When people talk about 'prompt engineering', it often sounds like a technical specialism — something adjacent to programming, a skill that belongs to a particular kind of person. I'd argue that designers are better positioned to do it well than almost anyone, and most don't realise it yet. The cognitive work of writing a good prompt is almost identical to the cognitive work of writing a good brief: you need to be clear about the problem, honest about the constraints, and specific about what a good outcome looks like. Designers spend their careers doing that. We just haven't been applying it here.

Why most prompts are bad

Most prompts fail for the same reasons most briefs fail: they're vague about what they want, they don't specify constraints, they don't explain context, and they measure success by the feeling of getting a response rather than by whether the response is actually useful. 'Give me some ideas for improving this onboarding flow' will produce ideas that are generic, untargeted, and very possibly irrelevant to the specific situation you're working in. The AI has no context. It fills the gap with generalities.

The other common failure is conflating instruction with intention. Saying 'write a summary of this document' tells the AI what to do but not why you need it or what you'll use it for. A summary for a design crit is different from a summary for a senior stakeholder who hasn't read the research. The same document, the same task, two very different outputs. If you don't specify the purpose, you'll get the default — which is rarely the right one.

The elements of a good design prompt

A well-constructed prompt for design work has four components. The first is context: who the user is, what the product does, what stage of the design process this is, and what constraints are already fixed. The second is the task: precisely what you're asking the AI to do, including the format you want the output in and the level of detail appropriate. The third is constraints: what the output should not do, what assumptions it should not make, what territory is out of scope. The fourth is success criteria: how you'll judge whether the output is useful.

In practice, this might look like: 'I'm designing a renewal journey for an enterprise software product. Users are IT administrators who manage 50-500 seat licences. They're time-poor and risk-averse. The current flow has three steps and a twelve percent drop-off rate. I want you to generate five alternative information hierarchies for the first screen only — not full flow redesigns. Each should prioritise clarity about what the user needs to do next. Do not suggest changes that require user account changes or new API integrations — these are outside scope for this sprint. A good output would let me clearly see five meaningfully different approaches that I could sketch and test within a day.' That prompt gets something useful. 'Redesign the renewal flow' does not.

How design thinking maps to prompting

The design process trains you to separate problems from solutions, to name constraints explicitly, and to define success before you start. All of that maps directly to better prompting. 'How Might We' questions, for instance, translate naturally into prompt structures: they're specific enough to be actionable, open enough to allow creative responses, and framed as problems rather than solutions.

Iteration also maps. Good prompting isn't about getting the perfect output in one shot — it's about treating the first response as a starting point and refining from there. 'That's close — now make the third option more specific to the scenario where the user has renewals due for multiple contracts simultaneously' is prompt iteration. It's exactly what you do in a good design crit: you respond to what's in front of you with targeted direction, not a wholesale rejection.

The skills gap that isn't where you think

The designers I see struggling with AI tools are often struggling because they're treating prompts as search queries — short, keyword-based, expecting the AI to infer everything they left out. The designers who get value from them quickly are the ones who treat prompting like briefing — with the same care and specificity they'd bring to writing a creative brief for a junior designer or an external agency.

If you've ever written a good design brief — clear problem statement, explicit constraints, defined success metrics, named assumptions — you already know how to write a good prompt. You just need to apply the same standard. The AI will meet you at the level of specificity you bring. Bring more, and the output improves accordingly.