In this article
A serious survey of more than 900 designers across 60 countries puts numbers on what everyone sensed. And the most interesting result is not the adoption rate, it is what still gets in the way.
Figures about AI are rarely trustworthy: most come from vendors with a tool to sell. The report published by Designer Fund and Foundation Capital is an exception: more than 900 designers surveyed across more than 60 countries, around twenty interviews, and case studies run at Anthropic, Framer, Linear, Notion, Shopify, Sierra and Stripe.
The adoption figures
Weekly AI usage among designers moved from 54% in 2025 to 91% in 2026. Three out of four now use it daily. In one year, we left experimentation and entered routine.
Another notable shift: Claude overtakes ChatGPT as designers' preferred general-purpose tool, moving from 52% to 78% while ChatGPT falls from 88% to 65%.
Seven tools on average, and code in production
The number of AI tools per designer has doubled, from three to seven on average. And crucially, half of respondents have already pushed AI-generated code to production. The line between designer and developer keeps fading, exactly as tool announcements bringing design closer to code suggested.
When 91% of a profession uses a tool every week, it is no longer a trend to watch. It is the new working context.
The number one blocker: reliability
Here is the study's most useful result. The main obstacle cited is neither price nor technical difficulty: it is reliability and the uneven quality of results.
That deserves a pause. If the blocker were cost, the fix would be a subscription. If it were skill, the fix would be training. But the blocker is consistency of output, and there no tool will save you: it is a matter of framing. An explicit brand guide, a reference corpus, verifiable rules. In other words, a brand foundation designed for the machine.
The question is no longer whether, but how
These figures drain any interest from the for-or-against AI debate. It is already everywhere, every week, for almost everyone. The only open question is framing: how to get consistent rather than random results. That is exactly the problem I work on with clients, and the study confirms it is where things break.
Frequently asked questions
-
Is this study reliable?
- It is among the most solid on the subject: more than 900 designers in over 60 countries, around twenty interviews and in-company case studies. It is not vendor content selling a tool.
-
Should you use seven AI tools?
- No. The figure describes an average practice, not a recommendation. Two well-framed tools beat seven used at random.
-
Why is reliability the real blocker?
- Because a tool produces variations, not decisions. Without explicit rules and reference examples, two identical prompts give two different results.