It is still too early to say how these roles will ultimately settle. Everything is changing very quickly, and what seems like a clear trend today may look different six months from now. But there is one hypothesis that seems worth exploring. Amid this reshaping, experience designers have an opportunity to take on a more strategic role.
A boundary that is becoming more permeable
For a long time, the process was fairly linear: design conducted research, conceptualized, prototyped, and handed off a solution for engineering to turn into software. The handoff was almost a mandatory rite of passage.
That boundary is starting to become more porous. A 2026 Designer Fund and Foundation Capital report on AI and design, based on more than 900 designers, found that 65% say they are taking on more product or engineering responsibilities, while 40% are seeing the opposite trend, with PMs and engineers increasingly taking on design tasks. It is a snapshot and does not necessarily represent the entire industry, but it is consistent with what many of us have been observing informally.
If this crossover continues to deepen, it does not mean designers need to become developers or that engineering will lose its specialization. It could mean that the same person is able to move further along the lifecycle of a solution before another discipline needs to step in.
Two possible directions for expansion
If this hypothesis holds, the designer’s role could expand in two directions at once:
- Toward the business: if producing an interface becomes increasingly easy, the differentiating value will no longer lie solely in knowing how to design it. Understanding business models, metrics, and constraints would allow designers to participate in defining the problem itself, not just the solution.
- Toward building: generative AI tools could enable what once ended as a specification or prototype to evolve into a real, functional experience, without necessarily having to go through another team.
Neither direction is guaranteed or automatic. It depends on whether each designer chooses—and is able—to develop these capabilities.
A debate that is far from settled
Not everyone agrees on how to interpret this moment, and that is probably a good thing. A 2026 update to the Designer Fund and Foundation Capital report shows that weekly use of AI for design tasks jumped from 54% to 91% in just one year, with 75% now using it daily. Half of the designers surveyed also say they have shipped AI-generated code to production, not only those who identify as “design engineers.” These figures support the hypothesis that the role is expanding.
But the same report also reveals friction. One-third of respondents feel that collaboration across teams has become messier as roles increasingly overlap, and there is still considerable uncertainty about what happens to junior designers when much of the execution can be automated.
There is also no consensus on how close designers should get to code. Some voices in the industry argue that the classic question of whether designers should code is already outdated, because it is now possible to build functional software without becoming an engineer in the traditional sense. Others argue that UX and frontend engineering require different ways of thinking, and that blurring those specializations can weaken designers’ ability to advocate for users amid other business pressures. We are more excited by the first interpretation, although we believe this is a healthy debate that remains open.
One limitation that comes up repeatedly in these conversations is quality. The most common complaint among designers already using AI is the lack of consistency in what it produces. One image that has circulated widely sums it up well: AI can quickly solve a large part of the journey, but what remains—the part that depends on understanding the brand, the user, and the context—is still human territory. Ultimately, this may be the best way to understand why judgment does not become less relevant in this scenario but, rather, more important.
When generating is easy, deciding matters more
What seems clearer is that these tools are lowering the cost of producing alternatives. A Lyssna survey of 100 UX/UI designers conducted toward the end of 2025 found that 93% were already using generative AI tools in their work, while 73% expected AI as a “design collaborator” to have the greatest impact on the discipline during 2026. KPMG reported something similar from the business side: 91% said they planned to increase their budgets for artificial intelligence applied to UX.
But the same Lyssna survey revealed an interesting tension. Fifty-four percent of designers said their clients want to jump on the AI wave without having clear use cases. It is a figure worth keeping in mind before assuming that generating more alternatives is, in itself, an improvement.
If a tool makes it possible to produce ten screens in the time it once took to produce one, the goal should not necessarily be to produce ten times as many screens. Instead, it could be an opportunity to explore more hypotheses and spend more time understanding the problem before building. Our ability to generate seems to be growing faster than our ability to decide what is worth generating. That makes design judgment, if anything, more relevant rather than less.
From specifying to building
All of this connects with a movement taking place on the engineering side, known as spec-driven development. Gartner described it in a February 2026 report as a way to scale AI-assisted development using machine-interpretable specifications instead of isolated prompts. The logic is simple. Rather than starting with code, you first clearly define what needs to be built, its constraints, and the expected behavior. That specification then feeds the AI agents that generate the code.
We see a natural connection between this logic and much of what UX has been doing since long before these technologies existed: researching a need, translating it into requirements, and maintaining consistency between the original problem and the solution that is ultimately built. If that connection deepens, handoff would no longer be thought of solely in terms of “how can I deliver this better?” but also in terms of “how much less needs to be reinterpreted from scratch?” It is a hypothesis, but one we find exciting.
What we are exploring at Baufest
With this hypothesis in mind, we created a pilot program for a group of experience designers to test AI tools and a spec-driven approach through an unusual challenge. The goal was to build a functional digital product from scratch and go through the entire process, from identifying the problem to deployment.
The entire Digital Experience Design team participated, and each person built their own functional product. Some focused on internal productivity, such as design system generators or assistants for processing research data that could eventually serve as accelerators within projects. Others emerged from needs identified internally or among clients. The entire team completed the program, which concluded with a demo day.
What we can say with greater certainty is where the limitation lies: getting a product to work is not the same as having software that is ready for an enterprise environment. A solution that will operate for years and evolve across multiple teams needs architecture, security, scalability, and testing. In that regard, software engineering remains irreplaceable. What has changed, we believe, is the distance between an idea and a working product in the hands of real users, making it possible to explore an opportunity or validate a hypothesis much faster.
An opportunity that is still open
We do not know for certain where all of this will ultimately settle. The pace of change means that any firm conclusion can quickly become outdated. But the question that interests us is not so much “what tasks will AI automate?” as another, more concrete one: What can a designer do now that they could not do before?
If the two directions of expansion we mentioned—toward the business and toward building—hold true, the advantage would not come from that expansion alone. It would come from simultaneously developing a stronger ability to understand context, make decisions, and critically evaluate what these tools produce. That judgment—the ability to question the obvious, understand needs, and make intentional decisions—is, for now, the one thing this hypothesis does not automate.