is a lesson-planning formula. Who, What, Feel, and Do forces you to make the pedagogical decisions first, then hands those decisions to the AI to execute. That is why it works.

A worked example

Compare these two prompts.

Weak prompt: "Create an activity on photosynthesis."

Who, What, Feel, Do prompt: "I teach Grade 8 science to a mixed-ability class of 35. Several students read below grade level. The concept is photosynthesis: what a plant takes in, what it produces, and where in the leaf this happens. I want students to leave feeling curious and confident, not intimidated by the equation. Design a 10-minute pair activity where students match inputs and outputs using simple cards, followed by three exit-ticket questions that move from recall to explanation."

The second prompt takes perhaps 40 extra seconds to write. It saves you the 20 minutes you would otherwise spend rewriting a generic output.

Why the "Feel" step matters most

Many teachers skip the emotional element because it feels soft. It is not. When you tell an AI that students should feel confident rather than tested, it changes vocabulary, pacing, and the order of difficulty. When you ask for curiosity, it tends to open with a question or a surprising fact. The Feel step is where your understanding of your own students enters the prompt, and it is the part no tool can supply for you.

If you want to sharpen the "What" step, it helps to write it as a proper learning objective. Our practical guide to Bloom's Taxonomy shows how to phrase objectives at the right cognitive level, which in turn gives the AI a much clearer target.

Advanced Prompt Engineering: The CO-STAR Framework

Who, What, Feel, and Do handles most daily tasks. But some materials have more moving parts: a parent communication that must sound warm but firm, a unit overview that needs a specific format, a case study that must match an exam board's style. For these, Manish Pandey introduces a second model.

CO-STAR originated with Singapore's government technology agency, GovTech, and structures a prompt across six elements.

A worked CO-STAR prompt

Say you want a reading passage that introduces Newton's first law before a practical.

• Context: "I teach Grade 8 physics in a CBSE school. Next lesson is the first in a unit on forces and motion."
• Objective: "Write an introductory reading passage on Newton's first law of motion."
• Style: "Narrative. Open with a short everyday scene, such as a bus braking suddenly."
• Tone: "Curious and friendly, never condescending."
• Audience: "13-year-olds, some of whom are reading English as a second language."
• Response: "Around 250 words, short paragraphs, three key terms in bold, followed by two discussion questions."

Every one of those lines removes a guess. The result is far closer to usable on the first attempt.

Which framework, when?

Think of the two models as layers rather than rivals. Who, What, Feel, and Do captures the teaching decisions. CO-STAR adds precision about style, tone, and format. In practice, many teachers start with the four-part formula and reach for CO-STAR when the output must fit a specific shape.

For more prompt patterns you can adapt, see our guide to using ChatGPT for lesson planning, assessment, and productivity.

Grounding AI in Your Curriculum: Researching with Google NotebookLM

The biggest worry teachers raise about AI is accuracy. General chatbots draw on vast amounts of internet text, and they sometimes produce confident answers that are simply wrong. In a classroom, a wrong fact on a worksheet is worse than no worksheet.

This is where Manish Pandey's first tool demonstration comes in. He describes Google's NotebookLM in plain terms:

"NotebookLM is basically a research assistant. And it reads any source you give it. And then it will answer your questions only from that source, not from the internet, not from AI hallucinating."
Manish Pandey, Suraasa

The principle is called grounding. Instead of asking an AI what it knows about the water cycle, you upload your Grade 8 textbook chapter and ask it to work only from that chapter. The vocabulary matches your book. The examples match your book. The scope matches your syllabus.

How to set it up

1. Create a notebook for one unit. Keep notebooks narrow. One chapter or one unit per notebook gives sharper results than an entire year's syllabus.
2. Upload your sources. Textbook chapters, the syllabus document, your own notes, past question papers. Use materials your school is permitted to use.
3. Ask a test question first. Ask something you know the answer to, such as "What are the three stages of the water cycle described in this chapter?" Check that the answer matches the text.
4. Then generate. Once you trust the grounding, ask for quizzes, analogies, mind maps, or summaries.

In the webinar, Manish shows how quickly this produces aligned multiple-choice questions, simpler analogies for students who are struggling, and mind maps of the chapter's structure.

A note on responsible use

Grounding reduces the risk of invented facts. It does not remove your role as the final checker. The U.S. Department of Education's Office of Educational Technology report on AI and the future of teaching and learning argues for keeping humans in the loop, with educators making the final call on what reaches students. That is exactly the teaching-assistant model Manish describes.

Data protection matters too. The UK Department for Education's position on generative AI in education urges schools to protect personal and sensitive data when using these tools. A simple rule: upload curriculum content, not student names, marks, or records. Check your school's policy before you begin.

Differentiating Content: Generating Tiered Assessments and Analogies

Differentiation is where AI saves the most time, because it is the task teachers most often have to cut. Writing one good quiz takes effort. Writing three versions at different levels of challenge, every week, for every class, is not realistic for most of us.

Once your chapter sits inside NotebookLM, you can ask for tiers directly. Here is a prompt built on the Who, What, Feel, and Do formula:

"Using only this chapter, create three versions of a 6-question quiz on the water cycle for Grade 8. Version A is for students who need support: recall questions with simple language and one picture-based question. Version B is for on-level students: a mix of recall and explanation. Version C is for students ready for stretch: application and 'what would happen if' questions. Students should feel that each version is achievable. Include an answer key for each."

What good tiering looks like

The key is that all three tiers assess the same concept. You are changing the route, not the destination. This keeps the class together and avoids the trap of giving some students permanently easier work.

Analogies for struggling learners

Manish also demonstrates asking NotebookLM for simplified analogies. When a student cannot picture condensation, a concrete comparison often unlocks it: the water droplets that form on a cold glass of water on a humid day. Ask the tool for three analogies at different levels of abstraction, then choose the one that fits the student in front of you. That final choice is professional judgement, and it stays with you.

Use the results formatively

Tiered quizzes are most powerful when they feed your next lesson rather than a grade book. Our guide to assessment for learning versus assessment of learning explains how to use quick checks to adjust teaching in real time. If you want more ideas for short checks, see our list of 25 classroom assessment techniques.

Audio Pedagogies: Creating Custom Study Podcasts for Auditory Learners

Not every student absorbs a chapter by reading it. Some understand far more when they hear an idea explained aloud, especially as a conversation. Until recently, producing audio for a class meant recording it yourself, which few teachers have time to do.

Manish Pandey shows how NotebookLM's audio overview features change that. From the same uploaded chapter, a teacher can generate a structured, conversational podcast or audio guide that walks through the material. You can tailor it to a specific age group, so the explanation for Grade 6 sounds different from the explanation for Grade 10.

Practical ways to use audio overviews

• Pre-listening homework. Students listen to a short overview before the lesson, so class time goes on discussion and practice rather than first exposure.
• Revision on the move. Many students have long commutes. A chapter summary they can listen to on the bus adds revision time without adding screen time.
• Support for students who struggle with reading. An audio version of the chapter lets them access the science content even when the text itself is a barrier.
• Absence catch-up. A student who missed three days can hear the key ideas before they return.

Prompting the audio well

Even here, the formula applies. Tell the tool who the listeners are, what they need to understand, how they should feel at the end, and what you want them to do afterwards. For example: "Create an audio overview of this chapter for Grade 8 students who find science intimidating. Focus on the three stages of the water cycle. Keep the tone encouraging. End by asking listeners to notice one example of condensation at home."

Always listen to the full episode before sharing it. Check accuracy, check pacing, and check that the examples suit your students' context.

Enhancing Memory Retention with AI-Generated Educational Music

Think back to your own school days. You may have forgotten most of the facts you memorised for exams. But you probably still remember a rhyme, a jingle, or a song you learned around the same time.

Manish Pandey builds his third demonstration on exactly that observation:

"Because music and memory, they are wired together in the human brain. So we don't forget songs. We can forget a lot of factual information, but songs, they stay forever."
Manish Pandey, Suraasa

Using Suno AI, an AI music generator, he shows how teachers can turn a concept into a complete song. In the webinar, he uses the water cycle, with its multi-step sequence of evaporation, condensation, and precipitation, and Newton's laws of motion. Teachers can choose the genre too, so the same concept can become a pop track, a rap, or a folk-style song depending on what will land with the class.

Where songs work best

Music is especially useful for content that has a fixed order or a precise wording students need to recall:

• Multi-step scientific cycles and processes
• Mathematical formulas and rules
• Grammar rules and verb conjugations
• Historical timelines and sequences
• Lists such as the planets, the parts of a cell, or the states of matter

A sample prompt

"Write an upbeat pop song for 12 to 13-year-olds about the water cycle. Name each stage clearly in the chorus: evaporation, condensation, precipitation, collection. Each verse should explain one stage in simple, accurate language. Keep it under two minutes so students can learn it in one lesson."

Check the lyrics like you would check a worksheet

Song lyrics are still AI-generated text. A rhyme can bend a fact out of shape. Read the lyrics carefully before you play the song to a class, and correct any line that oversimplifies to the point of being wrong. A catchy mistake is still a mistake, and it will be very hard to unlearn.

One effective routine: play the song at the start of the unit, again midway, and once more before assessment. Then ask students to explain each line in their own words. The song gets the sequence into memory. The explanation makes sure the understanding is there too.

Visualizing Lesson Concepts: Automated Diagramming with Napkin AI

Some ideas are much easier to grasp as a picture than as a paragraph. A process with stages, a comparison between categories, a cause leading to an effect: all of these become clearer when students can see the structure.

Drawing a clean diagram, though, takes time and some design skill. Manish Pandey's final tool demonstration tackles that problem with Napkin AI, which converts descriptive text into visual diagrams, flowcharts, and infographics.

In the webinar, he pastes in text explaining photosynthesis and gets back a structured process diagram. He then does the same with an explanation of the types of fractions, producing a visual that sorts proper, improper, and mixed fractions clearly.

How to get a strong diagram

1. Write or paste clear, sequenced text. The diagram can only be as organised as the text you give it. Use one idea per sentence and signal order with words like "first", "next", and "finally".
2. Generate several options. Napkin AI typically offers different visual layouts. A flowchart suits a process. A comparison grid suits categories.
3. Edit labels. Make sure the terms match your textbook exactly, so students are not confused by synonyms.
4. Simplify. Remove anything that does not serve the learning objective. A diagram with 12 boxes is rarely better than one with five.

Combining tools into one workflow

This is where the pieces fit together. You can use NotebookLM to produce a grounded, accurate explanation of a process from your textbook, then paste that explanation into Napkin AI to turn it into a diagram. The first tool handles accuracy. The second handles visual design. You handle the judgement in between.

Diagrams also make excellent anchors for a structured lesson. If you plan with the 5E model, a visual works well in the Explain phase, after students have explored the idea themselves. Our complete guide to the 5E lesson plan model shows where each type of resource fits best.

Dictation Workflows: Speeding Up Prompting with Speech-to-Text

By now, one thing should be clear. Good prompts are detailed. A strong Who, What, Feel, and Do prompt might run to 80 or 100 words. A full CO-STAR prompt can be longer. For many teachers, typing all of that feels like it cancels out the time saved.

There is a simple fix: speak your prompts instead of typing them.

Most of us talk far faster than we type. And when teachers speak, they naturally describe their students the way they would to a colleague in the staff room, with exactly the kind of context an AI needs. "My Grade 8 group is lively but loses focus after lunch, and a few of them panic the moment they see an equation." That sentence is rich prompt material, and it took five seconds to say.

Tools you probably already have

• Windows: built-in voice typing, opened with the Windows key and H.
• Mac: built-in Dictation, turned on in system settings.
• Google Docs: Voice typing under the Tools menu in the Chrome browser.
• Phones: the microphone icon on most smartphone keyboards.
• AI apps: many chatbots now include a voice input button directly in the chat window.

A simple dictation routine

1. Open a blank document and dictate your prompt using the four headings: Who, What, Feel, Do.
2. Read it back once and fix any words the software misheard, especially subject terms.
3. Paste it into the AI tool.
4. Save the best prompts in a running document, organised by unit, so you can reuse and adapt them next year.

That last step is quietly powerful. Over a term, your saved prompts become a personal planning library. Each one carries your knowledge of your students and your curriculum, and each one gets faster to adapt.

For more ways to build efficient routines around AI, see our guide to AI tools for educator productivity beyond basic ChatGPT prompts.

Why Prompt Engineering, Not the Tool, Is the Lifelong Skill

It is tempting to finish a session like this and think the takeaway is a list of apps: NotebookLM, Suno AI, Napkin AI. Manish Pandey closes his webinar by arguing the opposite.

"The tool is not the skill. The prompt is the skill."
Manish Pandey, Suraasa

His reasoning is practical. Within five years, he expects half of today's tools to be replaced by better ones. Interfaces will change. Features will move. Some of the apps in this article may not exist in their current form. But a teacher who can think clearly about who the students are, what they need to learn, how they should feel, and what they will do will be able to make any new tool useful on day one.

That is a reassuring position, because it puts the teacher back at the centre. The skill that makes AI work in a classroom is not technical. It is pedagogical. Knowing your learners, setting a clear objective, and designing a meaningful task have always been the core of teaching. Prompting simply asks you to write those decisions down.

What global bodies are saying

This view sits comfortably alongside the wider international conversation. UNESCO's work on AI competencies for teachers treats AI capability as part of a teacher's professional practice, with human agency and ethics at its heart, rather than a narrow technical add-on. The OECD Digital Education Outlook 2023 examines how digital tools can support teaching when schools build the right capacity around them. And the Brookings Institution's analysis of students in an AI world frames the challenge around helping young people prosper, prepare, and stay protected, which depends heavily on teachers who understand these tools well.

The common thread is that AI capability belongs to the teacher's professional toolkit. It is not something to outsource or ignore.

A four-week plan to build the skill

Four weeks, a few minutes a day. By the end, you will not just know a handful of AI tools for teachers. You will know how to direct any of them.

Make it count as professional growth

Learning to prompt well is genuine professional development. It changes how you plan, differentiate, and assess. If you keep a record of your growth, our guide to what counts as CPD and how to plan your year shows how to document new practice like this so it reflects the work you are actually doing.

The time you get back

Return to where this started. The resource teachers lack most is time. Used well, AI gives some of it back: the late-night worksheet becomes a ten-minute task, the differentiated version that never got made now exists, and the student who learns by listening finally has something to listen to.

What you do with that time is the part that matters. More time to know your students. More time to think about the lesson rather than just produce it. More time for the parts of teaching that only a teacher can do.

The tools will change. The prompt is yours to keep.

For the Love of Teaching.