Worth
building.

I love building. You have an idea at 10pm and by 2am it works. It’s the joy we had as kids, showing our parents a necklace made of noodles: “Look what I made. One of my ideas made it into the real world.”

My mum always said it was beautiful. Then I met clients. They ask why I made it in the first place.

I still make noodle necklaces every week, because that’s how I learn. My early mistake was handing them to people. The first things I built with AI were finished in my head before anyone was ready to use them. I’d built them for the joy of building, not for the people meant to use them. There’s a word for that now, slop, and I see it everywhere.

01The bottleneck moved.

Can we build it?

Should we?

For decades the hard question was “can we build it?” Building took teams, budgets and skills. Strategy existed to ration that capacity, and “it’s not possible” made plenty of choices for us.

AI has removed most of that cost, and most of that excuse. A working prototype takes days. A small custom tool that never paid for itself now does in a week. Production still takes real engineering, but the first version, the one that makes you fall in love with an idea, is almost free. Soon anyone will be able to build something. Few will have the discipline to ask whether they should.

That discipline is the effort of choosing what’s worth doing, for whom, and whether anyone will still use it in a month. Nothing says no anymore, not the budget and not the tech, so every no has to come from us.

Building got cheap, so the expensive mistake is now building the wrong thing beautifully, at speed, for people who never asked.

02Slop is what happens when nobody decides.

Everyone now agrees the hard part is knowing what to build. Almost nobody is doing it. We’re producing more than ever and less of it works: tools built for users we think we understand, AI agents nobody keeps, decks generated in an afternoon and never read properly. As with the noodle necklace, making it becomes the point.

Carelessness and spray-and-pray are the easy explanations. The bigger one is that AI is overwhelming everyone, and handing over our critical thinking has never been easier. Things move so fast that most teams are just keeping up, and their projects end up gathering dust in a repo. Gartner expects over 40% of agentic AI projects to be cancelled by the end of 2027. The faster AI moves, the more keeping up becomes a way of falling behind with better excuses.

When strategists, designers and developers feel overwhelmed, they delegate decisions, taste, architecture and code. All that delegation pulls everything toward the average. AI raises the floor, then turns it into the ceiling: good becomes whatever everyone else would have made.

So slop is a discipline problem more than an AI problem. It’s what happens when making outruns the reason for making.

03How I decide what’s worth building.

Discipline that depends on willpower fails the moment people are overwhelmed, so it has to be built into the way the work gets done. I build it in with four habits. Each one forces a decision before anything gets delegated.

1. Hack first.

   The best signal of what’s worth building is someone already hacking it. People who build for themselves never make noodle necklaces, because they’re both the maker and the user. Every company has a handful of them, running workarounds nobody approved. Shadow AI is unpaid product research: every workaround is a brief nobody wrote and nobody is reading. I find those people, fork what they’ve made, fit it to everyone’s routine, and automate everything but the judgment.

2. Money, magic, machines.

   Magic: it has to feel different to the person using it, not just work better. Money: it has to pay for the business, otherwise it’s an idea and not yet an innovation. Machines: map where the technology is going, even if no one can say when, and build for that, or the product becomes technical debt.

3. Fail fast.

   Put prototypes in people’s hands early, to learn before anything launches. If it breaks, good: they’ll tell us why while it’s still cheap to fix.

4. Lead with a vision.

   Most teams adopting AI have their heads down, using it to speed up today’s work instead of reimagining it. I get them to look up and ask where this is going and what their work becomes when it gets there. Saying no is only half the job; the other half is showing people what’s worth wanting.

04The short version.

Make all the noodle necklaces you want; the discipline is never handing one to someone who can’t feel why it matters to them.

Strategy without building is opinion, and building without strategy is slop, so let’s do both. The people who matter next will be the ones who can tell what deserves to exist, for whom and why, and then build it to prove it.

If you’re building something and you’re not sure it’s worth it, let’s find out. Fast.

Robin