A poster that used to take me a day now takes a minute to generate. Last night I built an animated website in one evening without writing a single line of code. By every estimate, designers should have half their lives back.
Yet I haven't met a single designer who has more free time. Deadlines are shorter, revisions are longer, and by the evening there are so many options that your eyes start to blur.
So I tried to figure out what is going on. Are we bad at working with AI? Were we sold a fantasy? Or is this just a bubble that is about to pop? The answer turned out to be more interesting than all three.
It feels faster. It turns out slower

In 2025, researchers at METR ran a simple experiment. Sixteen experienced developers worked on real tasks, some with AI tools and some without. Before they started, they expected AI to make them 24% faster.
With AI, tasks took 19% longer. The strangest part: even after the experiment, they believed AI had made them 20% faster.
The study is small, and it is about developers, not designers. But the same picture shows up in much larger surveys:
| Study | Who was asked | What they found |
|---|---|---|
| 2,500 workers and executives in 4 countries | 96% of executives expect AI to raise productivity. 77% of employees say it added to their workload | |
| 52 interviews, 300+ public AI initiatives | 95% of organizations got zero return from generative AI |
Computers went through the same thing in the 1980s. Economist Robert Solow famously observed that you could see the computer age everywhere except in the productivity statistics. The growth did come, but only after people rebuilt the way they worked.
Where the saved time goes

I see three reasons, and none of them is that the models are bad.
1. Generation got cheap. Choosing didn't. A designer used to make three options and bring the best one. Now you can get 200 in an hour. Each one has to be looked at, compared, and rejected for a reason. Nobody counts that time, but it eats all the savings.
2. Clients know generation takes a minute too. So "by Friday" turned into "by tomorrow", and one round of revisions turned into five. Economists call this the Jevons paradox: when something gets cheaper, we consume more of it. Cheap images didn't reduce the work. They increased the appetite.
3. We dropped AI into the old process. Brief, references, sketches, approvals all stayed the same. There is just a new "generate" button in the middle. According to Upwork, 47% of employees using AI don't know how to reach the productivity gains their managers expect. And in the MIT report, the 5% of companies that succeeded rebuilt the process itself instead of buying one more tool.
There is one more thing METR showed. With AI, the feeling of speed and actual speed drift apart. Generating gives you the sense that work is happening even when it is stuck. For designers, it looks like an hour of rerolling one frame instead of five minutes thinking about what exactly is wrong with it.
So is it a bubble?

Fair question. In June 2026, JPMorgan estimated global AI infrastructure spending at $5.5 trillion through 2030. The largest cloud providers alone are spending about $650 billion in 2026, increasingly with borrowed money. Investors argue about one thing: whether demand will catch up with the spending.
For designers, I think something else matters more. Even if the financial bubble pops, the tools won't disappear. After the dot-com crash in 2000, the internet didn't go away. The companies that couldn't make money on it did, and the ones that could grew in their place.
AI looks similar. If there is a bubble, it is in expectations: "buy a subscription and everything gets faster." The data above shows it doesn't work that way. But the change in the profession has already happened, and it won't roll back.
More designers, less foundation

There is another tension in the industry. Open Threads and it looks like everyone is a designer now. The profession has ballooned and is changing in front of us. But behind a lot of those generations is a lack of basics: composition, grids, typography. AI doesn't fix that. It mass-produces it.
That is especially true for product designers moving from interfaces into visual work. A component library keeps a UI consistent, but there is no Storybook for light, texture and framing.
What I do to make the speed real

- I decide before the first generation, and I set a limit. Light, background, palette and grid are written down in advance. Only the object changes in the prompt. If three rerolls don't get the frame right, I stop and put into words what exactly is wrong.
- I select side by side. Every frame of the series on one screen, then I remove the one that doesn't fit. Comparing them one by one in a chat window takes many times longer.
- I show the client less. Three directions with a rationale, not a folder of 50 images. The more options a client sees, the longer they take to decide and the more revisions they send.
None of these steps requires a new model. They are all about who is running the process.
The bottom line

We weren't fooled. We were sold faster execution and assumed we had bought faster work. AI speeds up making things. It doesn't free up a designer's time on its own. The person who frees up time is the one who decided in advance what they are making and knows how to choose quickly.
That animated website from the first paragraph took one evening for exactly that reason: I knew what feeling I wanted before I typed the first prompt. Without that, I would still be rerolling it.
Has AI saved you time, or eaten it? And how have your deadlines changed in the last year?