Judgment cannot be prompted

When I started my career as a UX designer in advertising, the agency still had paper cutting boards and large stacks of materials to craft presentable campaign ideas for clients. There were two delivery guys on staff whose main role was to transport materials between print shops, the agency, and our clients. I never did any print ads that way because everything was digital by then, dependent on my Mac Pro’s GPU speed and FTP transfer speed. My apprenticeship as a junior designer lasted less than a year because I only needed to learn the software and get guidance from seniors. I still remember the first catalog I designed, which actually made me proud because the impression and look had a real designer’s touch—this time by me. I had gained the skill to transfer my imagination and judgment to digital designs.

We are in a similar era right now.

Roland recently demoed how Figma, a design system, and Claude can be combined to create interactive, on-brand prototypes. His prompt included a product line, a specific phase of a product selection flow, and a note that it was aimed at private customers. The output was an interactive HTML page with various states, showcasing Claude’s interpretation of a product selection feature for financial services. Designers used to complete similar tasks as part of job evaluations. Now, with a structured prompt, you will get output at the quality of a junior to mid-level UX designer, which means that designers competing for work have to bring more than they used to, just a few years ago.

The UX skill that does not scale

I place my emphasis on fundamental skills, which, regardless of technical progress, will remain universal. I treat AI as a translation tool that takes input and transposes it to another format. We humans have broader contextual memory and experiences that technology cannot replace, and to sustain this advantage, we need to maintain our direct connections with people. Statistics do not capture the complexities and intricate factors of everyday life that shape our decisions. To be accountable for the result, you need more than a good prompt — you need to understand what you are actually designing toward.

This is what working on a savings product taught me.

While conducting user research for the project, our team used interview questions that cross-checked one another. We asked people how much they would want to start saving with, and then, separately, we asked what they had done with their last bonus, raise, or unexpected windfall. The contradiction was evident: people who expressed a genuine wish to save had spent every windfall on goods and services, and even during the interviews, they did not recognize this. They were happy about their luck.

But the data also revealed a savings threshold at which compulsory savers changed their behavior and stopped withdrawing from their deposits. Combining these two findings gave us a specific goal — to help the customer reach that threshold, so that they could experience the value of saving smaller amounts and keep saving, this time consciously. The design solution was a digital coin jar that collected small, almost-invisible deposits, and we created a private celebration for the moment they reached their goal. The purpose was to change behavior by altering the belief that you need to save a large amount to become a real saver. Customers responded to the recognition with genuine appreciation. I would not have found the contradiction in a data summary.

In another project, I spent days working alongside emergency services professionals — people whose work carried significant responsibility, with no tolerance for error. I learned how they communicated, how they followed procedure, and how this mapped onto the software I was designing. I especially focused on one section that did not match the sequence described in the procedure and on how they were actually working with it. The procedure was correct, but it did not reflect how people actually worked in high-demand situations — and that mismatch added mental load. The design change that followed significantly improved task completion speed, because I was designing for people, not to meet requirements. I would not have found this in a requirements document.

Back to the Double Diamond

Regardless of how capable the tools become, we need direct exposure to our customers to develop a nuanced understanding, which is the only real basis for design judgment. The Double Diamond¹ — a model that divides design work into two distinct spaces: designing the right thing and designing things right — is built on exactly this premise. Roland’s demo showed that technology will enable us to work with designs at a scale we have never had before, but to know how to prompt, guide, and envision where the session should land, we need to make a judgment.

Designers need to develop and sustain their direct 1:1 connection with the people they design for — not through summaries or analytics, but through actual conversations and time spent in others’ contexts. We, UX designers, are problem solvers for our customers, and to develop our understanding of their needs, we must get familiar with them, the people.