We turned 700 students' phones into one live classroom, for AI and Excel

First published: November 2026. This post records an idea and a working product that we designed, built and took into real classrooms this month. We are sharing what it does and why. We are not sharing how it is built.

Two courses, one problem

We teach two short, practical courses to big groups, up to 700 students a day:

• AI Prompt Engineering Essentials: how AI assistants work, how to write prompts that get good answers, how to check what comes back, and how to use AI responsibly.
• Excel: From Beginner to Advanced: from your first cell to formulas, PivotTables, lookups and a dashboard.

Most of our students have a phone; few have a laptop in class. Three things kept going wrong in both courses:

1. The back of the room can't see the slides. A projector works for 40 people, not 400.
2. Slide decks get shared and copied. We put real work into the material, and a PDF on WhatsApp is gone forever.
3. "Interactive" usually means hands up. With 37 teams, you can't hear every answer, let alone mark it fairly.

The usual fix is five tools: a slide link, a quiz app, a voting app, an attendance sheet and a spreadsheet. Students switch apps, the teacher juggles tabs, and the lesson loses its flow. We wanted one thing, for both courses.

The live classroom

It runs in the browser, with no app to install, and is built around one rule:

The teacher drives. Every screen follows.

Students open one address on their phone, type their name, the class code and their team code, and they're in. From then on:

• Every phone shows the slide the teacher is showing, the moment it changes. Animations, reveals, timers and scores included.
• The teacher runs the whole class from a phone: next, back, reveal, timer, team picker, points, and a video that plays on every screen at once and loops until the teacher stops it.
• Students can't skip ahead, download the deck or save the slides. What hasn't been revealed yet never reaches their phone.
• Every phone is watermarked with the student's name, which makes a screenshot accountable.

Each course gets its own classroom site, with separate teams, scores and records, and its own look: navy and gold for AI, Excel green for Excel.

37 teams, one league

Each team gets a printed card with its own team code, separate from the class code, which changes every class. The first teammate to join names the team.

When a game is on screen, every phone in that team gets the answer controls. Any teammate can answer, the whole team sees the current choice, and the last answer before the reveal counts. The teacher watches "23 of 30 teams answered" and reveals when the room is ready.

At the reveal, every team is marked at once, fairly and once per question. Each phone shows "Correct, +1" or the right answer, and the league updates on the projector and on every phone ("You're 4th of 37"). Scores carry across the whole course.

The same game types work in both courses: multiple-choice rounds, card sorts that teammates do together, live class votes where the bars fill as students tap, and a random team picker that lights up the chosen team's phones.

The AI course: learn by playing the model

Teaching AI to beginners means fighting two myths: "it looks things up" and "it's always right". So the lessons make students experience how it works.

• Be the model. Teams build a sentence one word at a time, choosing the most likely next word, then see the odds. It's the clearest way we've found to show that an AI predicts text rather than checking facts.
• Strengths and weak spots. Teams sort tasks into "good at" and "weak at" on their phones: drafting and summarising versus exact citations and recent events.
• Name the failure. Real-looking cases of hallucination, bias and out-of-date answers, answered by every team at once.
• Compare the assistants. Students run the same prompt in several assistants and vote live on the winner, then defend their choice.
• Build better prompts. Lessons on what makes a good prompt, with rewrites, error hunts and a final showcase.

The Excel course: a spreadsheet that answers back

Teaching Excel from slides has always been odd: you show a formula and students watch.

So we put a working spreadsheet inside the slides. It looks and behaves like Excel, with a Name Box, a formula bar, colour-coded references, fill-down, sheet tabs, filters and conditional formatting, and it actually calculates. It runs on a real practice dataset of 120 student records.

Then we sent it to the phones. On a live-sheet task, each team types the formula on their phone:

Task 1 of 2 · 1 point Total in H2: add the Assignment (F2) and the Exam (G2). It fills down to H11, like the fill handle. =F2+G2 [ Send ] In H2 this gives 82 

They see what their formula gives as they type, then send it. At the reveal, every team's formula is filled down and checked cell by cell, the way Excel itself would calculate it:

• =F2+G2 ✅ and =SUM(F2:G2) ✅: any formula that gives the right answers scores
• =$F$2+$G$2 ❌ "1 of 10 cells right": the classic $ mistake, caught and explained
• 82 ❌ "That was a typed answer, not a formula": no shortcuts

It covers what a four-session course teaches: SUM, AVERAGE, IF/IFS, COUNTIF(S), SUMIF(S), AVERAGEIF(S), XLOOKUP, VLOOKUP, INDEX/MATCH, TRIM/PROPER, SUMPRODUCT, ROUND, SUBTOTAL, and the newer FILTER, SORT, UNIQUE and TAKE (yes, the answers spill). We checked it against an instructor answer key of 503 formulas, plus every number quoted on the slides. It matches them all.

Around the live sheet:

• charts drawn from the real data, plus a "fix this misleading chart" before-and-after
• a PivotTable you build on screen: Rows, Columns, Values and a filter
• a dashboard with slicers: tap "Computer Science" and every KPI and chart updates
• a formatting makeover that turns an ugly sheet professional one step at a time
• a worksheet-style design, with column letters, row numbers and formula-bar labels, so the whole course feels like Excel

What the teacher gets afterwards

With no extra admin, for either course:

• daily logins: who joined today and who didn't
• attendance across the course
• per-class reports: time connected, time actually on the page, reconnects, and attempts to copy or save the content
• CSV exports of everything

We record only what teaching needs, and students are told on the join screen what is recorded.

What we learned

• The phone is the classroom. Once every phone shows the same slide, the room's size stops mattering.
• Teams beat individuals. With 37 teams, every question becomes a conversation in every corner of the room.
• Instant, fair marking changes behaviour. Students argue about $ signs, and about whether an AI answer can be trusted, because their team's score depends on it.
• The teacher must stay in control. Phones never run ahead, and the teacher's phone is the remote.

What's next

We're running both courses with our own cohorts now, and refining them with each class. We'd like to hear from:

• schools, universities and training providers who teach large groups
• instructors who want their own courses delivered this way
• anyone who has tried to teach AI or Excel to a room full of phones

Leave a comment or get in touch. We'd love to show you a class running live.

The idea, the teaching design and the product described here are our original work, first shown publicly in this post. All names and records in the screenshots are fictional practice data.