A finished website and a working website are two different products
- Maciej Konarzewski
- 1 day ago
- 5 min read
There's a new kind of enquiry landing in our inbox this year
It goes roughly like this: "We built the site ourselves with one of the AI builders. It looks great, honestly. It's been live three months. Nothing's happened."
Three months ago I'd get one of these occasionally. Now it's most weeks. And I want to be careful here, because the obvious agency response is smug, self-serving, and misses what's actually going on. You know the one. "Well, you should've hired professionals."
Because the sites aren't bad. That's the thing. I open them expecting a mess and I get clean layouts, sensible sections, copy with no typos. The builder did its job. Everything a checklist would catch is fine.
And the site still doesn't work.
What "finished" actually means
When an AI builder hands you a website, what it's really handed you is a hundred executed decisions. Layout decisions, colour decisions, copy decisions, structure decisions. All made instantly, all made competently.
None of them made about your business.
The builder executed a hundred decisions. It just didn't make any of them about your business.
It doesn't know that most of your customers find you after a recommendation and land on the site purely to check you're legitimate. It doesn't know your best margin is in one service and the thing you want to quietly retire is another. It doesn't know the question every single customer asks on the first phone call, the one your homepage should answer before they have to ask it.
So it fills the gaps with the average. Average structure, average messaging, average everything. Which photographs beautifully and converts nobody.
A site that looks finished is not the same as a site that works. Looking finished is a design problem, and design problems are basically solved now. Working is a strategy problem, and strategy problems can't be generated. The raw material isn't out there on the internet. It's in your business, and someone has to go and dig it out.
The questions the builder never asked
When someone brings us one of these three-months-and-nothing sites, we don't start by redesigning anything. We start with the questions the builder skipped.
Who exactly is this for? What do they type into Google before they find you? The actual words, not the industry's words. When they land, what's the one thing we want them to do? What has to be true on this page for them to trust you enough to do it? And how will we know, in eight weeks, whether any of this is working?
That's exactly how the work started with a rendering specialist we took on in Bournemouth. They told us straight: they didn't want more traffic, they wanted better jobs from serious homeowners. Every decision on that project flowed from that one answer. No builder would ever have asked the question.
Boring questions, all of them. Every one answerable only by a human who understands the business. And every one worth more than the entire visual layer sitting on top.
How we use the same tools
Here's what I suspect surprises people most: we use AI at almost every production step, and it's made us properly fast.
First-pass layouts and section scaffolds in Wix Studio, quick copy variants to react to, resizing and adapting assets across pages and breakpoints. Work that used to eat days now takes hours. I have no nostalgia for the slow version of any of it.
Our rebuild for Torqueflow-Sydex is a good example of the split. A pump manufacturer with 42 pages and no way to find a pump. AI did plenty of the production lifting on that project. But the decision that mattered, restructuring the catalogue around two cross-referenced databases so engineers could find a pump the way engineers actually search, came from sitting with the problem. No prompt produces that.
The sequence never changes. The sitemap comes from keyword research and the discovery conversation, not from a prompt. The homepage message is decided before a single section is generated, because if it's decided after, the layout starts making your decisions for you. Then, and only then, the fast tools run. AI gets the hands. The thinking keeps the head.
Nobody's phone rings because their website was quick to make.
Speed of production is now table stakes. Your competitors have the same builders you do. What they don't have is your answers to the boring questions. Unless you never answered them either, in which case you're both publishing the same average site with different logos.
Faster was never the problem
I've been building websites for over ten years, and in all that time not one client has ever said their real problem was that websites take too long to make.
The real problems were always the same: not enough of the right people arriving, and not enough of them getting in touch. AI builders don't touch either. They compress the part between the thinking and the launch. That's useful, and I'm glad it exists. But compressing the middle doesn't help if the front end never happened.
Twenty minutes gets you a website. It doesn't get you a reason anyone should visit it.
So no, I don't think AI builders are coming for the agencies that do actual strategy. I think they're coming for the agencies that were only ever selling production. Honestly, that reckoning is overdue.
If you've got one of those looks-great-does-nothing sites and you're wondering which camp yours falls into, drop us a line and we'll take a look and tell you straight.
Are AI website builders good enough for a small business?
For getting something live quickly and cheaply, yes. They're genuinely capable. The gap isn't quality of execution, it's strategy: who the site targets, what it says, and how it converts. If you've done that thinking yourself, a builder can execute it. If you haven't, the builder can't do it for you.
Why does my website look good but generate no leads?
Usually because it was built from design defaults rather than from your customers. Common culprits: the most-searched service isn't prominent, the homepage doesn't answer the question every customer asks first, and there's no clear single action for a visitor to take.
What should I decide before building a website with AI?
Who it's for, what they search for in their own words, the one action you want a visitor to take, and how you'll measure whether it's happening. Those four answers turn a builder from a random generator into a fast production tool.
Do agencies like Vision Marketing use AI tools themselves?
Yes, daily. Layouts, first-draft copy, and asset production all get the AI treatment. The difference is sequence: the strategy is decided by humans first, and the AI executes it. Not the other way round.
.png)
Comments