Who are we designing for now?

Between Humans and AI: The new design dilemma.

AI is disrupting more than the software industry, and is doing so at a breakneck speed. Not long ago, designers were deep in Figma variables and pixel-perfect mockups. Now, tools like v0, Lovable, and Cursor are enabling instant, vibe-based prototyping that makes old methods feel almost quaint.

What’s coming into sharper focus isn’t fidelity, it’s foresight. Part of the work of Product Design today is conceptual: sensing trends, building future-proof systems, and thinking years ahead. But besides the current momentum, we still have to focus on real problems that bring real value as of now. This balance is sometimes challenging, but also creates opportunities to reform our thinking and approaches.

As AI agents become embedded collaborators in our systems, designers face a powerful and pressing question: Who are we designing for now? Suddenly, we find ourselves in the middle of a new Experience dilemma: designing for both people and programs. That means exploring new personas and reconciling different approaches: emotional intuition, logical execution, and the coherence of both.

Let’s have a look at the pitfalls of this dilemma and explore what we have to consider while designing for both humans and machines.

AI agents are users, but not humans. Designing for them requires new UX abstractions.

Product Design 101 is all about understanding human experiences: how something feels, how intuitive it is, how it delights. But agents don’t feel. They parse. They tokenize. They operate on pattern recognition, context, probability, and strict interpretation.

Designing for agents means building interfaces that are accessible and intuitive but speak clearly to non-human readers. Think structured data, semantic HTML, accessible roles, predictable metadata, and context. If your interface looks like poetry to a human but gibberish to an LLM, you’re probably in trouble for future intertwined human-agent interactions. The machine needs to know where it is.

AI-friendly UX is about building interfaces with consistent syntax, machine-readable cues, and the kind of semantic rigidity that would make a programmer weep with joy. It’s less about aesthetics and more about contracts. If the machine can use your UI itself, it is future-proof for both humans and the machine.

Empathy remains essential even when designing for emotionless AI agents.

This sounds like an oxymoron. How do you have empathy for something that doesn’t feel anything? But stay with me.

Empathy here isn’t about imagining what the AI is “going through” emotionally. It’s about imagining what it needs from us to avoid breaking. What will it do when it encounters an unexpected input? Will it hallucinate? Run in circles?

Designing for agents is more like designing for a little ‘smartass’: they’ll give you exactly what you asked for, not necessarily what you meant. So we need to be very literal and need a kind of systems empathy: predicting failure modes, guarding against ambiguity, and ensuring fallback plans.

I like to think of it like accessibility design. You’re not empathising with someone’s emotions, you’re accounting for their very different set of capabilities. An AI agent might break with too little or too much context. Finding the right balance is crucial.

Designing for both humans and AI agents creates tension and opportunity.

Humans love novel, expressive interfaces. Agents love boring, consistent ones. As Product Designers we have to make both happy now.

The semantic layer of your UI means labels, roles, markup, and structure. This is now our double-double duty. It helps humans navigate and understand context, but it also provides the hooks and clues that AI agents need to function reliably. As these things need to be properly thought through to fulfil accessibility standards, we have to think of how we can make them even better, but for machines.

Most UIs weren’t built with semantics in mind. We got good at building shiny things, not necessarily readable things. And now, AI agents are forcing us to go back and retrofit meaning into our interfaces. The better your semantic layer gets, the more an AI agent can use and navigate it.

UX personas must evolve. Agents are now stakeholders, not just tools.

Your traditional persona might be: Jill, a content editor with 10 years of experience and a tendency to overuse exclamation marks.

Now you also need: Bot-1, a translation agent that prioritises accuracy, operates on API-based triggers, chokes on idioms, and needs as much semantic context as possible.

AI agents have capabilities, limitations, goals, and even preferences (although let’s be real: it’s just prompt engineering). They need inputs in specific formats. They expect consistent feedback loops. They don’t get bored, but they do get confused. Even worse, they get weird if they don’t have enough or too much data to perform their tasks.

So yes, build bot personas. Define their strengths and constraints. Define their toolset and goals. And then consider how they interact with your human personas. Where do they collaborate? Where do they conflict? When does Jill override Bot-1? When does Bot-1 need Jill to clean up her copy?

The future of UX is multi-agent systems. Designed for collaboration, not just control.

We’re moving from solo interfaces to ensemble casts: humans, AI agents, workflows, and decision-making chains. Think less “tool” and more “ecosystem.”

This means we need to design for:

• Visibility: Who did what? Was it an agent or a human?
• Coordination: When should an agent take over, and when should it wait for approval?
• Trust signals: How do you help humans trust an AI decision? How do you help an AI trust human input?
• Context: What information is needed for the AI to work? What information must be looped to the human?

I like the idea of progressive disclosure for agents. They don’t need all the fluff humans do, just the relevant metadata, tokens, and context.

And then there’s the UI equivalent of the team whiteboard: dashboards, activity logs, annotations, and semantic breadcrumbs that help agents and humans understand the shared workspace.

So… Who Are We Designing For Now?

The honest answer? We’re designing for everyone: the human and the AI agent. As humans seem to be the emotional and intuitive mess, the AI agent seems consistent and strictly logical in the first place. The opposite is true more often than you think: Humans love logical journeys and consistency as well. AI agents break if they have too many tools, too much context or way too complex tasks defined for them. We have to consider both in a way that is shaped to the needs of our new personas now.

Achieving both will empower humans to do their work with fewer headaches and be more creative, and the AI agent to produce reliable outputs and help the human where it should. A collaborative ecosystem of both personas is key here.

Our profession is changing. The rise of AI doesn’t mean we stop caring about human needs, but it does mean we expand our understanding of who “the user” really is. It’s no longer just Jill who oversees the large bright button that you intentionally placed on the top right. It’s also the invisible agent trying to summarise Jill’s intent, and perhaps feeding it into another system entirely.

The good old days of being called a User Experience Designer or User Interface Designer are gone. We need a new label. Maybe one that reflects this broadened horizon. Call it AI/UX/UI Designer if you like. Because in this new era, our work must account for the full constellation of users AND machines in a collaborative aspect.