Last week we explored how AI was shifting design from craft to orchestration. This week, the data arrived to confirm what that shift actually means: the 10x engineer mythology just collapsed under its own weight, and taste became the only moat that matters—except most teams have no idea how to build one.
Figma's three-year study drawing from 8,403 total survey responses captures the precise moment personal productivity stopped being the thing to optimize for. Two years ago, 7% of respondents said AI changed how teams collaborate. Today it's 41%. When AI can 10x anyone, speed becomes table stakes. The immediate consequence: 90% report design and taste as equally or more important than before AI arrived, with 58% saying it's actually more crucial now. Critically, this isn't just designers protecting their turf—65% of developers agree. "Design, taste, and the user experience matters more than before, so we are trying to improve it every single day," one software engineer reported.
But here's the paradox: the same tools that elevated taste as the primary competitive advantage are actively undermining the ability to exercise it. The "jelly bean factory" problem with AI output has reportedly been discussed by designers like Andy Madrick at Notion: AI output looks right at a glance, following patterns that are statistically likely to work but rarely deeply considered. The result is a market flooded with algorithmically polished sameness—products that function perfectly well but feel interchangeable because they emerged from the same statistical distribution of "what usually works."
When everything works, people choose what feels intentional. The only gap that matters now is judgment.
When everything works, people choose what feels intentional. The gap between products that look functional and products that feel cared for used to be bridged by execution capability—could you actually build the thing you envisioned? That gap has collapsed. Now the only gap that matters is judgment: knowing what's worth building in the first place, and having the discernment to push past the first convincing output until you've made something that could only have come from you.
This creates what Figma's chief product officer Yuhki Yamashita describes as a common trap. Newer builders "latch onto the first idea and go deep," iterating and refining without ever questioning the starting point. AI agents accelerate this trap—they're endlessly helpful and agreeable, moving you forward but never pulling you out of tunnel vision. The old advice was to go broad first: map the option space into mutually exclusive categories, compare pros and cons in the abstract. But that approach fails too, staying at the wireframe level where you can't build real conviction about whether an idea actually works.
The teams pulling ahead have figured out how to go broad and deep simultaneously. At Figma, this means prompting AI to produce multiple interactive prototypes—different takes on the same problem, laid out side by side—then inviting teammates in to react not to abstractions but to concrete experiences. It's exploration in parallel rather than in sequence, using AI to test multiple directions at a resolution high enough to surface the problems that only show up when something feels real.
FloQast's Benjamin Ellis recently needed to rethink a workflow where accountants had to jump between multiple pages just to track down a single discrepancy. An early prototype looked promising, but the page had too many interconnected steps—each dependent on real data and logic—to evaluate from a mockup. Ellis built a working code prototype with an AI coding tool, complete with simulated backend and realistic customer data. The code-backed prototype let the team click through real scenarios where completing one step triggered the next. They immediately spotted flows that looked right on the surface but fell apart in practice—the kinds of interaction problems that a design file alone wouldn't have surfaced.
This isn't a productivity hack. It's a wholesale restructuring of when and how judgment gets exercised. Instead of designers creating specifications that developers interpret later—with all the judgment calls and craft decisions that interpretation entails—the prototype becomes the specification. The person with product vision can now test their ideas against intricate constraints directly, iterating on the actual thing rather than a representation of it. By the time the team commits to a direction, they've already pressure-tested it against real scenarios at a level of fidelity that used to require significant engineering investment.
Easton Thomas at Merkle used a similar approach from the opposite end. Rethinking how a telecommunications client organized products on its website, he asked Figma's AI agent to generate ten variations on one of his layouts, then ten more, until he'd generated about fifty options.
The pattern across these examples isn't "AI makes people faster." It's that cutting-edge teams have restructured their entire workflow to exercise taste at higher resolution and velocity than was previously possible. They prototype in code to test constraints. They treat prototypes as specifications. They generate dozens of directions in parallel to map the full possibility space before choosing. Most importantly, they've stopped accepting the first convincing output. Figma's chief design officer, reportedly a former sound engineer, describes taste as "the accumulated weight of intentional choices, repeated over and over again, until what you've made could only have come from you." The teams winning in this environment have built workflows that let them make more intentional choices, faster, with better information about the consequences.
The teams losing haven't attempted any of this. Figma's research identified four adoption patterns: grassroots (bubbling up from individual practitioners), directive (coming from leadership), nascent (early exploration), and unified (individuals and organizations moving together). At grassroots and directive companies, there's friction—people aren't on the same page, teams lack shared playbooks. "Ultimately it's more of an organizational gap, exposing some of our communication and knowledge gaps across squads," one principal product designer reported. These teams are still treating AI as a productivity add-on, something individuals use to work faster within existing workflows.
The gap between unified teams who've restructured around this reality and everyone else isn't narrowing. It's widening into a strategic chasm. Because taste isn't innate—it's exercised through iteration. The more rounds of refinement you can run, the more directions you can explore, the more interaction problems you can surface before committing, the better your judgment becomes. Teams that have rebuilt their process around going broad and deep simultaneously, around prototyping in code, around treating the canvas as a multiplayer space for collective judgment rather than individual craft—they're running more iterations, seeing more possibilities, catching more problems. The taste advantage compounds.
Seventy-six percent of product builders now do at least half their work on the canvas, with 60% spending the bulk of their time there. Even one in five developers would rather start projects on the canvas than in a terminal or prompt, because "you can't riff with someone else or put designs side by side" in a solo space. The number of designers participating in development doubled to 41% in the last year. The number of developers doing design work rose from 44% to 60%. These aren't just collaboration statistics—they're evidence of workflow transformation, of the boundaries between roles dissolving because the workflows that created those boundaries no longer make sense.
What looks like a taste crisis—the jelly bean factory churning out statistically plausible sameness—is actually a judgment crisis. The baseline quality of "good enough" has risen so dramatically that passive acceptance of AI output produces functional, polished products that no one remembers. Standing out now requires what Figma's research calls "craft": choosing rather than accepting, revisiting and interrogating each decision, pushing past the first few versions until the work has a point of view. This isn't about innate aesthetic sense. It's about building workflows that force you to exercise judgment repeatedly, at high resolution, with real information about trade-offs.
The teams that figured this out aren't moving faster than everyone else. They're moving faster together, with shared spaces for collective judgment, with processes that surface the right questions at the right time, with tools that let them test ideas at a fidelity that used to be impossible without shipping to production. The gap between them and typical teams isn't about AI adoption rates or individual skill. It's about whether you've rebuilt your entire workflow around the reality that taste is now the only defensible advantage—and whether you've structured your process to actually cultivate it.
When anyone can build and build quickly, the only edge is in what you choose and how well you shape it. Most teams are still optimizing for the wrong thing.