A survey of 443 designers from 43 countries showed that artificial intelligence is actively used at the idea-generation stage but is sharply pushed aside at the final stage. In the Discover phase, 72% of respondents; in Deliver, only 41%. The paradox is obvious: the closer to the real product, the less trust in the algorithm.
The reason is simple and economical. At the early stages, a mistake costs nothing — a failed idea is easy to discard or rework. The further along the Design Council's "double diamond," the more expensive a misstep becomes: it may already reach users. That is why designers readily entrust AI with generating options but reserve the right of final judgment for themselves.
Interestingly, even specialized tools lose out to ordinary ChatGPT. In the Define phase, it accounted for 52% of all mentions. Designers prefer a universal chat over narrowly specialized "gadgets," which researchers most often create specifically for the later stages.
What holds skeptics back is not convenience but usefulness: 40% of those who do not use AI consider it insufficiently effective. At the same time, 80% of respondents already use it in at least one phase. The gap between enthusiasts and skeptics persists, but the data were collected in 2025–2026, an era when today's models already seem outdated.
A similar picture is confirmed by Figma's experiment: AI cuts product managers' working time by 35%, but designers' by only 17%. The machine helps those with less experience the most, but it does not replace expert intuition in complex decisions.
Thus, for now AI remains a "starting" tool. It multiplies the number of options many times over but does not take responsibility for the final result. In the coming years, everything will change only when algorithms learn not just to generate but to reliably bring design to completion.
