Designing UX = designing prompts? Prompt engineering as the new UX skill

AI is indiscriminate. It will respond to anything (vague, misaligned, incomplete )and return something that looks finished.

We’ve been here before. We’ve seen wireframes built on assumptions and prototypes shipped without hypotheses.

Now we’re watching the same happen with prompting. Vague instructions fed into AI are producing surface-level answers.

The crux is in how we shape the conversation. If we don’t apply the same rigour to prompt creation that we do to design, we allow randomness to creep into decision-making under the illusion of intelligence.

Prompting is not separate from our discipline anymore.

It is our discipline.

We either take control of that, or we delegate our thinking to a model that doesn’t understand users, business goals, or consequences.

---

📌 What’s Inside

1. A prompt as a design brief
2. Prompting is an interaction design at the system level
3. Iteration isn’t optional. It’s the process.
4. Prompt libraries = the next design systems?

---

✍️A prompt as a design brief

To treat a prompt as anything less than a design output is a professional mistake.

Prompts demand structure.

There are usually four dimensions that define whether they will produce value or waste:

1️⃣ Empathy

Every output serves someone. Who is it for, and what are they feeling? First-time users, power users, hesitant adopters. They are emotional states with consequences. A prompt that ignores tone, comprehension level, and cognitive load will reflect that ignorance in its output.

2️⃣ Intent

If we don’t define the goal, the system will invent one. Asking for “UX ideas” is like asking to “make something clean.” Prompts must specify the outcome: a flow, a variant, a summary, or a rationale. Format and scope must be built in.

3️⃣ Boundaries

Design without constraints is pure negligence. Your prompts must surface what cannot be violated: factual accuracy, legal compliance, inclusive language, platform-specific limitations etc. AI doesn’t know what not to do unless we draw the line.

4️⃣ Verification

Don’t ship without testing. Prompts are no exception. Who will review the AI output? Against what criteria? For what risks? Treating prompts as final is how bad decisions scale fast.

🔧Prompting is an interaction design at the system level

Prompts are like the interface layers between our intent and an indifferent system. If that system lacks context, it will likely hallucinate. It will fill in blanks with bias and assumptions.

Writing a good prompt is less like giving an order and more like briefing a junior team member. You outline the problem, define constraints, and frame the tone. The model isn’t creative. It’s reactive. Its output reflects how clearly we described the task.

We can’t afford to let prompting become a wildcard in our workflows. It must be intentional, versioned, and treated with the same care as any other interaction design input.

Just as we design experiences, we must also design our conversations with AI.

🔄Iteration isn’t optional. It’s the process.

The first prompt will rarely work. That’s not failure but a signal.

Just like early wireframes, AI outputs need to be interrogated, tested, and refined. The feedback loop is where quality emerges.

If the output lacks clarity, the prompt didn’t clarify the audience. If it lacks focus, the intent was underspecified. If it contradicts brand or accessibility principles, boundaries were absent. The only way to improve the system is to design better inputs.

Prompting is fast, but speed is no substitute for structure. One well-iterated prompt is worth more than ten shallow ones.

A mature design process doesn’t treat iteration as waste, it treats it as where the actual work happens.

📚Prompt libraries = the next design systems?

We already build for scale. Prompting needs the same systematisation.

Reusable prompts, tested, scoped, and annotated, can be centralised and shared across teams. A design system for prompting is no different from a typography spec: it creates guardrails, reduces duplication, and raises the floor of quality.

It’s about making sure that what we ship (even when generated) is aligned with user needs, business strategy, and ethical practice.

---

We are all already prompting. Often badly.

But the danger isn’t just bad outputs.

It’s decisions made on the back of those outputs. If we allow poor prompting to influence product direction, we are complicit in letting assumptions shape strategy.

We have the skills: user empathy, strategic framing, system thinking. Prompting rewards exactly what we’re trained to do. But it punishes vagueness, laziness, and hesitation.

If we wait until the output to intervene, we’ve already lost the opportunity to shape it.

At the end of the day, prompting is not supporting work anymore. It is often the core of design.

It requires the same rigour as any prototype or design flow.

Because what goes into the input box is often the first move in the design process.

---

Subscribe on Substack⬇️

If you’ve found this content valuable, here’s how you can show your support.⬇️❤️

You might also like:

📚 Sources & Further Reading

1. Prompt Engineering by the Interaction Design Foundation

Share this article: