Agentic UX is the new black
Open your mind (to AI)
Hello, Goodbye Traditional User World
For years, the user experience (UX) field has been focused on designing for people — real humans with behaviors, motives, and issues. But now, we’re seeing a shift. Agentic UX (AX), or designing for AI agents that act on behalf of users, isn’t really new anymore — it’s quickly becoming a standard.
This doesn’t mean humans are out of the picture — it means they’re operating at a new level of abstraction. Somewhat of a hybrid I see forming. Instead of tapping through a UI, people more and more rely on ‘agents’ to carry out tasks on their behalf. Think: “Tell Siri to book me a reservation.” The UX is now about guiding how Siri thinks, decides, and acts — a whole new design model, while similar to a human — with less input needed.
LLMs x Humans
Introducing the elusive, yet tantalizing human side of technology — AI. I come from a long line of human-centric, digital consumer products, somewhat pre-AI explosion. Cursor, Claude, OpenAI, Gemini, MCPs, the list goes on have now entered the chat (pun mildly intended). We’re entering a landscape where LLM-driven agents aren’t just assistants — they’re intermediaries and even decision-makers. The UI is disappearing. It’s less about navigation and more about intent: how well the system understands what the user and how reliably it acts in alignment with that.
To understand it better, you should probably get familiar with these guys:
Cursor
An AI-native code editor built on top of VS Code, designed to pair-program with LLMs. It helps devs refactor, debug, and generate code.
Claude
Anthropic’s chat AI focused on helpfulness, honesty, and harmlessness. Known for long context windows and constitutional AI safety principles.
OpenAI
Think ChatGPT and the GPT models, pushing the boundaries of general-purpose AI. Their tools power a growing ecosystem of agentic and generative apps.
Perplexity
AI-powered research engine that combines web search with LLM reasoning. It cites sources in real-time, making it great for grounded, conversational research.
Gemini
Google’s AI engine powered by Deepmind and other magic. A powerful LLM capable of development, photo editing and more.
MCP
Model Context Protocol . MCP is an open protocol that standardizes how applications provide context to LLMs. Think of MCP like a USB-C port for AI applications. Just as USB-C provides a standardized way to connect your devices to various peripherals and accessories, MCP provides a standardized way to connect AI models to different data sources and tools.
We use it for AI agents. They handle data related jobs to be done, making developers lives easier.
Designers need to come to terms with a radical new reality: your primary “user” might not be a person anymore. It might be a reasoning model. Your Figma designs and prototypes now have to account for edge-case reasoning, interpretability, fallbacks, and invisible logic paths — not just layouts or interactions. We should be prepared to work in between as some agents that still deliver end user applications. An Agent may still be required to crawl a consumer product like Doordash to find and complete an order for you.
How 514 is embracing the situation, and what it means for design practice
At Fiveonefour, we’re leaning all the way in. Our design practice is evolving rapidly to meet this moment — we’re building agent-centric design systems, prototyping with real-time reasoning chains, and stress-testing how AI agents make decisions across user contexts. We’re not just designing screens; we’re designing behaviors, boundaries, and trust frameworks.
Common Agentic UX may include Code, Research, Writing, Generative UI, and or User Guidance. Ours may go beyond that.
We are introducing our own set of Agents targeting devs working with Data — Analyzing, Pipelining and more. The agents focus on key jobs to done, follow through on them and require little to no human input to follow through on. The workflows are highly conversational and built around different options users could take or that an agent may automatically take on your behalf.
Looking ahead
It’s different, conversational and there’s no going back, really. The temptation is to retrofit old paradigms onto new ones. We should reimagine design as a conversation not with the user directly, but with their digital paths more heavily top of mind. Things are more conversational and automated than before. And that means evolving principles, tools, and intuition.
Agentic Design isn’t just a trend. It’s the beginning of a new practice. One where UX doesn’t just serve the user — it provides agents with what they need to serve well. It’s a leap, sure, but it’s one we’re excited to take — and we’re building the foundation now, so probably best to grab yourself a shovel.
Learn more: