Sources checked 11 October 2026.
Two diamonds, side by side. About twenty years ago the UK Design Council drew the way designers work as that shape and named its four phases Discover, Define, Develop and Deliver. [1] Each diamond opens out and then narrows. In the first you widen your view of the problem, then settle on one. In the second you widen the range of answers, then settle on one.
The council's description of the first diamond is the line worth keeping: it "helps people understand, rather than simply assume, what the problem is". [1]
Generative AI has made the second diamond close to free. It has done very little for the first. That imbalance is, I think, the most useful way to judge what design thinking is worth now, and why the people who are good at the first diamond are about to matter more.
The idea, without the sticky notes
The idea is older than the movement. Herbert Simon wrote that "everyone designs who devises courses of action aimed at changing existing situations into preferred ones". [2] In 1973 Horst Rittel and Melvin Webber described the problems planners face as "wicked": no definitive formulation, no stopping rule, solutions that are better or worse rather than true or false. Their sharpest line is about the first diamond: "The formulation of a wicked problem is the problem!" [3]
The movement came later. Stanford opened its d.school in 2005 [4], and IDEO turned the approach into workshops, courses and posters. IDEO still describes design thinking as "a human-centered approach to solving complex problems". [5]
Then came the backlash, and much of it was earned. In 2017 Natasha Jen, a partner at Pentagram, gave a talk called "Design Thinking Is Bullsh*t", aimed at the Post-it notes, the hexagons and the oversimplification. [6] The historian Lee Vinsel called it "a 'movement' that's little more than floating balloons of jargon". [7] MIT Technology Review's 2023 retrospective put the weak point plainly: "Execution has always been the sticky wicket for design thinking." [8] In 2023 IDEO cut 32% of its staff. [9]
The evidence for its value is thinner than the slogans. McKinsey's 2018 study of 300 listed companies found that the top quarter by its design index grew revenue 32 percentage points faster than their industry peers over five years, and shareholder returns 56 points faster. It is a correlation, from a consultancy that sells design work, and the differences between the other three quarters were "marginal". [10] Jeanne Liedtka, one of the field's leading researchers, has written of the "lack of a body of research that demonstrates systematically design thinking's ability to deliver value". [11]
So the ritual deserved its critics. The core, spending time with the people who have the problem before deciding what the problem is, never needed the sticky notes. It is the first diamond, and it is the part that just got more valuable.
The second diamond got cheap
In 2023 four innovation researchers at Wharton and Cornell compared 200 product ideas from students with 100 from GPT-4, and 100 more from GPT-4 shown seven good examples. US college-age respondents rated how likely they would be to buy each product. GPT-4's ideas averaged 46.8% purchase intent against 40.4% for the students. Of the 40 best ideas overall, 35 came from GPT-4. The authors estimated the model produced about 800 ideas an hour; a person working alone, about 20. [12]
Two details in the same paper matter for the rest of this piece. The best GPT-4 ideas were rated less novel than the best student ideas. [12] And purchase intent measures a well-specified product for a known customer: the problem had already been defined for everyone.
Other studies point the same way. On a classic creativity test, naming unusual uses for a rope or a box, chatbots in 2023 beat the average person, but "the best human ideas still matched or exceed those of the chatbots". [13] In a field experiment at Procter and Gamble with 776 professionals, people working alone with AI produced solutions at least as good as two-person teams without it, and teams with AI were about three times as likely as individuals without it to land in the top tenth. [14]
Producing options, prototypes and polished drafts, the work of the second diamond, now costs a prompt.
Cheap ideas look alike
Anil Doshi and Oliver Hauser gave 293 writers up to five story ideas from GPT-4 and had 600 readers judge the results. Stories written with AI ideas were rated more novel and more useful, and the less creative writers gained the most. [15]
The same stories were also more alike. The authors' summary is the trade in one sentence: "With generative AI, writers are individually better off, but collectively a narrower scope of novel content is produced." [15]
A reanalysis of another team's brainstorming experiments found the same narrowing. When people were asked to repurpose a tennis racket or a garden hose, 20 of 96 ChatGPT-aided ideas included the word "sprinkler", against 12 from people working unaided. [16]
The effect is real and modest. A 2026 meta-analysis of 19 studies found "a small but statistically significant homogenization effect" from AI use, and its authors read it as a reorganisation of creative diversity rather than a loss of creativity. [17] The most striking evidence for who benefits comes from four million artworks on an art-sharing site: text-to-image tools raised output, and the artists who gained most were those who "explore novel ideas and filter model outputs for coherence". [18]
That is the first diamond again. Exploring which idea is worth having, and filtering for the one that fits, are the parts the model does not do for you.
The creative mind lives in the first diamond
Employers seem to sense this. In the World Economic Forum's 2025 survey of more than 1,000 employers, creative thinking was the fourth most common core skill, named by 57%. Among skills expected to grow in importance, it ranked fourth, behind only AI and big data, networks and cybersecurity, and technological literacy. Design and user experience, as a separate skill, was core for only 25%. [19]
No study I found measures problem framing directly, so what follows is my reading of the pattern rather than a result. When the second diamond is cheap, everyone can produce polished answers, and the answers drift towards each other. What stays scarce is the work the model cannot start without: deciding what problem is worth solving, for whom, and recognising which of fifty fluent options is the one that fits. Rittel and Webber said it in 1973. The formulation is the problem.
That is what I mean by the rise of the creative mind. It is not a talent some people have. It is a set of habits that sit in the first diamond:
Spend time with the people who have the problem before you open the assistant. The Design Council's first phase is "speaking to and spending time with people who are affected by the issues". [1]
Write the problem down in one sentence, in their words, before asking for solutions.
Generate your own ideas first. The narrowing comes from starting where the model starts.
Then use the model for volume, and keep the filtering for yourself.
Before you ask a model for ideas, write the problem in one sentence, in the words of the person who has it.
Sources
[2] Simon: The Sciences of the Artificial, 3rd edition (1996), p. 111. As quoted by Keirl, PATT40 conference, 2023.
[3] Rittel and Webber: Dilemmas in a general theory of planning. Policy Sciences 4(2), 1973.
[6] Core77: Natasha Jen's "Design Thinking Is Bullsh*t" argument. August 2017.
[7] Vinsel: Design thinking is a boondoggle. Chronicle of Higher Education, May 2018.
[8] Ackermann: Design thinking was supposed to fix the world. Where did it go wrong? MIT Technology Review, February 2023.
[9] Fast Company: Ideo cuts a third of staff. November 2023.
[10] McKinsey: The business value of design. October 2018.
[11] Robbins and Fu: Blind faith or hard evidence? R&D Management, 2022; quotes Liedtka 2020.
[12] Girotra, Meincke, Terwiesch and Ulrich: Ideas are dimes a dozen. Working paper, July 2023.
[13] Koivisto and Grassini: Best humans still outperform artificial intelligence in a creative divergent thinking task. Scientific Reports 13, 2023.
[14] Dell'Acqua et al.: The cybernetic teammate. NBER working paper 33641, April 2025; not peer reviewed.
[15] Doshi and Hauser: Generative AI enhances individual creativity but reduces the collective diversity of novel content. Science Advances 10(28), 2024.
[16] Meincke, Nave and Terwiesch: ChatGPT decreases idea diversity in brainstorming. Nature Human Behaviour, May 2025; a reanalysis of Lee and Chung 2024. Figures from the preprint.
[17] de Rooij and Biskjaer: Generative AI makes creative output more homogeneous. European Conference on Cognitive Ergonomics, 2026.
[18] Zhou and Lee: Generative artificial intelligence, human creativity, and art. PNAS Nexus 3(3), 2024.
[19] World Economic Forum: The Future of Jobs Report 2025. January 2025.
One email when we publish. Research, product decisions, and what teams report back.
