"I typed it and it appeared": building a custom classroom widget with AI in 60 seconds

Every teacher has one: the weirdly specific tool you’ve always wanted that no app store has ever stocked. Mine was a phonics sound-button board — a grid of this week’s graphemes where tapping one says the sound aloud, with a quiz mode that plays a sound and asks the class to point to the grapheme. Too niche for anyone to build commercially. Exactly the kind of thing I’d have paid for on a Sunday night in September.

Last month I typed a description of it into Boardee’s AI App Builder — one sentence, plain English, no code — and about a minute later it was sitting on my class board. Working. Tappable. Making the sounds.

I want to walk through exactly what happened, because “AI builds you an app” is the kind of claim I’d have rolled my eyes at, and the reality is both more mundane and more useful than the pitch sounds.

What I actually typed

Here’s my first prompt, verbatim:

“A grid of buttons showing this week’s phonics graphemes: ai, ee, igh, oa, oo. Tap a button and it plays the sound. Add a quiz mode that plays a random sound and the class has to say which grapheme it was, then reveals the answer.”

That’s it. No mockup, no settings, no tutorial. The builder thought for the better part of a minute — long enough to sip coffee, not long enough to check email — and produced a widget: five chunky buttons, tap-to-hear, and a quiz mode with a reveal button. It dropped straight onto my board next to my timer like it had always lived there.

Was it perfect? No — and this is the part that matters. The buttons were too small for a projector, and the quiz revealed the answer instantly instead of waiting for the class to commit. So I typed two follow-ups: “make the buttons much bigger, readable from the back of a classroom” and “in quiz mode, wait for me to tap before revealing.” Each tweak took another short wait. Three prompts, maybe four minutes end to end, and I had the tool I’d been idly wishing for since roughly 2019.

The widget saves with the board. Monday morning it was just there, between the name wheel and the visual timetable, and my Year 1s used it like it was any other part of the room.

What my colleagues built in the staffroom

Word got around, as these things do, and the staffroom test drive produced a nice cross-section of what teachers actually want:

• A “working walls” randomiser (Year 5 teacher): tap for a random past topic, class does one-minute retrieval. She’d been doing this with lolly sticks in a mug since forever; now the mug is a widget with a shuffle animation.
• A traffic-light exit ticket counter (science): three big buttons — got it, nearly, lost — tapped by students filing out, tally on screen. Is it rigorous assessment? No. Did he have data by 3:31 that he’d never had before? Yes.
• A “hands in” group readiness board (my deputy): each table taps in when ready, board goes green when all six tables have. Her prompt was literally “like a restaurant order screen but for tables of children being ready,” which the builder somehow understood.
• A French verb conjugation flipper (MFL): shows a pronoun + infinitive, tap to reveal the conjugation. She iterated four times to get the exact verb list and tenses — the record for pickiest prompt sequence, and the tool kept up.

Notice the pattern: none of these are apps anyone would ship. All of them are the small, specific, this-classroom-shaped tools that teachers have been approximating with card, Blu Tack and mugs for a century. That’s the actual product here — not “AI writes software,” but the long tail of tiny classroom tools finally being buildable by the person who needs them.

Honest advice for your first build

Describe the behaviour, not the technology. The best prompts read like you’re explaining the tool to a teaching assistant: what’s on screen, what happens when someone taps, what the class sees. My failed first attempts were vague (“a phonics game” produced something generic); specificity in, specificity out.

Expect two or three rounds. First drafts are 80% right. Budget for a “bigger text,” a “change these labels,” a “wait for my tap.” The iteration is the workflow — treat draft one as a sketch to react to, not a verdict on the tool.

Start small and single-purpose. One job per widget. My colleague who asked for a combined behaviour-tracker-slash-timetable-slash-quiz got a muddle; the widgets that shine do one classroom-sized thing. If you want two tools, build two widgets — they can sit side by side on the board anyway.

Check it like you’d check anything before projecting it. The builder is generating your content as well as your buttons, so proof-read labels and (for my phonics board) verify the sounds against your scheme. Two minutes of checking, same as you’d give a worksheet from any source.

Know when an existing tool already has you covered. Before building a voting widget, look at the poll; before building a picker, the name wheel exists and is better than what you’ll prompt in a minute. The builder is for the gaps between the built-in tools, and the gaps are where its minute is best spent.

The cost, plainly

App building draws on the AI allowance: the free tier gives you a batch of credits to start — genuinely enough to build and refine a widget or two and see whether this is your kind of tool — and a 14-day Pro trial takes the limits off if you want to keep going. Pro (from $9.99/month) is there if, like me, you get the bug and start generating things weekly. No login is needed to try the board itself; the widget you build saves right onto it.

The phrase in this post’s title came from a colleague watching over my shoulder at 8:15 on a Tuesday, and it’s still the most accurate review I’ve heard: I typed it and it appeared. Fifteen years into teaching, I’ve stopped expecting software to fit my classroom. Turns out the fix was software that lets the classroom describe itself.