One minute, one tablet, one draft
Yesterday at a client's office. On the table sits a quote we wrote months ago. Now it's back, with a new assumption attached: with AI, this has to be cheaper. Then the client picks up his tablet, opens ChatGPT and speaks into it: design me a website. A minute later he turns the tablet toward us. "That was quick."
It was. It wasn't better.
This is the most expensive mistake companies are making with AI right now: they confuse fast design with good design. The line comes up in almost every first conversation: "AI does that in five minutes. Why does it take you two weeks?"
What actually happened in that minute
The draft on the tablet looked tidy. That's the problem, because tidy is the average. AI works with the mean of everything that already exists. What comes out is the website of the industry, not the website of the business.
We showed the client three things:
- The draft has no stance, just a random seed. Ask the same question next year and ChatGPT returns something else. What fits today won't fit in January.
- Nobody asked the most important question. Who reads this page, and what should they do next? A tool only answers what you ask. Ask the wrong question and you get the wrong answer in record time.
- His hour is worth more in his trade than in front of a chat window. Spend an afternoon explaining to ChatGPT what you mean, and you've spent an afternoon not working.
On top of that comes an obligation that keeps growing: since August 2026, Article 50 of the EU AI Act imposes transparency duties for certain AI-generated content. Generate blindly and you also have to work out what needs labelling.
Two paths to the same launch. Drag both handles to the flag and watch the scale below.
What we learned in that meeting: technology convinces nobody
A confession. We like to argue this topic with technology: clean tech stack, a design system with clear hierarchy, behaviour logic for every device. All true. And all invisible to someone who has just seen a finished layout on a tablet. The client sees speed. He doesn't see the difference between a layout and a brand, because that difference only shows after years.
What worked in the end was something else: we've been working together for years, and he trusts us. We found a compromise. The client sets the basic design concept himself, we implement it. That saves him a four-figure sum, including on usage rights, because less design work comes from us. We also said clearly: we won't vouch for the timelessness of this design. Whether the site still fits the brand in five years is now his call. He no longer has our guarantee for it. Both sides were fine with that.
With a new client we would have reacted differently. With a long-standing client who says openly that he needs to save, we trade scope for clear expectations. The relationship is worth that to us.
Three questions before any AI starts in our studio
We use AI every day, as a shortcut through routine groundwork, not as a shortcut to the result. Three questions keep us on course.
- Which decision should the design enable? Before a tool starts, we settle what the visitor should do afterwards: call, enquire or buy. The question first, then the tool.
- Where does the saved time go? We use the speed to show something tangible early: a clickable demo instead of endless moodboard rounds. The time AI saves us in building goes into the decisions that matter. Speed shifts work; it doesn't replace it.
- Could this sit on your competitor's site? Our simplest test. If a draft would work just as well on the competition's website, it isn't finished. No matter how fast it came together.
Add to that our open workshop: you watch live in the design tool or in the code. You see where AI gives us an edge in the project, where real human work is needed, and, if you're open to it, you pick up media literacy and an understanding of how things work along the way.
Why clients sense the difference in the end, even when they can't name it, is the subject of part two next week: The Chalkboard.
TL;DR
Fast is a tool. Good is a decision.
AI makes design faster. It only gets better when someone puts the time saved into thinking instead of into more variants.
Go ahead, turn the tablet toward us
You already have AI drafts for a logo or website and don't know whether they're any good? Show us. I look at the concept, Daniel at the technology. You get honest feedback in three sentences, with no project manager in between.