19. Upgrade from Vibe Coding to AI-Native Product Design (Metalab's Myles Palmer)

Welcome to the product contact podcast. There is one agency that has become most synonymous with AI-Native product design. They've worked with Windsurf, Mid-Jerney, Suno, The Atlantic, and Uber. Today's guest is Miles Palmer, design director at that agency, Medalab. In this episode, you'll learn what separates Medalab's AI-Native design process from how most teams are building right now. Why feature parity between AI products has become fast and cheap, and what that means for where the real advantage now sits. We talk about how Medalab moves from first meeting to a tangible prototype in a matter of weeks, and why that speed actually makes it work better, not worse. And where design fits into products that are becoming more agentic, proactive, and less dependent on a chat box. As much as the rise of AI products has been a technical pursuit, we are increasingly seeing that the products that actually break through are doing it by redefining the design of the experience. So we're watching this play out right now in the battle between Anthropic and OpenAI, the digital products, Cloud Code and Codex, and also OpenAI's move into physical hardware are all fighting to win on intelligence and design at the same time. So intelligence alums stop being the differentiator once every major lab reached comparable capabilities. These disabling experiments are needed because AI hasn't actually cracked most of the widely used consumer use cases. People actually rely on every day. We've covered this in a recent articles at productimpactpod.com. The technology just hasn't had its iPhone moment yet. That leap that made a normal person's daily life measurably better rather than just technically possible. So Arpe actually found the same gap in his research. He recently studied how 16 to 18 year olds actually use AI for a digital innovation product. And the numbers are worth sitting with. A third of them don't use AI at all. They don't trust it. Another third won't use AI for anything where accuracy actually matters. And if you want to learn about that research or run something similar with your own target audience to spot product market fit issues before they cost you, reach out to Arpe directly. His contact is in the show description. So this is the so-called AI native generation. Employers are counting on these teens to walk into the workforce effortlessly replacing millennials and older Gen Z, who supposedly never got comfortable with AI. But if Arpe's data holds its scale, and we've seen other data around this, the industry is heading toward a real reckoning on sentiment and adoption. Not from the generation everyone assumed would resist it, but from the one everyone assumed would just embrace it without question. And design is the lever that turns AI from a theoretical benefit into something that actually improves the world around the people using him. And that is exactly the conversation Miles has spent his career inside of. And it's why we wanted him on. So we write about this exactly what separates AI products that create lasting value from the ones that just look impressive at productimpact.com. And if your product needs help closing the gap between what it's capable of and what people actually adopt, that's what PH1 research does. So founded in 2012, we worked with Spotify, Microsoft, Mozilla, a range of publicly funded organizations, amongst others, to find exactly where that adoption breaks down. So more info in the show description. All right, let's jump into it. Hi, Miles. Thank you so much for joining us. So before we started, you were just talking about how your team calls you the AI-pilled designer. Do you mind telling us a little bit about what that means and how you would actually describe yourself? Yeah, I think the team maybe get a little bit sick of me talking about AI and talking about all of the silly little things that I get up to. But what I would define it as is somebody that uses AI day to day constantly as a thought partner or as a-- whether it's a coder or a kind of like, I see it as my little companion that does tons of stuff for me. And primarily, that is coding. So I love to make things, whether it's building my own tools or building tools for other people to use, or it's using AI to get rid of busy work, right? Like, I don't want to rename Figma layers or make components myself by hand or anything anymore. So I've pushed everything that I don't really want to do onto AI and anything that I can make a tool for that will enable me to do something better in the future and getting AI to build me one of those so I can use it. So that's probably why I can maybe consider AI build. Could you imagine even four years ago telling your boss, I'm just going to build my own little app, take care of my little job for a million five-code all. But some of the reason we want to speak to you more than anything is that Metal Lab probably touches frontier AI products and brands more than any other business. It's not necessarily that your AI pill is that you quite literally are building the future of so many different industries, whether it's entertainment or SaaS or any platform you can really imagine, news in the case of Atlantic. How do you feel about constantly being thrown into the fire and figuring out something novel every time? I think it's thrown, right? Like for me, it's why I work in an agency. So I've been in house a little bit before and it wasn't for me, right? The pace wasn't quite the-- Whereas working in my lab and working in an agency where every, let's say, 12 weeks, you're thrown a different problem, you're thrown a different curveball, you've got to get in there, work out what's fact from fiction, get to know the personalities that are on a project, and then also produce and create great design out the other side of that. That's what I love doing. So yeah, I'm thrilled to work at Metal Lab and love working with the other S-Ray into clients that we have. Cool. So founders are really excited that AI assisted coding means that they can now ship a prototype in a weekend and be in market in no time at all. There's a lot of founders that are listening here. Can you share any advice with them? Yeah. So firstly, I'm excited too. I think it's a really thrilling time to be a maker, the time that it takes to go from having an idea to having something as somebody's hand is shorter than ever. I think my tips remain similar to what they used to, just on a much quicker loop, right? So don't make something and sit on it for ages and ages without showing it to someone. The longer that it sits only in your mind and you're the only user, the more that it just becomes something that's crafted to you rather than for actual real people. Try not to get overly attached to things, right? Like have a rough idea and some good conviction around it, but make it and put it in people's hands, see what they think, listen, adapt, and balance that input that you get and what you observe, right? Some people sometimes will say one thing, but they'll behave in a totally different way. I think AI is an incredible accelerator for creating a first round of something, but it's not really good to create a beautifully crafted, precise and high fidelity output. Yeah, not at least the level that we want at the L app, that's something that we're always trying to improve and create things that get us closer to those high fidelity dreams. But you've got to recognize as a founder that polished delight and craft can be the difference between somebody opening something once and using it and then using it every single day. There are plenty of products out there in the market, the win because they're simply more accessible and more beautiful than a counterpart. So knowing that design can be your differentiator is really important and then finally, don't undervalue trust privacy and scale, right? We've all seen the horror stories of, you've made something on-lovable and it's exposed to an API token and now all hell is breaking loose. Yeah, don't underestimate that. Something that works for 10 people and something of the works for tens of thousands of people is very different. I have a bit of a different experience with AI, which is that from the minute I started using this generation of LLMs and particularly the vibe coding stuff, I loved it. I really enjoyed it. I thought it was powerful and made me feel powerful and made me feel like I'd do anything. But a year ago, I kind of hit this chasm, kind of pathetic feeling because while I was able to, get some basic things out there, some sites, some landing pages, basic apps, once you go beyond the stuff that you know really well and go into things that you don't know so well, there's a bit of a loss. There is a bit of a sense of I'm not sure how to do it and these platforms make you feel like it'll work. But then ultimately, there's a fundamental flaw whether it's architectural stability or the quality of the data that you're accessing, different APIs or some plan no longer free that it thinks it was. Now, I want to know as someone who's stuck in this cycle of experiments where sometimes I'm scaling up and learning new things, how should I identify when I need to stop and say design? I need design. Design is the core of my problem. - My relationship with AI changes every single day. Sometimes I'm like a caveman, there's just discovered fire and I'm like, oh my god, look what I've done. And then I'll look on Twitter and see somebody do these dates three weeks ago and I'm like, oh, damn. Oh, some days I feel like I'm just bashing my headaches the wall, I can't get anything that I want and it's not going my way. So yeah, I definitely relate to that feeling. - I want to know something that's a bit close to me. So I worked in agencies for a really long time and I still have dear friends that work in agencies and a lot of them are struggling and are being asked to do like, 15 different jobs and one go. But Metalabs seems to still be attracting these, you know, super vanguard AI forward clients. So Arpeid mentioned some, but like mid-journey, sooner, the Atlantic, Windsor, and a bunch more. So while these other agencies are struggling to stay alive and putting way too much pressure on the people working in them, do you feel like Metalabs secret sauce is the approach to design, your approach to product? Do you just have like a crystal ball inside the org that no one knows about? Yeah, I think Metalabs is very fortunate. We definitely don't have a crystal ball. That's for sure. I can say that we don't have that. We do like to say that we have a secret sauce, though. And maybe one day we'll actually release that sauce. We're very fortunate in that this year is our 20th year, right? And for all of those 20 years, we've been focusing on one thing, which is building great products. And we have a really tight set of disciplines, where the less design, research, strategy, brand, whatever it is that come together to all in service of building those great products. And I think the thing that we all know is the quality of the work that we produce is ultimately what defines us. And so there's a relentless kind of culture here to make the project that you are on the best one possible. And to put that workout into the world. And to come up with new ways of doing things or working collaboratively with each other to produce the best design output. But ultimately, I think it comes down to the people, right? Like Metalabs is a company full of incredibly talented individuals that care about what they do, but they also care about each other. And I think that's why I love working here. It's a company full of human beings. They want to look out for each other and make sure that yes, we're doing great work, but we're also doing it in a way that feels sustainable. And so I think that's what's different to many other companies. You know, we are honest with each other. We're honest with our clients. We don't claim to know everything and do all of these skills. But we'll definitely have a point of view. We'll get in there with our clients and we'll act as true partners. And I don't think that's the case for all agencies. Let's be honest, a lot of businesses don't have the money that the Atlantic or a Sunomite. These other brands, maybe they're not so determined. Maybe they don't have the budget. There's so many ways to look at it, but they would love to work with you. What are some of the things that they can do internally? Let's call it as a mindset, steps, approaches that will at least get them in the right direction of learning how to use AI. And avoid the traps of getting too overconfident in the, let's call it the monotonous seamness effect that AI brings to many businesses and projects. What AI has done is it's made feature parity really easy and quick, right? Like to get to this baseline of everybody's got these these same features is as much as Claude will tell you that's three weeks of dev time. It's actually probably like 20 minutes. It's still not very good at estimating its own work. And I think the thing that I always point towards being differentiator is having courage, right? Taking risk and being able to understand when something needs another go around, right? Like when you need to iterate on something again, when it could be pushed and when you could go further and never settle in for something that is just good enough. I've been having a lot of conversations with research and design and product leaders in organizations about, okay, well, yes, you might be really well versed AI or really AI pelt. But how does that then transition within the organization for people that are feeling more hesitant or like how are you sharing that knowledge and sharing that confidence? Is there any advice that you would have to other orcs? Because there's a lot that are just like, I don't know, I've just been told to get it AI and I don't know how to make that actually so that I'm doing the best work I can possibly do. Yeah, it's a constant challenge, right? In the past, it was always difficult share learning between projects, but now it's like exponentially even more right? You're trying to share learnings between projects and share learnings about AI use along the way. What I try to do for me at least personally is share what works and what doesn't work. I think if you only ever share what works, you're creating this false sense of success, right? That this thing is the, it's always going to be great. It's always going to go well for you. So sharing where it hasn't worked for you and acknowledging that is really important. And I also think the more senior leadership that can engage with AI and demonstrate and share with everybody, this is what I've been doing. This is my journey, this is how I've been using it. The better, right? Like creating that culture of it's okay to be wrong and it's okay to experiment and it not go right right from the very top is really important. So at Metalab, we've been encouraged to uh, efforound and find out, right? Like we're not sitting here right now. I was like, Hey, you guys have got token limits. You have to use these kinds of tools. You've got to do XYZ. It's like, look, if you've got ideas, if you want to play with things, go and play around with it. And I think in the future, in 18 months, maybe we'll be looking at consolidation and token limits, but right now, just encouraging people to play and have fun. That's the most important thing I think. I love that. The issue that I'm seeing is that we're hitting a point where too many people are building too many things. On one extreme, I'll have myself, I've built maybe six, 10 things. I don't even remember anymore, honestly. And some of them are useful things and some of them were play things. Some of them maybe get commercialized. But then on the other end, and it's coming from direct experience, speaking to people at conferences, there are situations where there are vulnerabilities coming up in your ecosystem because there's too many agents roaming around. Perhaps it's unprotected keys, access levels, who knows. Obviously, they should be in a sandbox. They probably are. Right. But the point of it is that we are in this moment of abundance. And it's going to be hard to manage the inventory of these things. Now, for me, going back to my stuff, a lot of what I built is just straight garbage when I look back at it. Are you building those three things? Who are they full? Well, one of them is for people like me who are trying to improve my workflows, where I'm just trying to test how far I can push orchestration, what the models are capable of and how to fine-tune them. Some of them could be for agencies. And agencies are really unique use case. And in truth, the process and how it fits into the ecosystem is more important than the power under the hood. I made another one, which was like a tourism platform to help groups plan together and collaborate and figure out what they're going to do and build schedules and vote on things. And it was pulling from tens of thousands of Google API calls. And honestly, it cost me thousands of dollars to actually build the entire system because the amount of data necessary to ingest into it. But then I went on the strip, I tested it out, and I just kept myself constantly preferring to use Google because Google always has access to the most accurate real-time information both in terms of what businesses are doing, their hours, their locations, and obviously where I am and what I'm doing. What would help me recognize what I'm being an idiot and need help from a designer to simplify these systems down? I think why I would say is like, did you learn something along the way of creating those things? And if you did, well, I learned that I need a good designer. That's definitely what I learned. Perfect. In that instance, I think recognizing when you need design goes beyond just like what it looks like, right? It's not necessarily just about what it looks like. It's how does it feel in the hand as you're identified, right? Like is there something that is actually delightful to use? Is this solving a problem? Does this need to exist? Like, can you actually compete with Google? No, probably not. And so therefore was it worthwhile making in the first place, right? And so I think good design leadership, the first thing that it always needs to do is work out, is this serving a need? Does this have a purpose? Is anybody engaged with this? And if the answer is no, well, it doesn't matter how great it looks, population exists. If you have created something, people are engaging with it and they are using it. Leveling up design, I think now is more important than ever, right? We've got to this point where these sea of copycats, everything feels the same. It's either going to be cream colors and serif type faces or it's going to be indigo blue and it's like terror we put together, futuristic. So Miles, you want a non-pain job just really unforgiving, because you can help me with my tourism app? I mean, yeah, I think I have this same problem, right? Like I'm a designer and I do this. I'm on this call with you right now. And the thing that I've got for some of my notes and stuff next to me is a markdown editor that I built myself. And why did I need to do that? I didn't. There's plenty of markdown editors out there. We don't need another notes app. Why? But I did it because I could. And sometimes that's fun and exciting. But what I think is the problem is, you spoke about that false sense of confidence, right? I could have made this markdown editor and be like, oh my god, that's a business. That's something that I put on the internet and I can sell to other people. And the truth is, it isn't. But AIs made me think that it is because it works and it's in my hands. And that's where I'm like, the first thing you always need to do is give it to somebody else because somebody else will always point instant holes in your idea, right? You went on that travel trip and you went, oh, I could just use Google. And I think that's That's the most important thing still is coding whatever you're creating in that. hands of other people and see how they respond to it. - Yeah, and just don't listen to like social gurus, everybody who tell you that you get super tough of everything. - Exactly, yeah. We don't wanna know where your A on R is. Every five seconds, yeah, let's not do that. - Now the dark side of this is that founders think they don't need designers. What do you wanna tell the founders who are listening or the one of you founders who are thinking, "Yeah, I'll hire that designer later. I don't need design right now. I don't need researchers, I'm just gonna build it." - At some point, you're gonna have to pay the debt that you are creating. - Right. - Like you can find a level of moderate success without a designer and you could maybe have a business that will be serviceable to you or whatever. But along the way, you're gonna make a lot of implicit decisions that are gonna become hard to reverse. If you wanna take it to that next stage and you wanna progress, you're gonna need to level up your design, your user experience, whatever aspect your brand, whatever aspect. And at some point, that's gonna kinda come back to bite you. So working with that from the very beginning is always gonna be very beneficial, right? Like rather than trying to retroactively bolt on a chat interface or whatever. I think the other thing is, because AI has made it cheap to get to future parity, we're all being dragged towards the same thing. And so investing early in design defends against, that's the average, right? Like your defending against everything becoming just the same. And I also think products are gonna change. They're gonna become more agentic. We're gonna face a lot of design problems that people haven't solved before. - Your self-admitted AI pills. And I wanna know, how long is it before you can just say to cloud or whatever tool we'll be using in that alien future, make it look better and actually does it. - I think a while, 'cause like what does better mean? - Right, we're gonna have to break down something that can be quite subjective, right? Everybody likes to talk about taste. I hate the idea of taste, the idea that taste is this differentiator drives me crazy, because it's like, well, everybody's taste is different anyway. So I think it's gonna be a while before you can put in very simple prompts to get towards something that is better or higher quality. What I do think is coming is if you have a design system or you have a definition of design that is pretty well mapped out, you will be able to use that to direct AI and create better quality outputs. But the creation and sourcing of that system is still gonna need a designer, right? You're still gonna need somebody to create that framework and give it a unique perspective rather than just have an AI self-direct itself to, quote unquote, good. Design basically. - Yeah, or we're all gonna have to become linguists and semi-attitions. - And I mean, people are trying to get right, that creating skills that describe different kinds of designs and different movements or whatever, but to me, it just feels like a walk around or like forcefully using AI where you don't really need to, right? Like, I think AI is a great first draft. I love it to create a first draft of something, but then I still jump into Figma and design things. I still jump into code and refine the actual styling or something myself, because that's what I good at. - So I wanna ask though, they've introduced the design that ND file, there's tons of tools out there that can scrape the style sheet from sites. There are sites where they've built allegedly the entire instruction set to replicate known products out there. Is design a solve problem? Or is this one of the last big milestones of an LLM proving itself capable of really making some damage in terms of FTEs in product teams? - If everything was a derivative copy of each other, the world would be a very soulless place, right? And so I think if we get to this world where you can just take something off the shelf and replicate it and create it in your own image, the pendulum is gonna swing the other way, right? We're all gonna want something that is more individual, we're all gonna want something that is more expressive, we're all gonna want the opposite of that and you're going to need people to craft and design those things. In a way, this is the way that culture and the world always goes, right? I get swings and extremes always. But I think right now we are marching towards a homogenization, but I don't think if you look at the big successful companies out there that they started from that place, right? If everybody's favorite example is Airbnb. Airbnb didn't start from a place of like, hey, let's look like everybody else and try and make it feel the same so that we get traction. It's a very design led business and I think those companies will still win out. - So our top episode for over a year was when your colleagues, Sarah Vienna joined the podcast and detailed that we need a new relationship with AI. That was about a year and a half ago. Now that AI is being pitched as something that we can design work for free, what should our relationship with AI be? - A deeply personal one, I think is the main thing, right? It's gonna be individual for everybody. So for me, I treat it cautiously as a thought partner and an accelerator, but it's not a replacement. I think what's changed in the last year is, hey, there's not something that's going away and so you're gonna have to learn to live with it and work with it in whatever way what's best for you. I like to use it to enable me to be a maker, to take that busy work away, to format things for me, to do stuff that I couldn't do myself. And I think the big change, or the thing that's more concerning day and day out is how it will change our society, right? We already found it difficult to trust what was around us, and I think that's gonna be challenged more and more every day of what's real and what isn't real. And I think we're also all gonna be more fatigued, right? But it's screen-based experiences, or by, if you're a programmer, the amount of code that you're reviewing now is exponentially more than you ever used to. So, I think our relationship with AI is, it's gonna be deeply personal, it's gonna be little and often, I think sitting in these long sessions with AI and having it in everything that you do in the day around you is overwhelming and is too intense, and I think because of that, people are gonna look for the physical world and look for ways to reconnect with that physical world and to detach in a way. - For sure. Now, imagine MetalHab gets a lot of clients coming in and saying, "Hey, I just need an agent. Can you just throw an agent on there and make my business great again?" - Very few, which is fantastic, because we kinda get the opposite, right? Like most of our briefs are, hey, we have this product or the service of this business. We need to integrate AI into it, but we don't just wanna chat about, right? Like it always pretty much every single brief is, but we don't just want chat, which for designers is way more thrilling. We've gotta work out, okay, how do you actually embed AI throughout inexperience? So thankfully, yeah, we're not getting too much, hey, just ball a chat ball onto it. - For the sake of our audience, please help us understand what exists in AI beyond the chat interface, because it feels like AI and chat are so synonymous that we can't even imagine other paradigms right now. - It does, and I think this is also because AI has been reduced to chat, right? Like I think there are certain things that, you could look at algorithms as being a certain kind of artificial intelligence, right? Like whether that's kind of YouTube algorithm or Twitter algorithm or whatever, but I think in an interface, everything has been reduced to a form box. I think proactive and pre-emptive experiences are what we look at the most. So anticipating what a user might need based on what they're doing somewhere else in a product and proactively serving them interface that can help them solve that problem, whether it is the personalization of software, right? Like everybody uses say a tool like Descript, I thought we're recording all right now. Everybody uses it differently. There's lots of different user personas that might come in and out of that thing that have a different workspace need and based on those people's usage, how can we reorient it around that? So I think whilst we're still in this kind of, hey, everything's reduced to a prompt box, going beyond that is about personalization and tailored inline use of AI, right? Like it's not just bolted on the side, but you can almost use it to take action on anything on the page or on the screen from any moment. - Can you give us some examples because really here in 2026, I think we've moved beyond the radicals and the possibilities and the design materials and all of these conversations that designers are having in 2024 and 2025, I feel like we need to start talking about what agentic design actually looks like. Now, I know that Metal Lab worked on the Uber platform and to me, Uber feels like something that is highly agentic, but we don't recognize it. It makes me question how much of the future of AI design is something that we aren't even aware of and we don't see. - I think transit is a really interesting one. I'm fortunate enough to live in London in the UK and so public transport's pretty good. I think, or at least in my opinion, the ultimate goal of public transport is to design away any interface, right? Like you don't want to be sat in a companion app looking things up. You want it to know what you're doing. You want it to know your daily routine, right? You've got a commute in the morning the same set times day and day out. You want it to be proactively serving notifications or information that materially changes your journey. and so are I think for public transit. getting away from an app, right? Like embedding deeper an OS level would be the most interesting experience possible. Like how can we get rid of planning? I'll get rid of that kind of stuff unless you really need it. So I'm going to challenge you one step further because I actually do work with a public transit organization. I've worked on several projects with them around trip planning and understanding when the investors are coming and most recently about alerts. And the one in particular that I've been working with, they have PIA issues where they need constant consent to be able to communicate. And if consent is removed from one modality, it's removed from everything. So how do we approach it when, because I agree with you, like removing the interface and making it just a seamless communication back and forth, you know this about me is great. But then we have the issue of consent and communication. Have you seen anything where we're able to kind of figure out around that? Ultimately, I don't think we should ever create workarounds for consent, right? Like if somebody doesn't want an application to do something, they should be able to revoke it and therefore not engage with those features. In terms of making it easier, I think these things are going to be offloaded onto the operating systems of the devices that we use more and more and be baked across that iOS, the latest one, like the age verification thing is all just baked into the OS, right? Like it's not on an individual app level anymore. And so therefore I think like this ambient use of AI that I think we're moving towards that people will want to hook into. We're going through that earlier, doctor phase where it feels very uncomfortable. And so coaching people through that, taking them gently through that, like progressing through the level of trust that you give to an AI is important. And I think as we move through that, you'll become more and more common for people to just be like, yeah, okay, sure you can have my location. I think we've seen the same with like privacy and data. That raises an interesting question about personal identifying information. So many AI tools are absolutely useless without context about the user. And while we're vibrating these tools, it's great. We can give it our data. We're comfortable importing things or connecting this or that. But put an application out in production and people will not be willing to share that. The API calls are not available where that ecosystem is unfriendly agents or it's not stable to them. Have you seen examples of what a genetic data permissioning might look like? Because from one perspective, it could be every agent has to constantly specifically ask for that information, which could get frustrating and undermine the entire autonomous eye. Or it's that permissioning contacts personal information is like an infrastructure layer. Not really yet. No. And I think we're going to move to you. In other way, that will manage health data. Like it's pretty good. It's relatively granular. It's on that kind of level. I think that's the system that we're probably going to start to move towards about the attributes of ourselves or kind of data and information. But I don't think we've reached the stage where AI is deeply embedded enough with it are normal day to day for that to feel like a problem for people to go after. Right. The side of work might use the AI is very limited. And I think that's where the majority of society still is. It's going to be interesting to see what Google and Apple tried to do because they're ecosystem owners. And ultimately we trust them because we're using their products. 100%. And I think the interesting part is like, do you trust them more than the government as well? Right? Like if we think about medical records and stuff, like who do you have government infrastructure? And what it looks after that stuff is our kind of like digital twin of all of these things. Who do we want to be the ultimate owner and controller of that? And where does that information ultimately get stored? Well, if you look at the example, 23 in me, the platform that 15 years ago, millions of us sent our genetic makeup to and they stored it. And then they never selling it for pennies on a dollar because they went bankrupt. Clearly, we as consumers are much more willing to give up valuable and critically value information about ourselves. So I just wonder how we can stop this in an agent arrow. I have a tend to agree with that. I think with data and privacy, the boat's kind of gone for some degree as well. Right? Like we are now in a world where so much of our data has been sold off for pennies. Like you say, that it's really hard to like put that back in the bottle basically. I wonder though if it's not even that people want to have ownership over their data because of the privacy perspective of it. But more so, it's like if you're the owner of your digital twin of your data, you then have the right or the ability to move it across different agentic systems. Whereas if it only is owned within like a government system, for example, okay, well, are they using cloud? How do you then control what you do? And using the health data, one, for example, I find it really interesting because I actually don't even know how to access most of my own personal health data. And I have to pay my doctor money to get a digital record, even though I know it's all digitized and it all exists within a digital system. But I have no way to access it, which is kind of bonkers given the state of the world and technology today. But I think that shows like how long some of these things can take to catch up, right? Like it's a bit different here in the UK with your health data is all with, you know, chess and you can access it all very easily. But I think aligning on a global approach to this kind of a thing and having everybody move at the same rate is an impossibility. And so therefore it's going to be. You can design the most frictionless experience and deliver the most value in return. Okay, let's bring in another core use case that hasn't seemed to change much. What about e-commerce? How can agentate capabilities change the way that we shop? In truth, most e-commerce today, it doesn't look that different than five years ago. You might have a smarter chatbot. What might the future look like? I think e-commerce, the most exciting thing I've seen is doji. I'm not sure if you guys have seen the doji app, but you can create your AI likeness. And then you can cure your wardrobe. You can try on different looks. Accuracy of it is pretty nuts, right? And so you can do this social shopping and gain the confidence to buy clothes that I just don't think you can get in our typical virtual world. So that visualization aspect of e-commerce, I think is a big one. Ikea, for example, has always been a very good visualization of furniture in your home and all of that kind of stuff. But giving buyers the confidence to purchase something before they hit the button is a big one. And then second, I think personalization, right? Like how can we get people exactly what they're looking for with human authored verification? So like using AI, like, I don't know. If I'm going to buy a new pair of headphones, I probably do quite a lot of research. I go off and read Reddit threads. I go off and do all of this stuff and spend all of this time. How can I get that serve to me better in a way that I know is verified, but I don't have to spend two hours scrubbing the internet for it. I still think that's pretty untapped. Okay. You're bringing up a really interesting approach. And I just want to make sure I understand it, which is we look at the core use case that traditionally has been the business foundation itself. So econ reshopping AI enables us to look at the individualized jobs we've done. Yeah. It's a tough one. I think for some legacy businesses, they're being forced into it, right? They're seeing what's coming and they're going, oh, we have to make a change. We don't know what that change is. And therefore, can we find some people to help us with that? I think others are maybe a bit slower to recognize that the times are changing. And I think the one of the most effective ways that we've been able to work through large organizations that need to change is showing things rather than telling things. So we like to get into projects as quick as possible. Start making prototypes, right? Like start visualizing things, start bringing all of these things to life and breaking down the problems, not in an abstract, but using researchers and strategists to help us frame those problem spaces with a tangible visual side by side. And I think that then gives everybody in those large organizations the ability to understand, okay, I can see what that might be as well as the evidence and the strategic approach to why we would do it. And I think that has always some agencies have done very deep discovery processes where they go into organ organization and spend six months mapping the whole thing out. A metal lab, that's not something that we do. Typically, we've got a week of kickoff and onboarding where we get to know a client and then the second week we are making stuff and we are like, let's it, we're going to make things, we're going to put it in front of you every single week and we're going to work together to craft things. And so yeah, that deeply embedded partnership, I think, is a great way to operate. Yeah, so that's it. It sounds like a lot of making and refining, just like write it off the bat. Okay, so I have another area that I want to pick your brain on then. So universities are struggling in terms of figuring out their place in this new world, but they are incredibly essential to the success of a society in our ecosystem in general. And they just, it feels like university websites haven't evolved at all. How should they be evolving in this AI world? Yeah, this is a funny one. I have a personal kind of like background in education. So I used to be an associate lecturer for design and things like that. So I have a little bit of opinion on how a university is running in the UK at least. A university website has been irrelevant for as long as I can remember for like 10 plus years, right? Like it's had not really many purposes. I think the primary thing that they should be doing is inspiring people, right? When you're a prospective student working out where you would like to study and why you would like to study somewhere. And what course it is, you want to be inspired and you want to feel engaged and that this is the place for you. And I think many universities of the website at least have become a kind of like car glove compartment, right? It's like they serve nobody and everybody at the same time. You rarely go there and when you don't really know what's in it. So I would split the experience today very virtually. I'd have something that was super inspiring and enable the different departments to express themselves in the way that they would like to. Don't make it super templated. Make it this fun and exciting thing. And then for everything else, I would probably put it into a conversational interface, right? Most people are going to a university website with a question and they probably want to ask that in natural language. And I think for most of those things, you are practically ready to Google search. So I replace everything else into that. And that would be my brutal take on university website. Last one in terms of this exercise. Imagine that anthropic came to you and said, you know what, it's time to revamp Cloud Code. How would you approach that brief? First, I would say, thank you. This is exciting. I think I'm a big cloud user. So I'm very familiar with it. Not once in my time as I used it. Have I been like, I have a problem with this interface. My biggest gripe with Cloud Code is the adoption, the understanding and sharing of information and connections between myself and my colleagues. Right? Like, if I'm somebody that is very familiar, compared to a colleague that isn't a ways of using Cloud Code are just so different. And there's no way for them to learn that other than through brute force. And there's no way for them to get access to the things that I'm doing of them doing like a forward slash skill thing. So I would say actually anthropic. I think the interface is totally fine. But you need to work on how people adopt, understand, share, and use this product and give them the kind of like, the training wheels to get there, right? Because I think everybody having a slightly different set up or a different hook into the computer causes and all problems. Part of my expectation is that within a year, we're going to be model agnostic and the harness is going to be the thing that we purchase. And it is the core of what we run our business on. Do you think we're going to be moving away from Cloud Code and Codex and more into these customized business operating system harnesses? My hunch is yes. And I think it's because-- or at least on an enterprise level, I think so. Because I think many businesses will find themselves wanting to tweak and customize it a way that they just can't get out of the box. I think businesses creating their own internal tooling now is faster and cheaper than ever before. It used to be something that you just didn't do. It was too expensive. It was incredibly hard to maintain. I think now if you have a custom harness that can see across your organization and access all of your information, and it's all pulled in one place, and it's an installer, I think that's going to outweigh the maintenance cost of creating those things. So my bet would be yes. I think we'll see a lot more custom harnesses and businesses that are moving in that direction. For the day-to-day stuff that I'm asking Chachi PT, no. It really makes me worried that there's going to be such a massive divide. Yeah, but I guess the internal creating a custom harness is a way to get around that in some ways. Imagine that instead of you setting up everybody within your business with Cloud and giving them all the skills and do the stuff, they just double click an installer, and it's one thing, and it's done. And it's all there for them. They don't have to go through all of those hoops of MCPing and to figure out whatever it is. Terrific, isn't it? It's a pain. I'm very familiar with the terminal, and I hate it too. So I can see if it was your first exposure to all this stuff, you're just like, why would I do this? What, why? I agree with that. OK, so wrap us up. Which AI products are you most excited about right now? And tell us why and what makes them so special. I'm going to give you a cop-out answer, and I'm very sorry. But there's not one product that I'm crazy about right now. I think what I'm most excited about is that someone can go from having a problem in their day to day and creating a solution for it and making these little micro apps. And I think that is super exciting. We have a channel-- a battle app called Waywo, which is what you're working on. Two people last week posted, hey, I've made this little app. It helps me day to day almost for their partners' business, and they work as an architect and measuring rooms and stuff like that. And the other one was for somebody with a new ball and an algorithm with different aspects of their day to day as a new father. And I think AI enabling people to create those solutions for themselves is really exciting. So yeah, that's what I'm most excited about. And where should people follow along with what you're working on? You can find my opinions always on the internet somewhere. So you can find me on LinkedIn or you can probably follow my Instagram, maybe or something like that. But yet, typically, I'll put stuff out on LinkedIn and come follow me that and see what I have to say. Thank you so much for having me on. Really appreciate it. Thank you for listening to the product impact podcast. If you enjoyed this episode, make sure to follow us and leave a comment to inform the conversation.