How many times have you run into a perfect design that, once moved into code and published in the browser, looks slightly different from what you wanted? I'm talking about fonts.
In Figma, a mockup can be pixel-perfect and accepted by the team with delight. But as soon as the design ends up in a real production environment, the fonts look slightly different.
Sometimes that "slightly" comes at a high price: for example, technical captions set at font size 12 (yes, I mean you, Figtree) can turn into a hard-to-read mess.
You rush to the developers, ready to complain, but it turns out they followed your instructions precisely.
And then you remember a simple truth that you yourself once promised not to forget: fonts in browsers don't render the way they do in Figma, so they end up looking different, and every browser and system has its own way of doing it.
Often, by the time you discover this, the design is already published and nothing can be changed quickly. So you have to go through rounds of approvals and make the edits gradually.
This is where the huge advantage of an AI-based workflow over the "classic" one shows up: a designer now gets access to the in-code version of their interface much earlier and can spot and fix the problem on the fly.
Figma is still good as a tool for initial sketches, but it feels like most of the work will soon be concentrated elsewhere.