"It's got dark mode!" Sweet, but has it got a customer?
Someone showed me their product recently. It had a lovely little dashboard, a nice little onboarding flow, and a dark mode switch. They'd built it in nine days with AI, and before this, they'd never written a line of code.
"Who's the paying customer?" I asked.
"Ah, well, we've got about five different segments we could go into."
"But who are you going after? Who's currently paying to solve the problem you're fixing with this?"
“It can solve many problems for different sectors, I haven’t spoken with anyone. But years ago I used to sell to this sector so I know they’ll love it.”
Uh-oh.
This little exchange demonstrates something called phantom progress - activity that looks and feels like progress, but isn't really getting you any closer to demonstrating value. (Which is the bit we need, it’s the bit that pays the mortgage.)
It's also the most dangerous type of progress, 'cos from the outside it looks like you're making loads of it but if someone takes a closer look they’ll see you’re lacking a few of the basics.
Don't you just love it when the dashboard's ready before the customers are?
AI made the facade free (booooo)
For years, building software was slow and expensive. And while that's time and money you don’t always have (trust me, I know), it has a hidden benefit: it forces you to think about the hard stuff, the stuff you really need to get your point across.
If your MVP is going to cost you £45k and it’ll take six months to build, you'd sure better make sure people want it.
AI has removed that friction. You can now go from idea to working prototype over a weekend.
I saw my sister do it recently with her maths game .She had an idea, plugged it into Loveable and voila - an AI designed Wizard was making me feel bad about not remembering trigonometry.
In some cases using AI in this way is a big positive, especially for those founders that don’t have the same resources as everyone else. All of a sudden, people who couldn't afford a software build now have something in their hands to show. Yay!
But there's a chance it's also fuelled a resurgence of the "build it and they will come" mentality.
Boooo. (Not sure why you’re getting a lot of panto noises in this article.)
Phantom progress: has AI brought back the "build it and they will come" mentality?
"I've validated it" means different things to different people. When you hear someone has validated their idea, I don’t think it hurts to question what that means.
Think back to the founder I mentioned at the start.
When we dug into it, his version of "validated" was identifying the demographics: "works in property investment, mid-40s, UK" and worked out his TAM.
But could he tell me what they were using to solve the problem?
Could he tell me what was wrong with the current offer?
Could he tell me which channels reach them?
Nope. Was he interested in digging into this? Unfortunately not, off he went, to develop another feature.
Startup validation comes down to three questions:
Who's the audience and what are they switching from?
Where’s the gap in what they are currently using for you to fall into?
What's their level of urgency and agency?
If you can't describe your ideal customer in terms of the situation they're in and the problem they want solved, you're not there yet. If people are already paying, you're not creating a market. You're offering a better way to spend money they've already decided to spend.
Once you've collected this info, you suddenly realise you only need to build one feature - the feature that delivers the promise.
Not a dashboard. One feature that delivers on one clear promise to that niche. Let that determine what’s in your AI tool.
'Cos everything else is decoration.
I'm not an AI laggard/hater - pinky promise.
This isn't an argument against AI.
And as I write this I can feel the AI vibe coders clenching their jaws… but what I am trying to suggest is the shortcut isn’t always the best route. Greg Engle from Re-Wired suggests that with AI we’re offloading our thinking, and I think that sums up nicely what I am saying here.
This little update is to get you thinking about the order in which you use AI when building something you want people to buy. Use AI to build fast once you know what you're building, and for whom. Used this way, in this order, it'll be the best thing to happen to you.
If this article resonates, before charging up Claude this week, try this simple test:
Write down the niche you're serving, the urgent problem they have and what they're paying to solve it.
Go and find five people who can confirm all three.
If you can't do that, I'd suggest you may need to do a little more groundwork, and no amount of AI-built front ends will fix it.
But fear not! These articles may help you get there sooner:
Going to market? Pick one competitor and forget the rest
I'd rather run into traffic than make a sales call
Millennial Alex doesn’t want to buy your product.
In the meantime, if you want a nice little chat about anything I’ve raised - or, you support teams who are currently working this stuff out - I’m always happy to chat.
