AI makes it faster to build. That is not the problem anyone needed solved.
I am writing a book about why good startups fail — and why most of that failure is visible before it becomes irreversible. It is called The Avoidable Startup Failure. If the argument in this article resonates, you can register for early access at anjali.no/book. No newsletter. Just a note when it is ready to order.
Every seminar, investor get-together, and conference I attend right now has the same energy. A founder on stage explaining how they built their product in a week. An investor describing how a solo technical founder can now do what used to require a team of ten. A panel nodding along about the democratisation of building.
The room loves it. I find myself sitting on my hands.
Not because it is wrong. A solo founder can build faster now than at any point in history. The cost of creating a product is collapsing.
But building has almost never been the hard part.
Think about the startups you know that failed. Most did not fail because the technology did not work. Most failed because not enough people saw the value. Because the sales pipeline never filled. Because the cost of selling never matched the willingness to pay. Because the founder spent two years building something excellent that the market received with complete indifference.
That is not a building problem. That is a demand problem.
And demand has not gotten easier. It has gotten significantly harder.
Demand is always a human problem. Even when it is a business problem — a company needing better tools, faster processes, lower costs — it is always a person who feels it and a person who decides to act on it. A specific person, with a specific habit, who has made peace with an imperfect solution because switching is hard and convincing the rest of the team is harder.
People are slow. They are biased. They act on feeling as much as logic. They have already solved most of their problems — imperfectly, expensively, frustratingly, but solved them. The default, for almost everyone, almost all of the time, is to continue as before.
I am not talking about innovators and early adopters — the people who find new things exciting rather than exhausting. Those people exist and they matter. But very few businesses can survive on early adopters alone. At some point you have to cross to the slower majority. The people who do not read tech newsletters. Who adopted their current tool four years ago and have built their whole workflow around it. Who need to see something work for someone they trust before they will consider it for themselves.
That crossing is slow work. It requires patience, repetition, and time. No AI tool touches any of it.
Inertia — the customer’s preference to continue as before — is getting stronger, not weaker.
Because the before is getting better.
Someone who might have needed your tool to solve a workflow problem can now open Claude, Lovable or a similar and prompt their way to a rough version of it. Whether that rough version is actually good enough does not matter yet. The perception that it might be is enough to delay the decision to buy. Maybe indefinitely.
Every acquisition channel is getting more expensive, more contested, and less predictable than it was five years ago. Paid search costs more. Organic reach is narrower. Cold outreach conversion rates are falling.
The founder who launches a beautifully built product today is launching into more noise, not less, against a customer base that is more resistant to switching than ever.
So here is where we are.
The supply side got cheaper. Dramatically cheaper. Faster. More accessible.
The demand side got harder. More expensive to reach. More resistant to switching. More crowded with alternatives, including alternatives people now assemble themselves for free.
And the loudest conversation in the ecosystem is about the cheap part.
When building becomes cheap and fast, the temptation to build before proving demand becomes almost irresistible. Why spend three months talking to customers and running small experiments when you can have a working product in a week? Skip the hard, ambiguous, uncomfortable work of finding out whether anyone actually wants this — and go straight to the thing that feels like progress.
The result is not faster companies. It is faster journeys to the wrong place.
A product nobody wants, built in a week instead of six months, is still a product nobody wants. You just got there faster.
The founders I worry about are not the ones who cannot build. They can all build. Some of them can build extraordinarily well, extraordinarily fast, with a small team and an AI assistant, agents and infrastructure that would have seemed like science fiction ten years ago.
The founders I worry about are the ones who have mistaken the ability to build for evidence that they should. Who launched because launching is now easy, not because the demand was proven. Who are six months in, with a beautiful product and an empty pipeline, wondering why the growth is not coming.
The answer is not a better product. It was never a better product.
It was proof, gathered before building started, that someone wanted this badly enough to change. A specific someone. A slow, biased, habit-bound human being who had to see enough reason to stop doing what they were already doing. Getting that person — and then the next one, and the one after that — to move is work that no model accelerates.
It requires sitting across from a potential customer and understanding, precisely, what it would take to make them act. That work is slow. It is ambiguous. It produces nothing you can ship.
It is also the only work that determines whether everything else matters.
The ecosystem does not need better builders. It has plenty.
It needs founders willing to do the demand work first — before the building starts, before the runway burns, before the window closes.
That is the part nobody is celebrating. That is the part that needs to change.
I love this take, and it's one that’s very much needed. AI’s impact on democratising execution velocity raises many important questions. Not only that, it is also forcing a rethink of what really matters in building durable businesses that solve real customer problems. Building fast and at scale is easier now; the more productive question is whether anyone wants what you are building. I think the way to approach it is to build fast enough to test and iterate, gather real user feedback, drop or pivot what doesn’t work, and scale what does.
direction > speed. great wisdom in here Anjali, thanks for sharing :)