Building Trust: Designing Product Ecosystems in the Age of AI

Digital products no longer exist as standalone apps. They live inside complex ecosystems of interfaces, AI systems, legacy infrastructure, and workflows that all have to work together. In this episode of Patterns, Chris Strahl talks with product design leader Andi Rusu about what it takes to design reliable digital experiences in environments where multiple systems—and increasingly AI—are shaping how products behave.

Drawing on experience at Disney, Sonos, Axon, and Microsoft, Andi explains why trust is becoming the central design challenge in modern product development. As AI becomes embedded in digital products, the job of design expands beyond crafting interfaces to shaping how complex systems behave, how decisions are made, and how users understand what’s happening behind the scenes. The conversation explores how designers can balance abstraction and transparency, when friction actually improves the experience, and why human judgment still plays a critical role in building trustworthy AI-powered products.

We’ll explore:

• Why modern digital products behave more like ecosystems than individual apps, and how fragmentation across systems creates new design challenges for product teams
• How AI is becoming a new layer inside product development, influencing how workflows, decisions, and automation shape the user experience
• Why trust becomes harder to maintain in AI-driven products, especially when systems make decisions users cannot see or easily understand
• Why human judgment still matters in AI-powered design, and how designers balance abstraction, transparency, and intentional friction to create reliable user experiences

View the transcript of this episode.

Check out our upcoming events.

If you want to get in touch with the show, ask some questions, or tell us what you think, send us a message over on LinkedIn.

Guest

Andi Rusu is a product design and research leader focused on creating user-centered experiences across complex product ecosystems. He has led design teams and initiatives at Disney, Axon, Sonos, Microsoft, and Deloitte, helping organizations deliver impactful digital products at scale. He has also taught experience design at Cornish College of the Arts, the University of Washington, and the School of Visual Concepts.

Hostt

Chris Strahl is the host of the Patterns podcast and a pioneer in modern digital product design and development. As the co-founder and CEO of Knapsack, he is a leading voice on how AI can fundamentally reshape the way teams design, build, and deliver digital products with a human-centered approach

Sponsor

Sponsored by Knapsack, the design system platform that brings teams together. Learn more at knapsack.cloud.

Transcript

Patterns Podcast – Alex Wilson

[00:00] This podcast is brought to you by Knapsack, the Intelligent Product Engine helping teams design, build, and deliver digital products at the pace of ideas. Knapsack creates a system of record built for both humans and AI, giving product teams the data structure and alignment they need to deliver with speed, scale, and confidence. Learn more at Knapsack.cloud. Hey everyone, welcome to the Patterns podcast. Each episode we sit down with the leaders and builders to find how modern digital products come to life. From systems and tools to culture and decision-making, We dig into what's driving real impact today and shaping the future of how teams build. Hey everyone, welcome to the Patterns podcast. I'm your host, Chris Stroll.

[00:43] Today I'm here with Alex Wilson. Welcome back to the show. It's really great to have you. Last time you were here, we talked a lot about design systems. Today, along with the change over to the Patterns podcast branding, I'm also gonna be talking to you about something we haven't talked a lot about before. And that is real deep practical adoption of AI, especially as it relates to this kind of fractured transition that we're all living through in the age of rising new technology. So for those that didn't listen to last show, why don't you go ahead and introduce yourself and say a little bit about who you are, what you do? Thanks for having me back. Yeah, I'm Alex. I am the senior engineering manager behind the design system at T-row price. I'm also working on AI activation in parts of the firm and

[01:27] working on trying to roll out like different AI strategies and how people can use the tools was available to them and then create new workflows. It's a really exciting time. We've gone pretty far with the design system that we have, and we're continuing to push it forward with more AI capabilities. There's a wealth of knowledge of gained over time in setting all of that up, and I'm happy to talk about it and share some of those points today. I think that this all starts coming from a place of tension. I know that we feel it in AppSack. There's a lot of change happening. I actually talked to one of our principal engineers yesterday, And I said, hey, what's going on? What's in your mind? Just kind of a generalized health check. And he said, I don't do the same job I did two months ago. And I don't think I'm going to do the same job I do now

[02:11] two months from today. And I think that kind of gives this sense of how things feel at the moment, where things are changing so fast. And there's also this mentality of only native AI companies are really the ones that are being successful. And I think this leads to a lot of interesting forces both inside of companies, inside of markets that I wanted to dive into for a moment. Where does reality fall for you on this? When you're thinking about the people you see an industry and in market doing this type of stuff, do you feel that attention? What does that look like for you? Obviously, startups, you've got a smaller group of people, so you can move a bit faster than a larger enterprise.

[02:53] And there's also red tape that accumulates the larger the company gets. In terms of moving quickly, you know, you see a lot of these startups adopting these AI technologies and rapidly bringing them into their existing workflows. And then for larger companies, the challenge looks like how do we bring on the right tools? How do we evaluate tools? Where are we spending our money? How do we best spend the money? Figuring all of that out, making sure that you're also helping train people up at the at the same time without too much of a delay. It's all about encouragement of use, trying to find and raise those different use cases about. So that way people kind of understand what others are doing around the company. It is very different. If you're within one of those large companies, you may be looking at someone

[03:38] that you might be friends with at a smaller company that is already further ahead, but that shouldn't stop you. There's lots of opportunity to do things outside of work that you can then bring that back into work to better your process. like I started setting up an open call for myself. So that way at home, I can have my own AI-agentic system. And that's helped me learn how to set that up. You know, I brought some of that knowledge back to where I work and I'm able to implement it better there. Even with the tools that we have that may not match what I have at home. But yeah, I think it comes down to individual attitude, the challenges of a large company in general that we've seen for years.

[04:18] and just how invested each company is in adopting AI. That's another piece of it too. There's a lot of really interesting forces of play and what you were just talking about, right? There's the traditional risk aversion of enterprise versus the scrappier mentality of startup organizations. Coming from a obvious startup myself, but a startup that works a lot of enterprise companies. I think that tension's really present, but it's present in different ways. For me, for example, when I look at the market, I look at where venture dollars are going, because I compete with a bunch of other startups for venture capital. The vast majority of it now is following to AI native companies. Moreover, you have different metrics that are emerging in startups around

[05:01] how you measure the success of startup company. You look at it not just as like, hey, are you getting three to five X growth? You look at it as, hey, you're getting three to five X growth where you have like three employees instead of what would take normally like 30. So that ARR per headcount is becoming this metric that is really rising. I think there's also this notion across big enterprise companies and more mature startups of it's kind of cold out there if you're a SaaS company. It's a bit of an AI winter for SaaS orgs. We were talking before the show and I was saying, like, look, we're renegotiating the vast majority of our SaaS agreements right now, largely because people are churning left and right at a big SaaS business. And this is probably gonna be the first time in quite a while,

[05:46] where you watch, I think, a pretty big contraction of enterprise SaaS kind of across the board. And I'd be curious what you think on the idea of what a pressure is existing market that are causing that sort of thing to happen and why organizations are suddenly like reevaluating a lot of how they're taking a posture towards SaaS or towards software more generally. This year thing is really important here. There's a lot going on this year. There's a lot that's been released. I think a lot of companies, large and small, that have taken their own path and have taken their time, a lot of them are coming to the same place where they're getting more efficient and then suddenly they start asking the questions and that's the first step.

[06:28] The question is, if we can build this, why do we want to pay for that? But that question's always been around. It's build versus buy. But now the difference is that if we're building with an agent capability and we have the ability to maintain it more efficiently than we used to. Then that provides a better path forward to take that on as a company rather than spend money with a SaaS company. Now there's obviously like trade-offs and you may have a robust product that you wanna invest in and you go to the vendor for that cause they provide that specialized skill. But something that is small that someone could create and you can maintain that, it would cost less than a vendor relationship. That's the route that companies are looking at

[07:12] because it's cost savings. That's important to the business. Yeah, and that's one of the things that SaaS is pitched for a really long time, right? It is don't pay the full sticker price, spread that sticker price over the course of a relationship, and SaaS companies rely on that relationship being five to 10 years that you're going to be able to have that same and ideally ever increasing amount of revenue that's coming to the door from these companies you work with. I think what's interesting is that was always pitched as a total cost of ownership savings. Like don't go invest over and over and over again these different tools that you're ultimately going to like not meet market or have to reinvest in or have to maintain. Now with agents, a lot of these solutions can be self healing. They can be self optimizing.

[07:54] You can have a lot of things that happen that aren't just about like, hey, this solution is in particularly hard to build. But also it doesn't break and when it does we can fix it faster than a vendor can and that is really different in terms of a market dinner also with SaaS in general like how long the contracts last is likely to look different you might have a Contract the last year rather than three years because it's just providing a stop gap so that way you can get something built internally I could see that happening But it also changes the way you build software internally. Everything needs to be thought about in a more modular way. Come back to the design systems. You got to think in that way to begin with. But you're not building this software. You got to think, well, this SAS product that we have now

[08:39] could go away at any time. There might be a decision that this just isn't here anymore. So how do we make sure that the architecture of what we're building is scalable and is sustainable even without that product that we could have just brought on for the company. So that's really interesting. I wanna dive into that architecture point for a second because I think this is really emergent right now. How are organizations setting up their AI ecosystems? How are they building these clockwork mechanisms where you have some really big gears, some things that are like big scale systems, like what LLM do I use where? And then down to very small bits of clockwork. Like, hey look, there's this agent

[09:23] that needs to do just this one thing, or skill that does just this one thing. For example, I do lots of email intros between other humans, like a big part of my value is my network. And so I got sick of having to hand right intro emails based on some interesting anecdote or context. So making a cloud skill that basically says, here's my one on one intro skill, is a little tiny piece of clockwork and a much bigger engine about how I do like communications management. And so I think there's these roles that are emergent, that are really interesting, where you have beyond just the breakdown of design code, the traditional stuff we talked about, how everybody kind of gets to be a builder. You also have people building different parts of the machine, different parts of the system.

[10:07] You have some of these people that are focused on the really big gears, and they have a lot of people that are focused on this really small pieces of clockwork. I have a sense of how this is working in Napsec, even though it's really new since the new year. How are you thinking about the way you put people in roles where they're designing these systems? Who do you decide who builds the big gears? Who do you decide who builds the clockwork? And how do you have some people that all they care about is they're looking at a watch? How do they contribute to that process as well? I think some of this is fluid because we're learning and things are changing so quickly. But teams are gonna look different, I think in general, and I'm not focusing on head count, right now.

[10:49] I'm talking about like team structure and the way the teams work. You look at agile processes that have been pushed for years. And it's worked well for many, even if you have your own flavor. Everyone has their own flavor and that's fine. But it all depends on the fact that you're like working on this team where you're updating this board, that you're like doing this longer term planning, shorter term planning with in sprints and saying like nimble on your feet. But now we have the ability to work in a way where you can get so much more work done in a shorter period of time. So how much time should you be spending in those other tools like a GIRA or something like that? It would take you more time to refine the work, to add tasks to the work,

[11:31] and do all of that setup than to get the work done. And so I think there's like this ways of working challenge that's number one. And then there's also how many people do you need focus on one area? And could you, if you have a group of people, have people focused on different reason. That's what you mentioned about the small gears and large gears. And how do you provide a way for them to quickly iterate through things? So that way we can come up with new solutions very quickly while still focusing on the long term plan. I think it comes down to creating these really small prototypes, really quick wins, just to check to see if something works. It's kind of like what we've called a spike for many years.

[12:14] And then out of that comes something that is a slightly more robust to check to see, okay, it just is fit within the overall process, the overall structure, what we have as this system. And then finally, you've got like that full on build step. And that's where you're really putting in all of the effort. And that's like a high level view of what could be. But it's definitely going to be changing quite a bit about how we function as teams. And how large the teams need to be. It could be that each team is down to a couple of developers, a designer, someone who's leading on the product side, and maybe that's what you need for that scope of work. But you have many more of these pods that exist within the space. I really like one of the things you said, and actually it's funny.

[12:57] I mentioned a very similar thing in an all hands yesterday, which I'm considering making a redefined core value of DAPSAC. And that is we prototype before we plan. And I think that it's really interesting how these traditional models of things like product management are breaking down in the face of enabling everybody to basically direct instead of just do. You have this place where suddenly people have this empowerment, like you feel like a magician, like a wizard, like I can go make stuff again. And so there is like almost this compulsive behavior around making stuff that I think is something that everybody is just embracing.

[13:40] We're trying really hard to say, okay, this is now easy and cheap. How do I take easy and cheap and make it useful? And oftentimes the fastest way to take easy and cheap and make it useful is not to go through the full cycle, but to just iterate on that easy, cheap thing enough times that you inevitably land on value. And I'm sure that that is going to get refined over time because it does shift a lot of burden of work, which I think is interesting also as a topic in and of itself, where humans traditionally are high input low output. You take in a lot more information than you synthesize every single day into deliverables. Well, with AI, a human can have a remarkable

[14:22] increase in their output. But everybody's input level isn't rising at the same rate. So you end up dozens upon dozens upon dozens of things to read process and then take and synthesize new deliverables out of. Ultimately, that should lead to more productivity, but there is also this problem statement in there of your burden of work is now different. Instead of being the person that has to spend much time writing the code, now all of a sudden I have 50 pull requests to review. Instead of being the person that is setting the company strategy and forcing everybody else to read it, I now have 200 pages of inputs on company strategy be that I need to somehow synthesize. That's really interesting. And I'm curious how it affects things in your world. Does you deeply understand how technology works

[15:07] and how code flows and stuff like that? What is a modern SCLC or a modern workflow look like? That is much more focused on how people direct than do. Those that have had management opportunities in the past, I've seen them gravitate towards this way of working a bit faster just because they're used to delegating. They're used to saying, like, hey, you know, I can't do this anymore. I don't have the capacity to do it. So I'm going to have to have someone else do it. I remember back when, so I was a developer on the design system and I led the design system and then I couldn't have time for that. So I got out of the code and I had to let somebody else take it on and say, well, hopefully that's right.

[15:49] And then you take one level up and suddenly not in the code at all. But you're still delegating. You're hoping that it's right. But if you think about the same type of process with code, writing code with agents, you're delegating now. You're handing off that responsibility. And you're like, I hope this works based on everything that I've shared with it in the past. All of this scoping information that I've shared with the model, you're looking for that output. I think that everybody now needs to learn how to become a bit of a manager in the way that they work. And yes, also roles are merging a bit. You like design and development and product. You see a product owner creating some app to serve some purpose for them. And that's excellent. Is it going to scale to be like the best technical at possible? Maybe not, but that's where those roles converge.

[16:34] Maybe someone from product is getting a little bit more into development and is able to build out something. But then that's where someone from the technical side can take that solution and really bring it to something that scales well that could be better for the customer. It could be better for them in turn, like whatever that looks like. Right now, the better thing to think about is it roles and role definitions. But how can you partner with the people around you that have the skills that you need to like build what you need to build? It's not about these rigid things anymore. It's just you're working together to solve a problem It is really interesting how you talk about something is very soft skill that we also do a terrible job in corporate America teaching people like how to manage Talk about something that is like a sink or swim sort of activity, right?

[17:16] And also like it's where a lot of Peter principal like hey look this person is not competent at their job but they can manage. That mentality exists more or less universally across corporate America, but it's also one of those things that there is a soft skill to delegation that isn't something that is put at the top of anybody's list, right? But I also agree with you. I mean, my entire job is delegation. As a CEO of a company, I have no time for anything. I can barely eat lunch or read the 15 things that I need to read that morning before I actually start work. And so AI is rocket fuel for me because now all of a sudden I have things that I can delegate this stuff to and it will reliably get it done. Now it might not get it done in the most perfect way.

[17:58] There might be an error rate that is present, but all of these things I find to be generally acceptable and also highly reliable. And that to me is really interesting where if you bring in that Sosco idea of like my super power with using AI isn't that I'm the best system designer, though I am a systems thinker. It's not that I'm the best coder or best at reviewing or understanding a user problem. That's more like the domain of engineers and product people. But I am really good at parting out bite-sized things for others to go act on. And that is something I hadn't really thought of until this conversation. That's something that's a really interesting skill for interacting with AI. What is also interesting to see play out is that those that already do some of that delegation

[18:39] or have some of those soft skills that you mentioned, they're more likely to also talk about outcomes rather than maybe the details of what's going on at the lowest level. And from what I've seen is that those that use AI to listen are really focusing on the details like functions, function names. So if you're focused at that level, then you're not going to gain the same productivity as somebody that's saying, I need you to build this thing. I need you to add this feature. And however it needs to be done, that's the way that it can be done. This is what I need and I think that's the difference also that we see with people outside of a company and outside of even tech There's some research out there that showing that those that don't have technical experience are actually gaining more productivity

[19:23] Because they're asking the right questions compared to somebody that is so low level that are so focused on how it's built That they cannot see that same level of productivity because they're so stuck at that level These soft skills these ways that we think about problems some are excelling greater than others because of how they process that information. There was a lot of people in corporate America that were engaged upon the quality of their output. I think that that's shifting from the quality of the output to the quality of the questions you're asking or the problem you're solving. It becomes not about what is the thing that I'm doing that represents my value the best, but what is the business problem that I'm trying to solve? Those people that can think about that business problem and those people that can think about how those outputs

[20:06] but it's ultimately get used to fundamentally move a needle, those are the people that you find see a lot of success. And I think this might be why this trend that I'm observing in market is happening. I spent a lot of time talking to CCLE level people at big companies. And in that time, one of the things that I've witnessed over the past six months is CPO's building software again. So somebody that hasn't been a software engineer in 20 years or hasn't been a product person in 10. And like of course, there are actually still product people, But they're not sitting there like defining out a backlog or doing a bunch of product visionary statements or user research or whatever, watching those people get excited about building because their ability to interact with that agent to get something that's actually

[20:50] moving the business forward is what excites them. I have a hard time quantifying or even really exactly expressing what it is, but there's a certain group of people that can understand how to unlock value differently than others. It leads to this sort of fractured state of adoption all across companies where, of course, you have some people that are skeptics. And that's fine. Like, you can be a skeptic. And you have some people that are like, well, full-throated embrace. And then you have a bunch of other people that are somewhere in between. And of those, you're sorting into these buckets of people that know how to get the value they're expecting and people that really don't. And I don't exactly know how to identify this. So I can say it, and like you said earlier, it's all emergent. And so who knows?

[21:34] There is something to be taught to people about how you reframe your mentality around what you're trying to get from an outcome. And that is fascinating, because it's like a human behavior, organizational behavior, human problem to how we use AI. And you look at the people that come into work every day, and not everybody has time outside of work to figure out what's next or pay attention to what's going on in the industry or stay up to date on LinkedIn news or what have you. they've got other things they're focused on, they're focused on their family, very reasonable, but then they're coming to work to learn. And that's where we then need to help shape the mindset there and like help them find their way with what's coming out

[22:18] and making sure that they're aware of like the different use cases and how they could work differently and what's benefiting others, some things that I hear from folks are that, yeah, this could work, but I don't see how it could work in this type of complex project. or like this is an older code base. I don't think it could work here. But then if you have like one person, prove out that that can work in that code base. Then that kind of brings down that wall. And then it allows people to see that maybe there is a possibility here of using these AI capabilities to creating more in that repo or refactor that repo. That dissemination of knowledge is a large factor and how well a enterprise can really gain momentum

[23:02] in their organizations and their adoption of AI tooling. If everybody's just continuing to work, the way that they have been, and there's not enough knowledge sharing, things aren't going to change. You'll have your leaders and you'll have your adoption curve, but it's gonna be slow going. Well, I think it's really interesting to bring that up because I think there was a lot of genuine curiosity about these solutions. Curiosity potentially on a level I haven't seen since the internet. But then there's also like this trepidation of how do I get started or how do I take that first step? And it's not that steep of a curve, but it's still a curve. It's a hurdle. It's a barrier. And I'd be kind of curious just some examples from you of where you've seen this stuff in the wild. And I'll start by sharing like one quick personal one and one professional one.

[23:45] So I coach baseball for my kids. And it's really great. We have an ecosystem that, like every little league, you're locked into this sports management ecosystem. I wouldn't say the aim of it. But pretty much every little league runs on it. And it doesn't really have an API and it has kind of like a UI that is pushing you a lot of ads and structure and all this other stuff like that because that's presumably how the company that serves it makes money. But setting things like a baseball lineup for a little league or understanding how people are doing pitching or how they're doing in fieldine is all things that's gated in that app. It was relatively trivial for one of the other other head coaches who is a designer. He doesn't have a lot of technical experience.

[24:29] He doesn't work at a technology enabled business for him to go and build his ultimate literally app. And he built it in a weekend and it's amazing. It sets lineups automatically, it sets all the fielding positions, it visualizes the diamonds. We can do all sorts of stuff in it. And he built this in a weekend. And that is a better app than this entire SaaS platform that exists that I'm sure our little bleak pays hundreds of thousands of dollars for. And that's a fascinating way of saying like, hey look, here's somebody that was like, man, this gotta be a better way for me to solve this personal life problem. And then just built it. And yeah, I mean, he has the advantage of some technical experience. So those hurdles were a little lower for him. But I think that it's getting lower every day. And that's kind of my point with that is

[25:13] people are gonna start incorporating this more regularly into their day-to-day life. The Napzock example is also really interesting. And it's another one of those things that is a barrier thing, right? So one of the things that every business struggles with is like, how do you keep your sales assets up to date? So we have deal rooms, just like most companies that sell down to price, have deal rooms. If you're buying software from us, you get a room that's spun up in that deal room is all your proposals and a welcome video and all the other things that you need to do and we use software for that. But it's woefully inadequate because those deal rooms, some of them are ephemeral, some of them have assets to get out of date, all this other stuff like that. And like it's a combination of, hey, here's this Google slide template or here's this Figma slide template here's this file in that file. It's kind of all over the place. And while it's supposed to be well sorted,

[25:57] it really isn't. So what we started to notice in our use of cloud is that the HTML assets that cloud generates are actually better than most of the stuff that salespeople could design using deck templates. And so we basically said like, okay, we're gonna spend one day at the company, we're gonna eat every single salesperson, a local clone repo. We're gonna build a URL where all of this stuff lives, and we're going to build what's called the Knapsack enablement system. And so Knapsack enable is now basically our sales kit for every single customer. All those assets live in a repo is HTML files and markdown files. They're all inherent brand from a core central brand system. We use our component tree to build them.

[26:40] And what's amazing is I have sales people every day committing to get to build these sales assets. and they're all sorted and organized and exist all in one place. And it's amazing because I have people, literally with a title like business development representative that are cloning repos and making edits, using Claude to create sales artifacts that are richer, more beautiful, more detailed, more personalized than anything we've ever made before. And that's amazing. That is amazing. I love that. My background's in development. I have a ton of passion for the craft of it. Like I can enjoy like putting together some really solid code and there's been a place for that. However, the barrier to actually get something out there is next to zero.

[27:24] There's not really much of a gate keeping from anybody to build anything. It's just what are the ideas that you can come up with and then how well can you execute on those ideas? How far do you want to take that product? One concern that I have is the amount of trash that we're likely to see over the years because so many people create products of their own. I fear that you've got somebody that had real ambition for something. Just to look and get hub projects, just look at open source and make a comparison there. There are so many people that start a project on GitHub, but never finish a project. Or they start a project and then someone comes to the table and say, hey, I love what you did. Can I add some features to that or when are you going to update it? When are you going to add this feature?

[28:08] We're going to add that feature. But the problem is not everybody has the time for that. We're not everybody was planning on that to be a part of it. So they might throw up a website and it served their purpose. But someone who else might see that website and start using that thing. And then something one day it just goes away because the maintainers weren't really maintainers of that thing. They had a reason to build it and it worked well for them. So then the question is how do we deal with that as people? How do we best handle that? It's a larger problem than one company. But at the same time, it's exciting that people can build their own thing. But I do think, and this is the part that I think is hard to come to terms with, is that implementation detail

[28:50] will become less and less of a focus than it ever has been. That's why I've mentioned the craft at the beginning because there's a lot of craft that goes into writing the code. But if you have an agent that's helping you build that, and you have an agent that's helping you maintain, I mentioned the comments of the naming before, why are we focusing on all those details if we can standardize in the architecture of the scalability, making sure that if it's a fun and experienced is it accessible? Like all of these things that matter, we should be focusing on, but how much should we be actually caring about the implementation details versus actually getting that product out? Hey everyone, I'm taking a quick break to tell you about NAFSX pattern summits.

[29:35] If you've never been, these gatherings are for senior product design and engineering leaders navigate in the complexity of modern digital production work. We bring folks together for thoughtful discussion based sessions we can share what's top of mind, learn from peers, and leave feeling renewed. Pattern summing servitation only, and intentionally small, so the conversation's same meaningful. If you'd like to join us, visit naps.doc.clouds-events to request an invitation. You're facing this all over the place. We have a lot of folks that we work with in retail, right? The degradation of the owned experience, or like the devaluing of the owned experience. One of the reasons why Shopify is killing it right now is they're able to take major brands that would have never considered using a platform like that for retail commerce and saying like it's a means to an end

[30:19] it's way faster and way cheaper than me building it ourselves and Shopify has market power and so they can beat up people like credit card processors for better rates with Amazon and with all these other like retail portals that exist now the owned experience has been eroded and chipped away at for more than a decade at this point. And there's a lot of organizations that are saying, why am I building this anymore? Why am I not just using a platform and then using AI to make the things on the platform? And I think that's going to continue. But there is this counterbalance that that I'd be curious to your opinion on. You can talk about this app fragmentation thing. Well, if there's a bunch of stuff that gets built, how are we all going to know what's worthwhile for us

[31:04] to spend attention on and what is it? And it frankly reminds me of the early days the internet before you had like 10 companies that had a stringlehold on most of our digital experiences. Yeah, sure. We all go to maybe a couple hundred different domains a year via an address bar. But the number of those I would say has been shrinking year over year over year, or at least the owners of those URLs have been shrinking year over year over year. And in a future where you're going to have a lot of people that are able to build their own or roll their own or do this thing, what does that look like? Does that break this stranglehold on this handful of software providers? What does it do for things like ads? We're like, it's hard to inject ads into an LLM today or into an agent today. Are we witnessing something that has this opportunity

[31:49] to be more likely early days of the internet where things were a lot more all over the place? But sometimes in a way, it's like, nostalgicly, at least for me, a little bit more fun too. Oh, it is fun. It's all fun. I mean, it's exciting and there's also a lot of change at the same time that we're dealing with. So it was during the beginning of the internet as well. It's a great comparison. I'll give my thoughts on where I think things are headed as far as the web goes. I really think that it will come down to a number of larger companies that have these chat portals like your perplexity, your chat, CPTs. you know, you look at Google and how often do you scroll down to the results anymore versus taking a look at what's on top. You'll have these companies that

[32:33] have these portals and then I think a lot of people will go there for a lot of finding their information, getting that. But that doesn't mean the websites go away. That's the way the brand is represented through anchors of the website. So like your homepage is built out like maybe your product page is built out. You're hiring pages built out and structured in a way that is static, but then things between those areas are all dynamic They're all ultra personalized. They're very focused on the user. You're still using the same branding But now it's all generated on the fly, but those anchor points are what really represent the brand So I think it'll be a composition of both of those things

[33:14] You'll have these places where people initially go up and then they go to the brand when they have something that they trust in in that brand and so So I think we're going to see major disruption in what the web looks like. Like you said, with the early days of the web, it just looks different at the beginning, I think we're going to see a real change over the next several years. You already see inklings of this, right? You have things like fast MCP that are delivering your UI along with actual context. So like, this is just the data. This is the way you should display the data or the way you should house the data along with it. to see a lot of infrastructure wrappers for this stuff. ZO.computer comes to mind, where you have something that is not just Cures

[33:58] Cloud, but here's like the infrastructure for your apps behind it, and for your scheduled tasks, and for all these other different things, they kind of are like, where's the landing point of the stuff you're making? It does kind of call the question in my mind, how much of these companies, there's SaaS products that are embedding agents into their products, what's the survival rate of them, and are those things useful because one of the things that I find now is I hate going to other people's UIs. I've generally felt for a very long time that even really clean UIs, companies like Linear that themeously are like, hey, we have a very minimalist take on UIs. I don't particularly love clicking around Linear. Stuff like Jira and whatnot. It's never really been like, oh man, I can't wait to get into there today and go sort of bunch of issues in the swim lanes. So just give me what I need and do the thing I need you to do. More and more,

[34:43] or one of my evaluative criteria for whether or not I like a piece of software, is how easily it is for my agent to get to it. And I don't really love using the AI that's embedded in products because usually that AI have find to be very limited to the things that exist there. That I feel like I should be able to access with my agent outside of that ecosystem with the right-off tokens. And so I'm kind of curious how you think about this whole like embedded AI, SAS thing. You know, I have my own opinions, but I really want to hear your take on. I think there's a place for both. I hear what you're saying and I agree with you, but I think it's also because of where we sit. We sit in a place where we understand how to make the connection with some of these products, like using MCP servers.

[35:26] There are a lot of people that don't know how to set up something that could even make use of an MCP server to access that type of data. And for those people going into those products and seeing that agent capability, that might be what's best for them. But like for other people that are maybe more technical, you can get much more out of the experience if you were able to have something like an MCP server connection to that vendor product. So I do think that there's a place for both. But really depends on the experience you want to have and what you want to create for yourself and how much value you can get out of that once you create it. To determine like if you actually end up using that feature. I think then when it comes to the companies they have to figure out for themselves, are there users ones that will actually make use of that feature?

[36:08] or are they users that would rather connect to it via some other mechanism? I think they have a place also. My take is slightly different, and then I think that there's gonna be this in a room period where people's habitual behavior is still like, I'm on my healthcare provider website, I need to use the things that are on that healthcare provider website to access things about my healthcare. But I also think that there's a bunch of stuff that is good practice that is maybe not great product yet, which is stuff that is like, I got a bunch of data cleanliness issues in a CRM, right? Like what agent that I never talked to that is operated by my CRM provider goes and cleans up that data for me? That's awesome, right? And that's very, very valuable. And so I think that that's how you take something that is good practice and make it great

[36:49] product is you think what you said, like go back to the user and say, are these users actually using this in finding value? And if not, are there other applications for this where they could get value? It might be a short-term problem because I talk about the fact that there are people that more technical or less technical. But who's to say that there's not something that comes about that makes it really easy for them to connect to MCP servers. And that's totally abstract. They don't even know that it is an MCP server, but suddenly they can get access to this data or to these features in this new way. And if that becomes more and more accessible, then those agent capabilities within the product don't even matter anymore. You may not even go to that site anymore if they can go through this other path. That's kind of like our future, right?

[37:31] Like that's how we see it is that when you think about a design system and you think about what is a design system exists for Is it really supposed to be this site? This is an elegant way to go browse around and experience components or you're really just trying to as a person Consuming that site understand how to build something and if you didn't have to visit the site at all If it just came to you via an MCP server or a clee or an API I wouldn't that just be a faster route to value for you? And you still need somebody that is able there to curate in the background, all the stuff that the Knapsack agent is doing and all the data sources and all the rules for those data sources. But ultimately, the consumption model is really what people want out of things like design systems anyway. But there is this inner point where people still feel like

[38:15] I need to go to mydesignsystem.com. And I don't know, maybe that'll never change, but I kind of do feel like as adoption becomes more normalized of this stuff, you're gonna watch a lot more people say like, are you guys not as important as our ability to access the data and do stuff with that data that's interesting? Yeah, I was thinking the other day, it's like why are we where we are now with the web? Why is it built the way that it is? And it's because we had to solve a problem that we had at a time at that early stage of the web. And it's progressed from there, but does it have to be that way? No. It just had to be that way at the time. That was the best way that we could handle

[38:59] with what capabilities were available at the time. And I really think that we're gonna see a lot of change in what the web experience looks like. Because there's nothing about it that has to stay the same. It's just at that point in time, that's how the model worked and the models likely to change. I think that's a really good way to think about it. It's almost like there's been this accumulated tech debt of 20 or 30 years of being able to operate in one paradigm and now we're changing it. And so a bunch of things have to change right along with it. As you think about managing that change, I kinda wanna take two perspectives and get your take on this. When you think about this as a leader in technology at a big company and you think about it

[39:41] as a junior developer that is looking at their first job, what advice would you give to those two different people if they were coming and asking me like, Alex, what should I do in order to adapt to this new world? A couple of years ago, I think there was a lot of concern. What's going to happen with the junior role? What's going to happen with the junior designers, junior developers, like what is the value proposition of bringing somebody new on? And I think it's becoming more and more clear that the value is in the mindset. The values in the mindset because those people that are coming into the industry, you know, if they're going to college, I've got a professor friend that works at a local university. and he lets all of his students use AI to do their work.

[40:25] But he does require that they reference the model that they used. He doesn't allow them to use AI on the exams. And if the exams are worth enough, then you really can't pass the class until you really know what you're doing. But I say that because this whole time, they've been learning how to work in this brand new way where people that have been in the industry for a while may not be at that point. So you're bringing on this fresh town for their perspective. The new ideas, the ways of working, I think that's very valuable. So for anybody that's looking to enter the industry, I'd really focus in on that and selling that as a skill. If you've got the soft skills and you're able to work

[41:08] with these AI tools in a way that's better than others, then market that. And then for someone that's at a high level of seniority, it's gonna be a very different conversation. It could be about like the overall strategy the company and figuring out like how AI plays a role in that and where it doesn't. It could be about how they best use that technology. I think in general companies are probably being pushed on pretty hard by their investors, whether it's smaller large to figure out what's next when it comes to AI or how they're best using AI. So I think there's going to be some level of pressure there. And I think that there's also pressure internally and externally. You see other companies doing things and you're seeing them do it well and I, oh, well, we You could do that, but you gotta be careful not to jump too soon. You can't wait either.

[41:52] You gotta really be strategic about how you use the technology, where it works best in your org, and not forcing it in a way that actually makes everybody perform in a much worse way, or it'll be less performant. I think that it's easy for someone at a very senior level to just say, hey, let's just do this. But I think that they also have to listen to their organization in all of the different levels to figure out really what's working and what's not. as they try to introduce these tools and new ways of thinking into the process. What are the stakes there though? Because you talked about, everybody is looking around and they're saying, like, what are my competitors doing? What are other companies in the industry doing? And the people that are doing this right, at least in the startup realm,

[42:35] they're the ones that are getting the venture capital, the ones that are getting the gains, to the point that it's in the neighborhood of 6X, the valuations of traditional SaaS businesses. And so if you're thinking about like the idea of AI as a metric for value creation, Whether or not you believe that valuation is inflated, dollars are flowing into those companies between 6 and 8 times faster than the generalized market. And so that means for every dollar I spend in the old way of working, I'm losing out on what amounts to an unbelievable amount of ROI. That is frankly a massive stake, at least in the startup world. And I wonder, does enterprise view that the same way? Are you sitting there saying like, maybe not you specifically?

[43:18] Are there a bunch of people sitting there saying like, hey, look, if I don't get on this train, I legitimately am losing six days to every day that my competitor does this the way that they're doing it. Speaking broadly, I think that any company that has had a robust product that they've already built is facing a real challenge right now because they've got to maintain this thing that they built. But they also have to really, really think their strategy around what that product even looks like. You can't pause the maintenance and the new requests coming in. You have to start on the new thing. Then it comes down to resourcing. And that's when I think companies are saying, okay, well, can we be more efficient? Because then if we're more efficient, maybe we could focus on both or a transition

[44:02] or adding to the older product in a way that we're moving in the right direction. There's lots of ways to slice it. But there's a real problem right now where everything's moving so quickly that companies don't have a quick move from current state to future state that they can make without some really drastic changes, so I think that the way that they work. It's a lot of one way doors between current product and new product. Yeah. So I can only imagine what some of these meetings are looking like when they're talking about the future of each product at various companies. You do see big moves in the industry, like Oracle, Laneoff, 30,000 people, block laying off a bunch of people. And all of that is under the auspices of like, we need to make the hard decision

[44:46] to reframe the company in a new way of working. And you said a moment before the show, like, hey, at least we're all being honest about this now. And I think that it has been that moment in industry. Sometimes in like late 2025, or early 2026, the people were finally willing to stand up and say, like, we need to embrace new ways of working. And that does come with a lot of changes to the overall organization, not just in the form of layoffs, but in the form of ways of working. And that ability to finally have this conversation around like how does this actually change things with the recognition that there's a lot at stake. I think that's a really good thing for our industry. Yeah, and we've also seen people make mistakes though. There's a lot of rehiring that's going on. Going back to that senior leader and how they're thinking about things.

[45:29] There's like strategy side of product strategy, but then there's also like the hiring and maintaining of talent. You don't want to lose talent, but you don't want to spend money. If you don't have to, and it's that balance and creating that balance and figuring out what works for you. But if you make some drastic changes, you might come to regret that I think we're seeing companies do that. I think there'll be some strategies that come out and made more known about the best way to handle this. But I think everyone's trying to find their own way. That's all part of the change. But I do think, like you said, having these really candid conversations about where we are is important. So that way everyone kind of understands the current situation. This is what we're dealing with, roles are changing. And on the personal side for the senior leader,

[46:14] I think that they have to embrace to also be a leader in their organization and show that they're not just telling other people, hey, go do this, but they themselves are embracing that technology. You have to model the behavior. Otherwise, you're just some person telling other people what to do. It doesn't work as well. And also in order to make the right calls, you have to put good people around you. That's always been the case. You also kind of have to know a bit about it yourself. So the learning there is a must. And I know that in technology, we've always said that there's always learning to do. You have to keep learning. You can't fall behind. But I think that that's going to start growing for everybody. Because if more and more people are using these technologies and these new ways, and we see this change all the time,

[46:59] you've got to embrace the change. You've got to embrace the fact that you're going to have to learn something new frequently. And that is inclusive of anybody in these roles that might be using these tools. Now, not just those that were builders before. So what's a personal project that you've kicked off? I've been working on the open claw. I've been working on building that at home, started with an old Windows computer that I had laying around from a number of years ago. It was just good enough to get it working. And I didn't want to spend a ton of money up front. I know everyone bought the Apple Mac minis. Yeah. All my stuff runs on a pi five. So I get okay. All right. All right. I just wanted to try out. I want to see what it was like. But I now and you're really considering getting a machine that can handle the

[47:42] performance. I really am interested in running local models. I think it's easy to pay for models. They're really good ones out there and they're worth paying for. But I do wonder for a lot of the use cases are some of the free models good enough. And could they... get the job done. I'd be interested to see what that shift looks like. Will people continue paying for the models for all of this work? Or will they invest in their own infrastructure and not have to pay for things longer term now? It's the way the pros and cons of like how much you're spending on the infrastructure itself. Yeah, one like unfortunately like three companies control the supply of most of the character right now. So the supply and demand side of things,

[48:26] the whole supply chain for AI sucks. But let's assume that gets solved. I think you have a really interesting point of like, if all of a sudden to be able to run this stuff locally for the straw LLM run out of my house, that's an interesting idea I really thought of. Yeah, I don't want to think about how much I'm paying when I'm building something. Like that's not what I wanna be focusing on. I wanna be focusing on the idea as I wanna be focusing on like what I can create next, or how does this help better my life? These are things that we used to dream about. It's like, oh, if only I had as personal assistant, It could just do these things, these things are now becoming available and it's exciting and it's accessible and you can run it for free if you want to. It might not be as strong as some of the ones that you pay for, but that might be fine.

[49:10] So I think that that's just a fun thing to work on and it also helps you learn along the way. Maybe your parents were different, but my parents used to talk to me about the first time that they would run out of house to go watch Sputnik crossover in space. They get really excited about that and they told me the end about going from that to the moon landing and how those felt like this gigantic quantum leap about how fast the world was changing for them. The distance between jewels, burn and the moon landing is five, six decades, right? But like the distance between us thinking about, man, wouldn't it be nice if I had a robot that could just do all these things for me that I just asked it to do versus that being now is simply a moment where I feel like the world is changing quickly, but it's also

[49:54] It really is, right? Like it was only maybe 20 years ago, we were dreaming this type of stuff. And here it is, it exists today. And it just kind of makes me really excited and interested in where we all take this. It exists for your average user. It's not gate-cubed by large companies that they're the only ones that can do it. It actually reminds me of also is when suddenly you didn't have to own your own infrastructure and anyone could deploy a website. The ease of the ability to do that, it was just like, okay, wow, you know, like anyone could start a business online. Like even if you weren't using some products that shot the buy or something like these, you could do it yourself. You are paying for it, which is fine, but like it broke down that barrier. I feel like that's kind of like how sometimes

[50:39] this feels is that you actually have it in your hands now. You have the control over what you want to do with it. I do think, nearly jidemic concerns about how far do we go with these models? How smart do they need to be? I'm not the one that's working on that. There are smart people that are, but I think that it's something to be mindful of, but I don't think it can stop us from learning how to best use these to our advantage. I mean, the technology exists, and it is a one way door, right? Like, it absolutely is. There's not really going back. I do think it is interesting to me how many times in the past couple of podcast recording talking about AI ethics have come up. Ethics didn't really come up that much

[51:23] when we were the design systems podcast. And so it is interesting that there's a bunch of things that are about ethics now and how far and what should we do? It's at different levels too and it's uncomfortable. It's uncomfortable. That's sometimes the way I could put in like, of introducing something, it's like, okay, this can improve our productivity and what's the impact to this group of people? What's the negative? And that negative can be uncomfortable. And sometimes that level of comfort is what pushes us to say, no, no, we shouldn't do this. And sometimes that comfort level is also what kind of us is to take that risk and say, you know what? This might be worth it. And then it really pays off that line is hard to figure out which side you go on sometimes.

[52:07] Well, in our regulatory frameworks and our structures for all this stuff haven't caught up to the speed. It takes no time to prioritize it still like a year to buy software. I don't know how that's gonna persist, given that if you were buying Nap Sox products a year ago versus today, you're buying something very different in fundamentally different in a lot of ways. It is an interesting thing to bring up. Alex, this has been so much fun. Thank you. I love having conversations like this. I love having conversations with you. Thanks for coming on and being willing to share your wisdom and all that your insights has been great. Sure, and thank you for having me. We'll have you back again soon. I'm sure this has been the patterns podcast. I'm your host, Chris Straw. Have a great day everybody. Hey everyone. to the Patterns podcast. You can reach us on LinkedIn using the link in the show notes.

[52:52] The Patterns podcast is brought to you by Knapsack, intelligent product engine, helping teams design, build and deliver digital products with the pace of ideas. Learn more at Knapsack.cloud.