It's Tuesday morning in the City of London. A room full of marketers and business leaders, here to learn how AI is changing everything.
And the conference's own website has gone down.
The host breaks the news in the very first session. The site people need for the agenda and the workshops isn't working. It gets mentioned again through the day, and by the second or third session the organisers have replacement agendas up that you reach by scanning a QR code. Apologies all round. On with the show.
Late morning, it's time for a hands-on session from Replit. The slide on the big screen says "Laptops out". Replit's tools let you build an app by describing what you want in plain English. People call it vibe coding. You say what you want, the AI writes the code, you try it and you tell it what to change.
Some of the demo doesn't behave. A few bits on the big screen won't do what they're meant to, and the speaker works round them.
Replit itself is working, though. So I open my laptop and start describing an idea for an app that can research prospects. No code, just a description in plain English.
By the end of the session I've got version one of a new app.. Built in the room, in one session, by describing it in plain English.
That afternoon, during another vendor's demo, the speaker waits for his results to come through, "if the Google Cloud will let me," he says.
So that's aiblLIVE in London. A day about the future of AI, with the website down from the very first session and the odd demo wobbling. And I walk out having built version one of a new app.
Why the wobbles didn't matter
I don't say any of this to poke fun at the organisers. They handled the website calmly and quickly, which is exactly what you'd want. And the day was excellent. It gave me so much in terms of ideas.
Every talk showed something impressive. AI writing briefs, analysing ads, running customer service, building its own filing system. Then reality turned up uninvited. A website fell over. A demo stalled.
If I'd taken that as my cue to sit back and wait for everything to work, I'd have gone home with a notebook full of ideas. I went home with a prototype instead.
The tools aren't perfect. They don't have to be. A rough app you can try beats a perfect one that's still an idea in your head.
Done beats perfect, live in the room
I wrote a book called Done Beats Perfect, and this was it, playing out live.
The organisers didn't cancel. They put up a workaround with QR codes and carried on. The speakers worked round the bits that wouldn't play. And I got on with building.
Most people wait until everything is ready before they start. The website has to be finished. The plan has to be perfect. The tech has to be sorted. Someone has to show them properly first. And while they're waiting, someone with a laptop and a rough idea gets on with it.
A prototype is allowed to be rough. That's the whole point of it. You can show it to people, find out what's wrong with it and make it better. You can't do any of that with an idea you're still polishing.
What vibe coding means for you
If you run a small business and you've been waiting for AI to settle down before you try building something, I've got news. It won't. There'll always be an outage somewhere or a demo that stalls.
So what would happen if you started anyway?
You don't need to be a developer. You need a problem you understand better than anyone else, and the nerve to describe it in plain English. Think about the job in your business you do the same way every week. The spreadsheet you keep copying. The homework you do on someone before a sales call. That could be your first app.
And when the tech wobbles, as it will, have a plan B for the important stuff. Know how customers would reach you if your website went down for a morning. AI makes you faster on a good day. Your own preparation keeps you going on a bad one.
A small challenge
Pick one small, boring job in your business that you'd love a tool for.
Spend ten minutes writing it down in plain English, as if you were explaining it to someone on their first day. What goes in, and what you want to come out the other end.
Then paste that into Replit, or a tool like it, and see what it builds. Your first go will be rough. That's fine. Done beats perfect.
