Short answer: no, but roughly half of what a UX designer did in 2022 is now a fifteen-minute task, and the half that remains is the half nobody was measuring. This page takes the role apart function by function, shows which functions have been absorbed, and gives you a way to work out where your own week sits. The employment picture is not the story people expect: the US Bureau of Labor Statistics puts web developers and digital designers at 220,100 jobs in 2025 with 5 percent projected growth to 2035 and 2025 median pay of $99,520, ahead of the 3.5 percent projected across all employment.
The Four Functions Inside the Role
Almost every argument about this topic goes wrong because “UX designer” describes four different jobs that happen to be held by one person.
Interface production. Screens, states, components, specs. This is what most people picture, and it is the function under the heaviest pressure.
Interaction and system design. How the thing behaves across time and edge cases: what happens on failure, on slow networks, on first use, on the hundredth use. Partly assisted, largely not.
Research and evidence. Finding out what users actually do and what they cannot do. Synthesis is assisted; the decision about what the evidence means is not.
Organisational work. Persuading a product manager to cut scope, negotiating with engineering about feasibility, deciding what not to build. Untouched.
The pressure runs almost exactly in that order. A designer whose week is 80 percent function one has a very different outlook from one whose week is spread across all four, and both are called UX designers.
What Has Actually Been Absorbed
| Function | Status | What is left |
|---|---|---|
| Screen and layout production | Heavily absorbed | Deciding which direction to pursue |
| Component and state variants | Heavily absorbed | System architecture and naming |
| Interface copy | Heavily absorbed | Voice, legal wording, edge cases |
| Design system documentation | Largely absorbed | Governance, versioning, adoption |
| Research synthesis | Assisted | Interpreting what users did not say |
| Usability session logistics | Assisted | Running the session, reading the participant |
| Accessibility auditing | Partly automated | Judgement about real assistive-technology use |
| Prioritisation and scope | Not absorbed | All of it |
| Stakeholder alignment | Not absorbed | All of it |
If you total the hours a typical mid-level designer spent on the absorbed rows in 2022, it is a large fraction of a working week. That is the honest scale of the change, and it explains why the experience of designers feels far worse than a 5 percent growth projection suggests.
Why the growth projection and the lived experience disagree
Two things are happening at once. The total volume of design work is not falling, because software keeps being built. But the number of people needed to produce a given amount of interface has fallen sharply, and hiring has concentrated at the senior end.
The result is an occupation that grows in aggregate while junior hiring is difficult, which is exactly the pattern that produces widespread anxiety alongside positive official numbers. Neither the optimists nor the pessimists are lying; they are describing different parts of the distribution.
BLS also published AI exposure categories with the 2025-35 projections, sorting occupations into Low, Moderate, High and Very high relative exposure, and states plainly that exposure “does not imply job loss, productivity gains, automation probability, or wage effects.” Design work registers on those measures because a model can produce similar artefacts. Producing the artefact was never the whole job.
Designing for Systems That Are Sometimes Wrong
There is one area where demand has appeared faster than supply, and it is worth understanding because it barely existed three years ago.
Conventional interface design assumes deterministic behaviour. Press the button, get the result. Every pattern in a design system, every error state, every confirmation dialogue rests on that assumption. Probabilistic systems break it, and the resulting design problems have no settled answers.
How do you show a user that an answer is probably right without either overstating confidence or making the product feel useless? Where do you put the correction path, and how do you make correcting cheaper than starting again? What does an undo mean when the action was an interpretation rather than a command? How do you design an interface that stays useful when the underlying model is replaced next quarter and behaves slightly differently? How do you disclose that something was generated without the disclosure becoming visual noise that users learn to ignore?
Teams shipping AI features are running into all five of these and mostly solving them badly. The designers who can reason about them are being hired quickly and paid well, because the alternative is a product that either misleads users or annoys them into abandoning the feature.
The reason this is worth naming in an article about replacement is the irony: the technology that absorbed a large part of interface production simultaneously created a design problem hard enough to justify hiring. That is what technological change usually does, and it is why occupation-level projections keep confounding the confident predictions.
What to Know Before You Change Anything
Hiring criteria moved before the work did. Reviewers now assume anyone can produce attractive screens, so a portfolio of attractive screens carries almost no signal. What carries signal is a documented decision that changed because of evidence.
Speed created a new failure mode. Teams now generate five directions in the time it used to take to produce one, and then choose badly because nobody framed the criteria. Being the person who defines what “better” means for this problem is the most valuable thing in the room.
Design engineering is the strongest single differentiator. Designers who ship working interfaces, whether hand-written or generated, remove a handoff and a translation loss. This shows up in hiring more than any other capability right now.
Accessibility is a genuine specialism, not a checklist. Automated auditing catches a fraction of real barriers. Understanding how someone actually navigates with a screen reader is expert knowledge that remains scarce.
Titles are drifting. Product designer, design technologist, design systems engineer and UX engineer are absorbing work that would have been posted as UX designer. Searching for one title makes the market look emptier than it is.
Where Designers Are Being Hired
- Complex and regulated products. Healthcare, finance, industrial and government software, where constraints dominate and mistakes are expensive.
- Design systems ownership. Naming, tokens, governance, versioning and adoption across many teams. Political as much as technical, and durable for that reason.
- Design engineering. Building the thing, not specifying it. Currently the clearest route to higher compensation in the field.
- Research with teeth. Not usability testing as a formality, but research that changes roadmaps, which requires standing as well as method.
- Conversion and experimentation. Teams that run experiments need designers who can form a hypothesis and read a result rather than only produce the variants.
- AI product design. Designing interfaces for probabilistic systems is genuinely unsolved: how to show confidence, how to design for wrong answers, how to build correction paths. Very few people are good at it yet.
A Decision Framework for Your Next Move
- Producing screens from clear requirements. Most exposed. Within twelve months, move toward either research with decision authority or building. Both are learnable and both change how you are evaluated.
- Mixed role in a product team. Reasonably placed. Your leverage is in framing problems: bring the criteria, not just the options.
- Specialist in research, systems or accessibility. Strong. Your risk is being seen as slow if you are not fluent with the tooling, which is a cheap gap to close.
- Trying to enter the field. Do not compete on portfolio volume. One project in a domain you understand, with a measured outcome, beats six speculative redesigns of consumer apps everyone has already redesigned.
The test that cuts across all four: in your last three projects, what changed because of something you found out? If you can answer concretely, you are doing the durable work. If not, that is the thing to build, and it does not require permission from anyone.
Common mistakes right now
- Presenting generated options without a stated basis for choosing between them.
- Treating an automated accessibility pass as accessibility work.
- Adding tool names to a portfolio instead of showing a decision the tools helped you reach.
- Producing more concepts because it is now cheap, which pushes the decision problem downstream.
Building the Fluency Hiring Panels Are Now Testing For
Interviews in this field increasingly include a version of the same question: where do you use these tools, where do you refuse to, and why. A vague answer reads as either avoidance or uncritical enthusiasm, and both cost offers.
A good answer requires having actually worked out where generation degrades quality in your own practice: which parts of research it flattens, where generated copy fails legal review, why it produces inconsistent system behaviour across states. Learning that deliberately, in a structured sequence rather than by accumulating habits, gives you both the fluency and the vocabulary to explain it. Pairing that with a certificate and one project that demonstrates the workflow gives a panel something concrete to evaluate. If you want a structured route in, explore Coursiv AI lessons and check current plan details on the official site.
FAQ
Will UX designers be replaced by AI?
Which UX skills are safest?
Is it still worth entering UX?
How do I show AI fluency in a portfolio without it looking gimmicky?
What is the highest-value thing to learn this year?
Your Next Step
Take last week’s calendar and split every hour into producing an artefact or deciding something. Then take your strongest case study and rewrite it around the second column: what you believed, what you tested, what surprised you, and what changed. Most designers discover their best work is described in terms of what they made rather than what they worked out, and fixing that is the fastest available improvement to how you are evaluated.