The Prompt Is the New Brief

Part 2 of the “UX × AI” series.

In the first part of the series, we dismantled the most dangerous myth in the design community right now — that AI is coming to replace the designer. We replaced it with a more accurate and more useful frame: AI is your new intern. Fast, tireless, well-read, and completely dependent on you for direction, judgment, and accountability.

In this article, we go one level deeper. Because if AI is the intern, the prompt is the brief. And the designer who truly understands it — not as a metaphor but as a practical framework for working with these tools — has an immediate, significant advantage over every other practitioner who approaches prompting as a technical skill to be learned from scratch.

They do not need to learn it from scratch. They already know how to do it. They just have not recognized it yet.

A skill you already have; you just haven’t named it this way

Let me describe a professional activity and ask you whether it sounds familiar.

You are given a task that requires someone else to produce creative output on your behalf. Before they begin, you need to give them a clear understanding of what you want — the goal, the audience, the constraints, the tone, the format, and enough context about the specific situation to allow them to make good decisions without asking you about every detail. You need to be specific enough that the output is relevant, but not so prescriptive that you eliminate their ability to bring something useful to the problem. And you need to do all of this in writing — clearly, concisely, and with enough structure that someone who does not share your full mental model of the project can still produce something genuinely useful.

Every designer who has ever written a design brief, a research brief, a creative direction document, or a content strategy brief has done exactly this. Every UX researcher who has written a discussion guide has done this. Every design lead who has briefed a copywriter, an illustrator, or a junior designer has done this.

Prompting an AI is not a new skill. It is an existing skill — brief-writing — applied to a new surface.

The surface is different. The underlying competence is the same. And this is one of the most practically important insights in this entire series — because it means that the designers and researchers who have invested in the craft of clear, specific, contextually rich communication are already better at prompting than they realize. While the conversation about AI is dominated by talk of “prompt engineering” — a term borrowed from software development that positions prompting as a technical capability — the reality is that prompting is a communication and design skill. And design professionals have been building it for years.

“If AI is like an intern, then the prompt is your creative brief — it frames the task, sets the tone, and clarifies what good looks like. It is also your conversation script that guides how the interaction flows and how ambiguity is handled.” — Smashing Magazine (2025)

What a brief and a prompt have in common

The parallel between brief writing and prompting is not superficial. When you look at what makes a design brief effective — and what makes a prompt effective — the structural similarities are striking enough to be instructive.

A strong design brief specifies the goal — not what to make, but what outcome to achieve. “We need a homepage redesign” is not a goal. “We need a homepage that converts first-time visitors from paid search into newsletter subscribers for a financial planning product targeting working professionals aged 28 to 40” is a goal. The specificity of the goal determines how useful the output will be.

A strong prompt does the same thing. “Write me some onboarding copy” produces generic output. “Write three variations of a welcome message for a financial planning app — the user is a working professional in their early thirties who has just connected their first bank account. The tone should be encouraging but not patronizing. Maximum 40 words per variation.” produces something you can actually evaluate.

A strong design brief specifies the audience, with enough specificity that the person receiving the brief can make design decisions that serve that audience. Not “urban professionals” but “first-generation professionals in Tier 2 Indian cities, primarily mobile-first, with moderate financial literacy and a high degree of trust skepticism toward financial institutions.”

A strong prompt does the same thing. The more specifically you can tell an AI who the output is for — their background, their context, their knowledge level, their specific situation — the more relevant the output becomes. The AI has no access to your user unless you put them in the brief.

A strong design brief specifies constraints — budget, timeline, technical limitations, brand guidelines, and regulatory requirements. Constraints are not obstacles to creativity. They are the conditions within which creativity becomes useful. The unconstrained brief produces unconstrained output that serves no one.

A strong prompt specifies constraints with the same intention. Format constraints. Length constraints. Tone constraints. The things the output must not do, as well as the things it must do. The AI, like the junior designer, performs significantly better when it understands the boundaries of the solution space it is working within.

A strong design brief provides context — the story of how this project came to exist, what has been tried before, what the organizational situation is, and what assumptions are being challenged. Context is what allows the recipient of a brief to make intelligent decisions in the moments when the brief does not give them explicit guidance.

And a strong prompt does the same thing. The AI has no memory of your project, your organization, your users, or your previous decisions. Every prompt is a fresh context. The designer who provides rich context — who treats the prompt as a context-setting document rather than a command — consistently gets better output than the designer who treats it as a search query.

The anatomy of a brief that works as a prompt

Over 25 years of practice, I have come to believe that the most valuable thing a UX professional can develop — more valuable than any specific tool skill, more valuable than certification in any methodology — is the ability to communicate design intent with precision, specificity, and contextual richness.

That ability has always mattered. AI makes it matter more, immediately and visibly.

Here is the anatomy of a prompt that applies brief-writing principles directly. I use this structure regularly in my own practice, and I have seen it consistently produce better outputs across tools — whether I am using AI for research synthesis, microcopy generation, user flow mapping, or competitive analysis.

• Role: Tell the AI who it is in this context. Not a generic role, but a specific one that carries the expertise and perspective you need. “You are a UX researcher with expertise in financial inclusion in emerging markets” produces different output than “You are a UX researcher.” The role primes the model to draw on relevant knowledge and adopt the appropriate perspective.
• Situation: Give the AI the context it does not have. The product, the user, the business context, and the design challenge. Treat this as the background section of a design brief — enough context for the AI to make informed decisions, not so much that you are writing an essay.
• Task: Specify exactly what you want produced. Not a vague direction, but a precise description of the output — its format, its length, its structure, and the specific decision or action it is meant to support.
• Constraints: Name the boundaries. What the output must include, what it must not include, what tone it must adopt, and what assumptions it must challenge. Constraints are not restrictions on the AI’s creativity. They are the conditions that make the output useful.
• Evaluation criteria: Tell the AI what good looks like. This is the most underused element in prompting, and it is the one most directly borrowed from brief-writing practice. A strong design brief tells the recipient how the output will be evaluated — what criteria it must meet to be considered successful. A strong prompt does the same thing. “The output should be immediately actionable by a designer who has never worked on a government services product before” is an evaluation criterion. It shapes the output in ways that a task specification alone cannot.

This structure is not a formula to be applied mechanically. It is a framework for thinking through what the AI needs to know — the same thinking that good brief-writing requires.

“At its core, prompt engineering is about intentional communication. The designer who can articulate intent with precision — who has spent years crafting design briefs, research briefs, and creative direction documents — is already building this skill.” — Parallel HQ (2026)

Where designers are better at this than they think

The design community tends to approach AI prompting with a degree of intimidation that the skill does not deserve—and that does not match the actual capabilities that designers bring to it.

Designers are trained in ambiguity management. One of the foundational challenges of design practice is making useful decisions in the face of incomplete information — understanding the problem well enough to move forward without having resolved every uncertainty. This is exactly the challenge of prompting. You will never have complete information about what the AI needs to know. You need to make judgments about what context is most important, what constraints are most binding, and what the most likely failure modes of the output are. These are judgment skills that designers exercise constantly.

Designers are trained in iteration. Design practice is fundamentally iterative — you produce something, evaluate it against criteria, identify where it falls short, and produce a revised version. Prompting is iterative in exactly the same way. The first prompt rarely produces the best output. The practitioner who treats the first output as a starting point — who evaluates it critically, identifies specifically where it falls short and why, and uses that analysis to refine the prompt — produces consistently better results than the practitioner who either accepts the first output or abandons the tool because the first output was inadequate.

Designers are trained in audience empathy. The core discipline of UX is understanding the person who will use the output — their mental model, their context, and their needs. Applied to prompting, this means understanding the AI as a system with specific capabilities and specific limitations and designing the prompt to work within those capabilities while compensating for those limitations. The designer who approaches the AI the way they approach a user research problem — with genuine curiosity about how the system processes information and what inputs it needs to produce useful outputs — develops prompting fluency faster than the person who approaches it as a command-line interaction.

Designers are trained in specification. UI specifications, interaction specifications, content specifications — the ability to describe an intended output with sufficient precision that another person or system can produce it correctly. Prompting is a specification applied to AI. The more precisely you can specify what you want — and the more accurately you can predict the gaps in your specification that will cause the AI to make poor assumptions — the better your outputs will be.

The failure modes: Where prompting goes wrong

Understanding why prompts fail is as useful as understanding why they succeed. And the failure modes of prompting map almost exactly onto the failure modes of design briefs, which means that designers have a significant head start in diagnosing and fixing them.

• The vague brief. The most common failure in both briefing and prompting is insufficient specificity. “Make it better” is not a design direction. “Write me some research questions” is not a prompt. The output you receive from a vague prompt is, at best, generically competent — it could apply to any project, any user, any context. It has no chance of being specifically useful to your specific situation, because you have not told the AI what your specific situation is. The fix is the same as the fix for a vague design brief: add specificity to the goal, the audience, the context, and the constraints until the direction is genuinely actionable.
• The over-specified brief. The opposite failure is equally problematic and equally familiar. The design brief that specifies every element of the solution, leaving no room for the designer to bring genuine judgment to the problem, produces compliant output without intelligence. The prompt that prescribes every sentence of the output produces something that needed the AI’s involvement for nothing except the typing. The skill is in specifying the right things — the goal, the audience, the constraints, the evaluation criteria — while leaving room for the AI to contribute something genuinely useful within those specifications.
• The missing context brief. The brief that assumes the reader shares your full mental model of the project — that does not explain the background, the constraints, or the reasons for the decisions already made — produces output that is disconnected from the reality of the situation. The prompt that assumes the AI knows your product, your users, and your design history produces output that is generic precisely because the AI has had to fill the contextual gaps with assumptions. Providing context is not inefficient. It is the investment that determines whether the output is useful.
• The no-evaluation-criteria brief. The brief that does not tell the recipient how success will be measured produces output that may be creative and competent, but that cannot be evaluated, because neither party knows what they were aiming for. The prompt that does not specify what good looks like produces exactly the same result. The AI will produce something that appears complete. Whether it is useful is a question that only an explicit evaluation criterion can answer.

What this means for how you work right now

The insight that prompting is brief writing has a practical implication that goes beyond individual productivity. It changes the conversation about who in a product organization should be leading AI integration — and it is a conversation that design professionals need to be part of, now.

The assumption — driven by the “prompt engineering” framing — has been that AI fluency is primarily a technical capability. The people best positioned to use AI effectively are the people closest to the technology: engineers, data scientists, and technical product managers. This assumption is wrong in the UX context, and it is leading organizations to make AI adoption decisions that underuse the capabilities of their design teams.

The people in most product organizations who are best positioned to write effective prompts for design and research tasks are the designers and researchers themselves. They understand the user. They understand the design problem. They understand what good output looks like. They have spent their careers developing the brief-writing skills that make prompting effective. The UX professional who is sitting at the edge of their team’s AI adoption conversations — waiting to be told how to use the tools rather than shaping how the tools are used — is ceding ground that belongs to them.

The World Economic Forum’s Future of Jobs Report 2025 projects that AI and big data fluency will be the fastest-growing skills demanded by employers between now and 2030, with 39% of core skills expected to change significantly in that period. The designers who understand that their brief-writing competence translates directly to AI prompting fluency — and who build on that foundation deliberately — are positioned on the right side of that shift.

Applying LucyUX to prompt design

The LucyUX framework — Listen, Understand, Conceptualize, Yield — applies to the design of prompts with the same rigor it applies to any other design challenge.

• Listen: Before you prompt, listen to the output that simple, under-specified prompts produce. Not to evaluate the output for usefulness, but to evaluate it for what it reveals about what the AI is missing. The gaps in the output are signals about the gaps in the prompt. A response that is too generic reveals that your goal specification was insufficient. A response that misunderstands your user reveals that your audience specification needs more depth. A response that goes in an unexpected direction reveals an assumption you made that the AI did not share. Listen to the failures as data.
• Understand: Develop an accurate mental model of what AI tools are actually doing when they respond to a prompt. They are not reasoning. They are pattern-matching against vast training data to produce the most statistically likely continuation of your prompt. This is a powerful capability with a specific limitation: it is very good at producing outputs that look like outputs it has seen before and less reliable at producing outputs that require genuine novelty or contextual specificity. Understanding this shapes how you write prompts — you use the AI’s pattern-matching strength deliberately by structuring your prompts to align with patterns that exist in its training, while recognizing that genuinely context-specific insights require your judgment to extract from the AI’s output.
• Conceptualize: Treat every prompt as a design object. Before you write it, think through the goal, the audience, the constraints, and the evaluation criteria. Draft it. Read it as if you were the AI — does this prompt give you everything you need to produce a genuinely useful output? Where would you make assumptions? Where would you go wrong? Revise accordingly. The prompt that has been designed — rather than typed impulsively — consistently produces better outputs.
• Yield: Measure the success of your prompts not by the quality of individual outputs but by the quality of decisions those outputs support. Did the AI-assisted research synthesis lead to better design decisions than your previous process? Did the AI-generated user flow variations open directions that improved the final design? Did the AI-drafted microcopy require less revision than your previous copy-drafting process? These are the yields that tell the honest story of whether your prompting practice is working — and they are the only yields that justify the investment.

“The foundation of a great prompt is how well you can describe the task the AI needs to do. It sounds simple, but a single missed step or poorly chosen word can make or break the prompt.” — UX Studio Team (2025)

When prompting is treated as a technical skill, design loses

There is a specific risk in the current framing of AI prompting as “prompt engineering” — and it is a risk that the design community needs to be aware of.

When prompting is positioned as a technical skill, it is positioned as the domain of technical roles. Engineers become the people who write the system prompts that shape AI products. Product managers become the people who define the AI workflows. Designers — positioned as the “creative” counterpart to the “technical” AI work — are handed outputs to make look good, rather than being involved in defining what those outputs should be and what purposes they should serve.

This division reproduces one of the most persistent structural problems in product development — the separation of technical execution from user understanding — in a new and more consequential form. The AI system that is designed without UX involvement at the prompting and workflow level is the AI system that optimizes for what is technically producible rather than what is genuinely useful for the human being who will interact with it.

The UX professional’s role in AI product development is not to receive AI outputs and make them usable. It is to shape the AI’s behavior from the ground up — through the system prompts that define how the AI presents itself, through the workflow design that determines what the AI is asked to do and when, and through the evaluation frameworks that determine whether the AI’s outputs are serving users or merely appearing to. These are all brief-writing challenges. They are all UX challenges. And they belong to UX professionals.

The Nielsen Norman Group’s research on AI in design workflows makes this point directly: the designers who are adding the most value in AI-integrated product teams are not the ones who are using AI tools most fluently. They are the ones who are bringing UX thinking to the questions of how AI should behave, what it should produce, and how users should experience it. Fluency in using the tools is table stakes. Shaping what the tools do is the strategic opportunity.

Your action this week

Take a piece of work from your current workflow — something you have produced in the past week. A research discussion guide, a set of microcopy, a user flow, and a competitive analysis. Something that required you to communicate a design direction to a collaborator or client.

Now look at what you produced and ask: What was the implicit brief behind this piece of work? What goal was it serving? What audience was it for? What constraints was it working within? What does good look like?

Write that brief out explicitly — as you would write it for a junior designer joining your team who has never worked on this project.

Now use that brief as a prompt. Run it through an AI tool of your choice and compare the output to what you actually produced.

The comparison will be instructive in ways that are specific to your practice. Where does the AI’s output match the quality of your work? Where does it fall short — and what does that shortfall reveal about the knowledge and judgment you brought that the brief did not capture? Where does the AI produce something that surprises you — a direction or a variation you had not considered?

That last question is particularly important. The prompt that is written well enough to produce genuinely useful AI output is the prompt that reveals something. Either it reveals what the AI can do that frees your time. Or it reveals what you bring that the AI cannot approximate. Or it reveals a direction you had not considered. Any of these is a valuable outcome. None of them happens when the prompt is vague.

My perspective: What I actually believe

The design community is at risk of approaching AI with a combination of excessive intimidation and insufficient strategic ambition. Excessive intimidation — because the technical framing of prompting makes it seem like a skill that requires learning from scratch, when in reality it is a skill that designers have been building for their entire careers. And insufficient strategic ambition — because the conversation remains focused on using AI tools individually, when the more important question is how UX professionals can shape the AI systems that are going to define the next generation of digital products.

Brief-writing is the prompt. Prompt design is UX. The interface between human intent and AI output is a design challenge — and it belongs to designers.

The practitioner who internalizes this — who stops treating AI prompting as something to be learned from engineers and starts treating it as an extension of the brief-writing and communication discipline they have already developed — will find that their AI fluency grows faster than they expected. Not because they have discovered a technical trick, but because they have recognized a competence they already had.

Your briefs were always prompts. The recipient has changed. The skill is the same.

---

Up next in the “UX × AI” series: “Stop Calling It Empathy: AI Does Not Feel Anything.” The design industry has developed a tendency to describe AI in humanistic terms — AI that “understands” users, AI that “empathizes” with needs, and AI that “knows” what people want. This language is not just imprecise. It is dangerous because it shapes how design decisions are made in ways that consistently disadvantage the real human beings design is supposed to serve. In Part 3, we examine the myth of AI empathy — what AI actually does when it appears to understand users, why that is fundamentally different from the understanding that genuine UX practice produces, and what gets lost when we confuse the two.

---

References & further reading

• A Designer’s Guide to Prompt Engineering, UX Studio Team.
• Prompt Engineering for Designers: Mastering AI Workflows, Parallel HQ.
• Prompt Engineering vs. Prompt Design: The UX Perspective on AI Personality, LogRocket.
• What Is Prompt Engineering in Design? Interaction Design Foundation.
• Prompt Engineering for Designers: A Practical Guide, Medium/UX Raspberry.
• 10 Prompt Engineering Techniques for UX Designers, Eidos Design/Siavash.
• The Future of Jobs Report 2025, World Economic Forum.
• AI and Big Data Top Fastest-Growing Skills 2025–2030, World Economic Forum.
• Using AI for UX Work: Study Guide, Nielsen Norman Group.
• State of UX 2026: Design Deeper to Differentiate, Nielsen Norman Group.
• How to Integrate AI in Your Creative Design Process, Parallel HQ.
• The Design of Everyday Things, Don Norman.
• Usability Heuristics for User Interface Design, Jakob Nielsen.
• LucyUX Process: Listen, Understand, Conceptualize, Yield, Tushar Deshmukh.

Featured image courtesy: Roman Budnikov.