Agentic Design Systems-Part 1

John Rodrigues explains agentic design systems: design systems with machine-readable metadata and AI agents that automatically detect and fix UI inconsistencies. Product teams waste 34% of development time rebuilding existing components; mature design systems cut defects by 34% and speed delivery by 40%.
Chapters
Welcome to another video. In this video, we're going to talk about agentic design system and what it is and where it's evolving and some patterns that I see. Just here in Silicon Valley, around $128 billion is invested in, you know, AI development and things like that. Very few people solving the AI slop problems, right? There's so much of AI slop, especially like in designs and things like that. There are companies like Taste Labs and stuff coming and trying to solve this. Usually people say that like you solve this slop problem by using taste, but I don't think so, like, taste is the solution for it because taste can be subjective and it might probably immerse like consumer brands and stuff. Companies are rebranding and exploring new visual language and stuff. It makes a lot of sense, but mostly in B2B products or larger teams or growth stage team where you're not inventing the visual language all the time, they work in systems. I mean, there's also visual language. I mean, there's definitely some kind of a visual language there, but then they usually work in like systems and it helps to scale or work in a mature organization. Design systems are core part of what you do. And taste, and honestly, like taste can be subjective in nature. Like my taste would look different from your taste. So usually that does not like end up, you know, this does not help us to create a consistent system in our organization. So I feel like that's where like the design systems and agentic layer to it becomes the most important. The agentic layer becomes important because the speed of shipping right now is like moving fast. Like then engineering teams are like moving fast. Design teams has trying to catch up with that phase. But even the market right now, the speed of product development is increasing. So we kind of have to adapt to that market change and also ability to able to ship fast without having to ship AI slops. I solve this problem, inconsistency problem, ability to ship fast. In one of the startups that I was like consulting, and I built like a mobile design system for them, which was like agent-ready. And then I've been experimenting a lot with building agents for my studio, and there are like multiple agents that helps me with my day-to-day operations, and they are proactive in nature. So that kind of led me to like combining this operational problems design systems team have and working in both the sides. I've seen the frustration. I believe like agents and, you know, having a proper, better automation that automates some of the manual repetitive work can reduce that, you know, frustration and help us to like ship faster. I was able to build like an agent-ready design system for my, the project that I was working on. And eventually in the upcoming videos, I'll also show some of those proofs. But which was mostly about making the design system agent-ready through like metadata and components able to read by agents and all that stuff. I'll share more details about it soon, but that proactive layer and the self-healing loops and everything is kind of a missing layer in agentic design systems. And most like some of the early adopters in the In2Design system, that community and all that stuff, they are doing great work of like, you know, exploring the space and things like that., but, uh, the adaptation or having a streamlined approach for that right now is currently missing. These efficiencies are not just like, uh, my observation. I've been in the design industry for 10 years and I observed it, but also, um, that's like data around it, right? Product teams, product teams spend 34% of the dev cycle creating the components that already exist. Companies with like mature design systems shift 40% faster and cut the UI defect by 34%. And to be honest, like, at this time in the market or time in the history, I don't have to ex— you know, convince somebody to have a design system. I think that we have passed that, and most growth stage startups and enterprises have adapted to it. Maybe startups, maybe not, but as the company starts to scale, the adaptation has increased, and that's where most design system teams are like positioned. This, uh, so what is actually is this agentic design system, right? So I went to Intu Design System. This Intu Design System is one of the most reputed platform for design systems work. So this is their definition of it: an agentic design system is an infrastructure that lets AI agent autonomously read, uh, reason over and build your— and build over your components, tokens, and guardrails. Unlike traditional, um, design systems for humans, it encodes intent relationship constraints as machine-readable context. So agents can observe, detect, suggest, fix inside the self-healing loop with the human oversight. So this is a self-healing loop, a loop is a very interesting— that's where the proactive agents comes into picture. And the second layer, there are two components to us, so there are two components to it, right? The first component is taking an existing design system and making it agent-ready. And most of the design systems have that big gap right now. And then the second layer is making it, you know, self-healing and things like that. For example, if you white-code something and the component is not matching the design system, right now there's like a lot of like governance and everything that goes into pictures. This can be automated. Agents can like automatically fix, observe, and do these changes., and constantly monitor through cron jobs and stuff like that. Those concepts do exist right now, but it's not really completely applied to the design systems operations and design system maintenance and design system building and all that stuff. Gap also is like, for the— over the years, right, design system was only built for humans, but now agents are also becoming the consumers of this agent, uh, of this design systems. So design system is almost going from from what we have traditionally to becoming an infrastructure layer to agents and also the entire teams, right? So if anybody is like white coding and creating prototypes, they're creating like generic UI, which is not consistent or aligned to the brand, or if the design system changes, it's not adapting. So there's this missing layer growing over the time so the agents can auto-monitor, fix it, and stuff like that. So you don't really have to spend a lot of time on the governance and manual and repetitive work and things like that. So even if sometimes, I mean, I, when I do cohorts and things like that, cohorts or helping with the clients and stuff, the agents naturally drift, right? When you prompt and wipe code, it's not aligned to the brand. But now you can like link your design system and you can do lots of like prompt engineering circles and kind of like fix it, right? And that's how we're doing it. But that's not really the right approach, and that's backward. I think people are trying to solve it by saying that I'll get good at prompt engineering, or I will, you know, find a method to kind of— I'll find some kind of prompt engineering method to kind of fix it. But that's really a backward approach with like with agents getting so proactive ability to take actions and do things in time is becoming a potential path for agentic design system. Just making your design system agent agent-ready is one step. That's the core thing. First, you need to set up that, right, and having a clear documentation and things like that, that agent-readable and stuff. But, um, that's only the one component to it. The entire— this entire harness design that goes behind it is also the missing layer right now. So harness engineering is one of the growing topic, and I've made a video on that as well. But that harness design and creating that agent system around that documentation and design system is the missing layer right now.. And when you combine this agent-ready design system along with this hardness, hardness design, this is where— this is what it actually makes it an agentic design system. This thing is not a theory or a future prediction right now. Most companies are already adapting to it. Like Spotify is doing some of the great work within— companies are already adapting to this. And doing this. But I think there's definitely higher adaptation for making a design system agent-ready. But I think that harness and the AI system around it is kind of a missing gap, which I believe eventually we'll all pick up. And I— some of the work that I'm doing at Human AI Studio, I'll also share this through my videos and my Substack. And my core thesis is like, design should— my core focus here to design should not just exist in a layer where you are just cleaning up the UI slop, but becoming the core infrastructure to the product team and the AI systems that builds the products, right? Interfaces are also changing now. Those traditional patterns are going away and the new patterns are coming, and the design system teams need to be able to deliver to that speed too, right? And over the time, there's always that friction between product and design system teams., and product designers and all that frictions because, and always there is this challenges. Uh, some are like very like people challenges and adaptation challenges and, uh, things like that. But I believe like, um, agents can definitely help to support to increase the velocity and speed and kind of create this agentic systems that are self-adapting and healing and becomes a core part for design systems team to able to, you know, ship faster and able to deliver and kind of automate some of the repetitive work which they don't like to do. So I believe, I do believe design, so I do think design system are going to be a core infrastructure to the product AI systems and design teams and things like that. But right now it's been like a little scattered and everywhere and it's very undefined territory right now. But Intu Design System is definitely doing a great job of bringing all this people from the companies together, sharing resources and things like that. But I think it's still an undefined territory yet, but it's definitely gonna be a core part of design roles and companies, how they are building and shipping products, especially when they are scaling and they are in a growth stage, or even like enterprises trying to make efficient agentic design systems and stuff. So that's about agentic design system, and if you want to work one-on-one with me, you reach out to me. All the resources are there in my description, and I'll see you in another video. Thank you for watching.
More from John

Building Orb AI in Public: Episode 6 Accelerator & Creator Focus #creator #orbiai

Jev: The Decision Layer Boosting AI Retrieval from 72% to 95%. #Jev #AIDecisionLayer #GenerativeUI

Design Engineer's Ultimate Tool Library: designengineer.tools #designengineer #designsystems







