The bottleneck was never production
Six weeks ago I set up a weekly search for evidence about where design is heading as AI matures. Seven reports later, the clearest finding isn't about AI at all.
It's that making things was never the hard part. Now that it's cheap and fast, the real issue is more obvious than ever.
Teams already know what they should build
In June 2026, Design Sprint X surveyed 100 practitioners across 11 regions. 96% of them use AI in their work. Only 12% said their teams reliably ship what their design sessions produce.
Why?
Idea quality was named by 20%, poor facilitation by 14%. Everything above those is about people, not craft: who was asked, who agreed, who owned the work afterwards.
The problem isn't the making, and AI won't fix it.
Pendo's 2019 Feature Adoption Report found that 80% of features in the average software product are rarely or never used, and that 12% of features generate 80% of daily usage volume. That was true long before AI tools arrived. Teams have always built things that looked finished and weren't worth building. What's changed is that there's no longer anything to blame it on.
This one is vendor research. Pendo sells feature-analytics software, and the figures come from 615 of its own customer accounts, measured over three months and published in February 2019 — the oldest data in this piece.
Speed is measured. Quality mostly isn't.
Every big survey of designers using AI finds large speed gains. Figma's State of the Designer 2026, a survey of 906 designers, found 89% working faster, 91% saying AI improves their designs and 80% saying it improves teamwork.
What almost none of them ask is whether the work itself got any better.
One study did ask both. Figma's AI report of April 2025 surveyed 2,500 designers and developers. 78% said AI made their work faster. 58% said it made their work better.
So the real finding is smaller than the one people quote. In the only survey that asked the same people both questions, the gap between the speed gain and the quality gain was about twenty points. That was January 2025, and designers rated the quality gain lower than developers did.
The absence is the finding. The industry has measured its speed gain over and over, to the percentage point. It has mostly never asked the other question.
"Is it better?"
Buyers already know what makes the difference
In September 2026 the World Federation of Advertisers published remuneration research based on 69 multinational companies with a combined marketing spend of $147bn. It asked what actually makes an agency perform well, and scored the answers out of six.
| Driver of agency performance | Score (max 6) |
|---|---|
| Great briefing | 5.6 |
| Respect and trust | 5.2 |
| High-quality feedback | 5.1 |
| Financial incentives | 3.9 |
The top three all happen before the work starts. Money comes last. This is the buying side saying, in its own data, that the thinking before the work matters more than the deal around it.
71% of those surveyed work in marketing procurement. This is the buying department's own view of where agency value comes from. The people whose job is to negotiate price put price last.
Paying for time is on the way out
The same research tracked how those companies actually pay. Over fifteen years, the hour has gone from the main way of paying to a minor one.
The fall isn't recent, and it isn't sudden. Paying for labour was 54% of agency pay in 2011, 33% by 2022, and 19% now. Two-thirds of that drop happened before AI tools were on sale.
Fixed fees and pay for outputs rose from 20% to 33%. Pay linked to performance more than doubled, 9% to 21%. Looking ahead, 63% expect more performance-based fees and 46% expect more value-based ones.
AI is speeding up a change that was already well under way, not starting it. Only 20% of these companies have changed how they charge because of AI. 61% plan to. The repricing is still in its infancy.
The companies surveyed here buy advertising, not design. The report also says creative and production work has moved onto outcome-based fees more slowly than media buying has. So the shift is real, but whether it reaches design work in the same way is a guess still.
The talent is thinning at both ends
If judgment is the scarce thing, the obvious move is to buy more of it. Firms are trying. Figma's Design Hiring Study 2026, published in February, found 56% of hiring managers reporting rising demand for senior designers, against 25% reporting demand for juniors. More than 45% named collaboration, systems thinking and product strategy as critical skills.
But the route that produces senior designers is narrowing. Seyed Hosseini and Guy Lichtinger at Harvard tracked about 62 million workers across 284,974 firms, using LinkedIn CVs and job ads from 2015 to 2025. At firms that adopted AI, junior employment fell about 9% compared with firms that didn't, within six quarters. Senior employment didn't shift at all.
The fall came from slower hiring, not layoffs. Junior hiring at those firms dropped by 5.03 a quarter compared with the rest, while junior departures fell by 1.78 and promotions barely moved. Firms aren't cutting the juniors they have. They've stopped taking new ones on.
Meanwhile the people already making the decisions don't say they're ready for them. Doug Powell's Emerging Patterns study at Washington University surveyed 119 design leaders and interviewed 11. Among those 119 people there were 93 different routes into the job. Fewer than 30% studied design formally. Only 21% had a strong early mentor, and 41% had none. Asked how well they had been prepared for the job they hold now, they averaged 2.8 out of 5.
Powell's own summary of the gap: "the absence of a clear playbook for moving into leadership."
So the field is handing decisions to senior people, narrowing the path to becoming one, and those seniors say nothing prepared them.
We have a training problem the discipline has never had a chance to solve, because production work used to be the thing being judged.
What cuts against this
Two things, and both are stronger than I would like them to be.
Design is growing, not shrinking
The Design Council's Design Economy research puts UK design output at £136.7bn, up 40% from £97.4bn in 2019, against 24% for the wider economy. That is bigger than UK retail, at £114bn. Employment in design reached 2.27 million, up 15% since 2020, while total UK employment grew under 2.5%. Note the figures are from 2023, published in 2026 — they come from before most of what this article is about.
Around 80% of designers work outside the design industry, and £90.6bn — two-thirds of all design output — comes from designers working outside specialist design agencies. Employment of designers outside the design industry grew 20% between 2020 and 2025, against 6% inside it. Design has been migrating into the businesses it used to sell to for years.
The market is bundling strategy back in, not splitting it out
If judgment were becoming something you could sell on its own, you would expect to see firms selling it on its own. They are doing the opposite. In the two weeks to 25 October 2026 there were 27 agency and marketing-services deals, which put October at roughly 1.9 times the previous year's pace. One Strategy Group bought the brand strategy firm Fuller Impact Group and said the point was to offer an alternative to the big holding companies. Gunsmoke bought Chirripo so clients could get brand platform, strategy and production from one team.
Buyers may well rank briefing first and still want to buy it attached to the work. A plan nobody carries out is worth nothing — and the people who have to carry it out have good reason to be honest about it, because they are the ones who will live with the result.
The question this leaves open
Put it all in one line. Teams ship 12% of what their sessions produce, and they fail on agreement, not craft. Buyers score briefing 5.6 and money 3.9. Paying for time has fallen from 54% to 19%. Demand for senior judgment is rising while the path to becoming senior narrows, and the seniors rate their own preparation 2.8 out of 5.
All of it points the same way: to the decision, not the thing delivered.
And yet nobody buys decisions on their own. The market's answer so far is to put strategy back next to the work and sell them together.
So the open question isn't whether judgment is where the value sits. Four unconnected sources agree that it is. The question is whether judgment can be sold on its own, or whether it only has a price when it comes attached to the work that follows.
Six weeks of evidence doesn't answer that. It does suggest who will find out: the people being handed the decisions right now, with 2.8 out of 5 preparation and no playbook.
Sources
Every figure above was checked against the study that produced it, not the article that quoted it.
| Source | Date | Sample |
|---|---|---|
| WFA, agency remuneration research | 29 Sept 2026 | 69 multinationals, $147bn marketing spend; 71% marketing procurement |
| Design Sprint X, Is design thinking dead in 2026? | 28 June 2026 | 100 practitioners across 11 regions |
| Powell, Emerging Patterns: Design Leadership in 2026 | Spring 2026 | 119 design leaders surveyed, 11 interviewed |
| Hosseini & Lichtinger, Generative AI as Seniority-Biased Technological Change | rev. 27 Oct 2025 | ~62m workers, 284,974 firms, 2015–25 |
| Figma, Design Hiring Study 2026 | 10 Feb 2026 | Sample size not disclosed |
| Figma, State of the Designer 2026 | undated | 906 digital designers via NewtonX |
| Figma AI report, via TechRadar | 28 Apr 2025 | 2,500 designers and developers, surveyed Jan 2025 |
| Design Council, Design Economy | Sept 2026 | UK national accounts; GVA measured 2023 |
| Founder Perspective, agency M&A roundup | 30 Oct 2026 | 27 deals, 12–25 Oct 2026 |
| Pendo, 2019 Feature Adoption Report | 5 Feb 2019 | 615 Pendo customer accounts, three months of usage |