Designing for a New Kind of User (aka the Agentic Revolution)
Understanding agentic AI and the shift to agent-first design
Alright, folks, buckle up. We're diving into some seriously transformative stuff—the rise of AI agents and what it means for designing digital experiences. If you're anything like me, you're probably used to thinking about users as... well, humans. But that's changing and fast.
These AI agents aren't just some futuristic fantasy anymore. They're here, they're learning, and they're going to be interacting with our products in ways we're only just beginning to grasp. Think about it: instead of a person clicking buttons on a website, you might have an agent autonomously booking a flight or managing a whole shopping list.
So, what does this mean for us as designers? It means we need to think beyond UX and consider AX—agentic experience.
AX: It's Not Just UX 2.0
Now, you may be thinking: "AX? Is that just UX with a fancy new name?" Not exactly. While there's overlap, AX brings a whole new set of considerations to the table.
You know how we obsess over making things look good for humans? Button sizes, typography, color palettes? Agents, for the most part, don't care about that stuff. They're focused on efficiency, on getting the job done.
Think of it this way: We spend a lot of time making onboarding flows smooth and delightful for people. Agents? They need access now. There's no time for tutorials or fancy animations. It's a different ballgame.
Okay, So How Do We Design for Agents?
This is where it gets interesting (and, honestly, a little challenging). We're going to have to rethink some of our core design principles. Here are a few things that are top of mind for me:
APIs are king (and queen): Agents primarily interact through APIs, so these need to be rock-solid. We're talking clear documentation, predictable structures, and maybe machine-readable formats. Think OpenAPI on steroids.
Data, not just display: We're used to designing for human eyes, but agents need data they can easily parse. This might mean creating dedicated data feeds or APIs tailored for agent consumption.
Security is paramount: If agents are going to be acting on our behalf, security becomes even more critical. We'll need to figure out authentication and authorization models that work for AI, not just people.
Agent-friendly error handling: Agents need to understand why something went wrong, not just that it did. Clear, machine-readable error messages are essential so they can try again or find alternative solutions.
Think workflows, not features: Instead of designing isolated features, we need to consider how agents will combine them to accomplish complex tasks. This means really understanding agent workflows.
A Glimpse into an Agentic Workflow
Let's walk through a simplified example to make this more tangible. Imagine a user wants to plan a weekend trip:
User input: The user tells their AI agent something like, "Plan a weekend trip to a beach destination within 300 miles, leaving next Friday."
Agent task breakdown: The agent breaks this down into sub-tasks:
Find a list of beach destinations within 300 miles.
Check hotel availability for those destinations for next Friday.
Compare prices.
Present the top three options to the user.
Agent actions:
The agent uses a "Destination API" to get a list of locations.
It uses a "Hotel Booking API" to check availability and prices.
It might even use a "Calendar API" to block off the weekend on the user's calendar (with permission, of course).
Agent output: The agent presents the user with a concise summary of the three trip options, including hotel details and prices.
User decision: The user selects an option.
Agent confirmation: The agent uses the "Hotel Booking API" to confirm the reservation and the "Calendar API" to update the user's schedule.
Can you see how different that is from how we normally interact with travel websites? The agent is proactive, chaining together different services to achieve a goal.
Why Should We Care?
You might be wondering, "Is this really important? Is this AX thing just a flash in the pan?"
I don't think so.
This shift towards AI agents has the potential to fundamentally change how software is built and used. Companies that embrace AX and build agent-friendly products will thrive in this new world. It's about staying relevant, staying competitive, and unlocking a whole new level of efficiency and automation.
Let's Build This Future Together
This whole area is still pretty new, which is exciting. We get to help define what AX looks like and what the best practices are. I'm personally diving deep into this, and I'd love to hear your thoughts. Are you seeing this in your work? What challenges and opportunities do you see?
Let's chat in the comments.