Every dominant interface eventually becomes a museum exhibit. The command line gave way to the GUI, the GUI gave way to touch, and touch is now giving way to voice and intent. Web builders are next. Wix, Squarespace, Webflow and Elementor solved a real problem fifteen years ago: they let people who couldn't write HTML put a presence online. They were the right answer for their decade. They are not the right answer for this one.
What drag-and-drop actually solved
Before 2010, building a website meant hiring a freelancer for two thousand euros or wrestling with FTP and a copy of Dreamweaver. Drag-and-drop democratised the web. A florist in Brno could sell bouquets without learning CSS. A photographer in Lyon could publish a portfolio without phoning a developer. That was a genuine revolution, and it deserves credit. The problem is that the revolution stopped. The interface that liberated millions of small businesses has become the new bottleneck.
Why describing is faster than dragging
Watch a small business owner build a site in Wix Studio. They open a template, scroll through forty section variants, drag a hero block onto the canvas, fight with the responsive breakpoints, realise the headline is too long, resize a button, and forty minutes later they have one page that mostly works. Now picture the same person sitting at a screen and saying, I run a sourdough bakery in Vinohrady, I want a warm site that shows my three signature loaves and lets people preorder for Saturday. Thirty-five minutes later the site is live. One of these is a craft. The other is a conversation. Conversations win because intent is denser than gesture.
The agency timeline is the real comparison
Most people compare AI builders to other DIY tools, which is the wrong frame. The honest comparison is to a small agency. A typical brochure site quoted today in Prague, Berlin or Vienna runs four to twelve weeks and twenty-five to ninety thousand crowns. Discovery call, mood board, two design rounds, development handoff, content delays, revisions. The output is often perfectly fine. The process is what's broken. When a baker can describe her intent on Sunday evening and have a working site by Sunday night, the four-week timeline stops being a feature of professionalism. It starts looking like overhead.
What this means for designers and agencies
Designers who think AI will erase them are reading the situation backwards. The drag-and-drop era turned designers into pixel arrangers and component janitors. Conversational creation hands that work back to the machine and frees designers to do what they were trained for: judgement, taste, brand strategy, art direction. The agencies that thrive in 2027 will not be the ones with the largest Webflow teams. They will be the ones who run AI builders white-labelled under their own brand, ship in days instead of months, and charge for taste rather than for hours spent nudging divs.
What 2027 and 2028 actually look like
By the end of 2027, expect three things. Most freelancers will stop selling brochure sites as a deliverable, because clients will refuse to pay four-week timelines for something a competitor delivers in an afternoon. Squarespace and Wix will quietly bolt voice and chat layers on top of their canvases, but the underlying paradigm will still be block-shuffling, which means they will feel like CD players with a Bluetooth dongle. And a new layer of tools, including Codrik, will compete on conversation quality and taste rather than on template count. By 2028, asking a designer to drag a hero section into place will sound the way asking an accountant to use an abacus sounds today.
The shift is structural, not cosmetic
It is tempting to file this under another design trend, somewhere between brutalism and bento grids. It isn't a trend. It is the same kind of jump that took us from typing ls to clicking a folder icon, and from clicking a folder to swiping a screen. Each of those shifts felt premature for about two years and obvious for the twenty after. Drag-and-drop builders had a fifteen-year run, which is a respectable lifetime for any interface. The next fifteen years belong to people who can describe what they want and to the systems that listen carefully enough to build it.