State of Design 2026 · Part 3 of 5
The anxious middle
The people who code most with AI feel more valuable. Most of the profession is on the other side. This article is about them.
Sentiment by role
How will AI affect your role over the next two years?
- less secure
- more valuable
* Researcher: n=23, directional. The three percentages per role do not sum to 100%; remaining categories are not in the extract.
View table
| Role | more valuable | less secure | Same | Net (pts) | n |
|---|---|---|---|---|---|
| IC Designer | 24.8% | 32.4% | 21.7% | −7.6 | 862 |
| Lead / Principal | 43.2% | 20.0% | 24.6% | +23.2 | 280 |
| Manager / Director | 35.4% | 23.6% | 22.4% | +11.8 | 161 |
| Design Engineer | 50.0% | 10.6% | 10.6% | +39.4 | 94 |
| Non-designer | 31.6% | 24.6% | 24.6% | +7.0 | 57 |
| Researcher | 17.4% | 39.1% | 21.7% | −21.7 | 23 |
Source: UX Tools, State of Prototyping Spring 2026 (CC BY 4.0)
I ordered the roles by the vibe coding rate from the first article. The sentiment order is nearly identical. Design Engineers: 50% feel more valuable, 10.6% less secure. Researchers: 17.4% more valuable, 39.1% less secure (n=23, directional reading).
More valuable minus less secure
The net, role by role
I did simple arithmetic: percentage saying “more valuable” minus percentage saying “less secure.”
- Design Engineer: +39.4 points
- Lead / Principal: +23.2
- Manager / Director: +11.8
- Non-designer: +7.0
- IC Designer: −7.6
- Researcher: −21.7 (directional)
Among designers, the IC is the only one with a negative net. And the IC is 58.3% of the sample.
Tension
The group that is the profession
Workflow satisfaction
Satisfaction rises with the dose, step by step
- Overall mean 6.49
Δ None → Nearly all: +1.46
View table
| Tier | Mean (1–10) | n |
|---|---|---|
| None (0%) | 5.93 | 557 |
| Occasionally | 6.12 | 274 |
| About half | 6.83 | 188 |
| Most of it | 7.16 | 258 |
| Nearly all | 7.39 | 201 |
Source: UX Tools, State of Prototyping Spring 2026 (CC BY 4.0)
Satisfaction with your own workflow, 1 to 10. People who generate nothing with AI give it 5.93. People who generate nearly everything give it 7.39. A difference of 1.46 points, and every intermediate step goes up.
Caveat
Correlation, not cause
Hypotheses
Two possible readings
Reading 1: adopting reduces fear. People who try it discover that AI expands what they can do alone, and start feeling more valuable. The fear is fear of the unknown.
Reading 2: fear blocks adoption. People who feel threatened avoid the tool that represents the threat, and fall further behind. The fear is rational and self-confirming.
I lean toward the second, because it explains the seniority paradox from the first article. Lead and Principal have autonomy and job security; they can experiment without the experiment looking like a confession. The IC, measured on delivery, can't.
Reading
A management problem, not an individual one
The useful question isn't “why doesn't the IC adopt.” It's “what in the structure of the IC's work prevents it”. Deadlines measured in screens delivered. No protected time to experiment. Tools that need security approval. No technical peer to ask without embarrassment.
None of those items is solved by an AI course. All of them are solved by decisions made by whoever organizes the work. And the cost of not solving them is having most of your team feeling like they're on the losing side of a transition that, by the numbers, the people who cross it rate as positive.
Limits
What the survey doesn't say
Question
For the audience
Does fear block adoption, or does adoption reduce fear? And what changes in your answer if you're an IC?
What to do on Monday: if you lead, set aside protected, explicit time for ICs to experiment, with a real task and no delivery expectation. Pair a Design Engineer with an IC for one sprint. If you're an IC, pick a small prototype and generate the whole thing with AI this week, just to see what happens to your own net.
Continues
In the next article
Asked to pick three investment priorities, designers put AI-generated code first, agents second and design systems with tokens third, ahead of canvas tools themselves. The fourth article is about why the design system became AI infrastructure.