Designing Beyond the Chatbot

Speaker A: Hello and welcome to the Bowie World show, the longest running web design podcast where we look at user experience, design, conversion, optimization and working in the web. But on this month's show, we're going to be talking to Josh Clark and Veronica Kindred about their, their new book and the role of the designer in the age of AI. My name's Paul Boek and joining me, as always, is Marcus. Hello, Marcus, how you doing?

Speaker B: Hi, Paul. I'm very well. Um, as I was just saying before we got on, I'm not going to say much on this episode, so. Bye.

Speaker A: What's the point of even introducing you then? I mean, how rubbish is that?

Speaker B: I know that that's not true and I'll have to interrupt at every minute.

Speaker A: Well, you've started the show by taking up most of the talking by saying

Speaker B: that's normal, saying anything that's normal.

Speaker A: How does that work? So let's talk about. Well, we'll talk to the people who actually want to talk to me then, which is Veronica and Josh. Nice to, to have you both on. Josh, you've been on before at some point back in probably 1975, something like that, I would have thought.

Speaker C: That's right. Back when. That's right, that's right. Yeah. I think maybe, or maybe 10 years ago when we were both what, I guess 15 or 16 years old.

Speaker A: When we were very young. Yeah, yeah, absolutely. And Veronica, you, you've got the dubious privilege of being on the show for the first time. It's lovely to have you here.

Speaker D: It's so nice to be here. Thank you so much, guys.

Speaker C: We, uh, we also don't plan to talk very much today, so.

Speaker B: Yeah, it's all going to be.

Speaker C: Thank you.

Speaker A: Nothing new.

Speaker C: Thank you. Thank you. Thank you. Thank you for the, the introduction though.

Speaker B: That was good.

Speaker A: It wouldn't be the first time on this show. It's just been me talking for 40 years, 45 minutes, you know, a bit longer sometimes. I do love the sound of my own voice. But anyway, so, um, we're going to be talking about Sentient Design. I love that. I loved this title. It felt very sci fi and slightly, um, dystopian. Like design's going to take over the world, which I'm quite up for really, um, in, in a positive way. But can you give, um, give people a bit of an introduction to the book and how it came about and who you had in mind when you were writing? Sorry, that's three questions in run. I'm rubbish at this.

Speaker D: Interviewing. First of all, you're the first person, I think, to say ever, that you love the titles, so that's an, uh, privilege for you. And I also see, like, Marcus shaking his head about design taking over the world. So I think this is going to be a great conversation.

Speaker B: I've got no problem with design taking over the world. I have got no problem with AI really either. I just do it to wind up. Paul, um, I think this is a fascinating subject. I've read up a little bit before, uh, before we, um, uh, got on the. Got on the show, but the idea of using AI, because I think I've already. I've already started talking when I said I wouldn't. Yeah. Um, the idea of AI being used for this is my interpretation. The right things, uh, the right design jobs, if you like, makes a lot of sense. And I think we're going further than interface design today. Maybe I might. Yeah. So, yeah, anyway, I've already, um. I've already taken over your response, Veronica, so I'll let you talk.

Speaker A: So tell us about the book, Veronica. What is it? What's it cover, and who's it meant to be for?

Speaker D: Right. Uh, Sentient Design is really about using AI as a design material. There's been a lot of talk in the industry about tooling and how to make people's processes faster, more efficient, better, and not really about the new kinds of things we can make because of the technological advancement of AI, uh, or rather LLMs in the past couple of years. Um, and it's really for anyone that touches web design, whether it's designers, developers, product people, anyone who is interested in kind of ushering in this new era of, dare I say, sentient design.

Speaker C: Well, I mean, picking up on that, Veronica and Marcus, on something that you were saying too, it's sort of like using it for the right thing. I think there's like a lot of anxiety and uncertainty right now about what AI does to design and that, and, uh, the kind of current fascination with how do we use it as our. In our tools and about speed and it's all about production. Don't get me wrong, production is important. How we make things is important, but I would argue it's not as important as the product. You know, let's look at product over production, what we can make instead of how we can make. And I'm not saying that how we make things, like I said, is unimportant, but, man, we can make things now, now that weren't possible before. And all that really matters in the end is what we deliver to our customers and what Happens next. What can AI enable as a material, as Veronica was saying, that we couldn't do before? We've been talking for decades about adaptive interfaces and personalization, and we haven't really had the technology to do it. What becomes possible now? Sentient design is the key, the practice. We're really trying to establish a practice of creating intelligent interfaces, which are experiences that are, uh, that have the awareness and the agency to respond to the user in the moment, to make design decisions in the moment based on the rules and guidelines that we, as the designer, provide. And it is exciting. I am excited. Like, in my 30 years of doing this, this is the most excited I've been about the creative opportunity for design. And, uh, I think that a lot of people are missing that opportunity and are rightfully anxious about what does this do to design. If you pivot to what you can do with AI for design, it becomes this remarkable elevating experience instead of replacing it.

Speaker B: Uh, that.

Speaker A: That idea that Veronica mentioned of design material, treating AI as a design material, um, really, really resonated with me when I first heard it, because it made me think back to the very, very early days of the web, where the designers who were moving into the field were basically print designers. So they'd spent decades designing for a particular medium, that of paper and print, and, you know, and all of the rules and conventions that existed around that. And that as we moved into the Web, we were having to adapt to a new interface that at the time, of new material that at the time was very restrictive compared to the. The material we were used to working with before. And then we came out of that, and there was a period of time when where there was a big debate about does a designer need to know code? Right. And the big argument that I always made is, yes, you do, because you need to understand the medium in which you're working, which is exactly the same with anything. If you're a painter, the way that you paint with oils versus watercolors versus, you know, acrylics or anything else is totally different. And so the medium that you're working in is incredibly important. And although admittedly all of this makes us sound a little bit pretentious, I think, um, it is really true that once you understand the nature and the capabilities of the medium you're working in, that's when you really start to unlock its full potential. You discover what it can't do, but you also discover what it can do. And that. That, I think is one of the big problems I'm seeing at the moment. Amongst designers, because, like you say, a lot of people are afraid of AI, which I totally get and I totally understand. I remember feeling the same way as a graphic designer when desktop publishing came along and everybody said that, uh, I'd be out of a job because there was desktop publishing there now, and people could do it themselves. But equally, I think once you start exploring it as a new medium, that's when it gets really exciting. And that seems to be what you've been driving at in the book. Is that a fair assessment?

Speaker D: Yeah, absolutely. And I. I love the metaphor with the painter and the painting. I think that's great. That's exactly how I think of it as well. Um, I think there's not, like, if you want to be a master designer in 2026, there is absolutely no excuse not to use AI. It doesn't mean you have to use it in everything. Rather, there's no excuse not to know how to use AI. It has to be a choice. And you can't make a choice out of ignorance. And I think a lot of the people who are not engaging with the new material, like we're saying, it has to do with fear and anxiety. Really well deserved and well earned fear and anxiety. But that doesn't mean that it's a smart choice to ignore it altogether, I think, especially when, um, you know, there's so many. There's so many jobs at stake. The future of design is at stake. So I think the more that we engage with the material, the more we're able to make choice and decisions, just like you're saying.

Speaker A: I was thinking. I disagree with you, Veronica. You said there that you. You can't make a choice out of ignorance. I have made a very good career out of making choices out of complete ignorance of the subject. Sorry.

Speaker B: But you've learned from those choices, haven't you, Paul? Uh-huh. Yeah.

Speaker A: Yeah.

Speaker B: Learn from failure.

Speaker A: Yes. Yeah, absolutely.

Speaker C: I do think that, like, both. You know, what, Paul, you and Veronica are both saying really is like a common tension at moments where we have a new design paradigm and a new material. You know, it's like the kind of, um, dislike that we're seeing or distrust for the new thing is 100% natural if it threatens what you knew before and if it comes from a place that you distrust, which, I mean, let's be honest, like, the big models and the people who are running on the company and the people behind them, there is some sketchy decisions being made here. Right. So, like, all of that makes sense, too. Yeah. And yet it also can be really powerful when used in the. In the right way. And I think, um, you know, Veronica, you were, um, just getting started, not in your career, but in your life, during the era that Paul was just talking about of this move into the web for design. And a lot of people didn't like it, right? A lot of people said, the web, it's not going to be a big deal. The design medium is stunted. Um, and I think that really, the opportunities to shift how we think and it's like, right, it's not for what we made before that if you try to jam what we do now into that new medium, try to use that same material, you're going to be disappointed. And we see it now, right? It's like, AI is weird. It is different from traditional computing. It's like it won't give you the same answer twice. It's not great with facts, or sometimes it is M. It's hard to know. It has this confidence that it's always right. So it's like, if you try to use it as an answer machine, it's going to be disappointing because it answers things differently. How can we use its weirdness as an asset instead of a liability? Is a big part of the interesting opportunity here. But I will say, as somebody who's been doing this for a while, Paul, you and I came up together at about the sort of the same time in all this. It has been incredibly helpful to write this book with someone new to their career like Veronica, because, you know, Veronica, you may not have the experience that I have. You started your career just a few months after ChatGPT came out. You're three years in. But you also aren't burdened by my experience that you don't, um, take the current best practices as the next generation's best practices. And I couldn't have written this book without you and without that perspective. She's tossing her hair, everybody.

Speaker D: Yeah, I think it is hard, like, this period of transition. Everything is so chaotic. People really want to know what to do, like, what they can do. And so they look to industry leadership, and industry leadership is maybe tired or themselves doesn't know what to do. And so there's this. I think there's a real craving for leaders and for people who know what they're doing. And the truth is nobody knows what they're doing, um, except for us. Read our book. Just kidding.

Speaker C: Well, I mean, it is. You know, I think something that we are really trying to do here is pull together the threads. Like, I think there's there's this ungrounded moment that Veronica is talking about. How can we sort of find um, stable foundations in what is emerging now, similar vocabulary. We've been building these intelligent interfaces in this new era of LLMs since the get go, um, and so have other people, but it's been kind of haphazard. Everyone's been doing their own experiments. The language is slippery around this. Um, design patterns haven't exactly been consolidated and had sort of a common structure. So what sentient design tries to do is one not only sort of inspire and give the literacy that we're talking about about what this new material can do, but what are the patterns, what are the experience archetypes that go way beyond text based chat and text, uh, based agents? What can we do with this in new ways? What do we call that? How do we make decisions for what we should be making given the problem and need we're trying to meet? And what even are those options and decisions? And that's what sentient design is. It is broad and deep. It's, it's, it is a good doorstop, friends. It's over 400 pages.

Speaker A: You know, this is, this is a good, uh, a good time as well. Right? What I mean by that is, you know, if you look back through the history of, of the web, this, this has happened time and time again from responsive design to the release of the iPhone to whatever else, all the way through this, this kind of period of time. There is this kind of big turning point where, where everything suddenly shifts. That is a time of incredible innovation. It's a time of a lot of sharing where we come together and we share and we work stuff out together and inevitably good stuff comes out of it. Yes. Bad stuff comes out too. I'm um, not claiming everything's great and rosy and lovely, but it, but it's actually when I get the most bored is when things are, you know, oh, uh, everything's worked out. Know we know how to design a user interface now for maximum engagement and, and for usability and readability and all of that. Then you get one of these, these curveballs thrown in and that's where it gets interesting. So, you know, as I've said many times on this show, is something I'm, I'm actually very excited about. And it, it's, it's the times when you haven't quite yet found the edge of what this thing is or what it's capable of. So I mean, you mentioned chatbots and you know, thinking beyond ch and, and I do feel that's a great example of this where at the moment, you know, a lot of us are going, oh, AI equals chatbots. Um, and, and chatbots are rubbish as a design element. Well, not always, but in, in you know, many cases. And so, you know, that idea of thinking beyond um, those conversational uh, interfaces and, and exploring other ways is really interesting as well. So I'd be quite thoughts around that about the idea of you know, a conversational interface. When should you be using it? When is it the wrong choice? What are the alternatives out there? Because it's hard to get your head around it sometimes.

Speaker C: No, you're right. I mean there's been this gravity well around chat for 75 years. Right. I mean this goes back to Alan Turing, uh, and his imitation game, the Turing Test where you know, his thought experiment was an intelligent system is one that can fool you through a conversation. And that thought experiment somehow became a design brief that intelligent machines are talking machines. Right. And don't get me wrong, it's like a text dialogue can be really powerful in some ways because the input and the output can be totally wide open. Um, and so there is utility for that. But open ended input puts it all on the user too to be like, describe what you want this thing to be. And friends, we are not a reading and writing society anymore, if we ever were. And I'm not sure that this is like it is tuned for some brains but not for a lot. And walls of text are not going to be a great solution for certain kinds of data. But I think one of the things that's exciting and then I'll kick it over to you Veronica, to talk a little bit more about this. It's like we have four postures of experience patterns that we describe in Sentient Design. These are four ways that the system can position itself relative to the user. Chat is one of those along with tools, agents and copilots. Uh, but we also try to think about what is chat beyond simple dialogue. We think of it as a conversational format that is a turn based exchange with a peer. I do something, the system responds. What are different ways for the system to talk other than text? What if the system talked in ui? What if the system um, engaged in a multiplayer reaction and we start to see a whole bunch of different experience archetypes. But I don't know Veronica, if you want to talk about the four postures, um, yeah, sure.

Speaker D: So Josh mentioned chat and you're talking about our idea that chat is not just conversation but turn based interaction. So we, we see that Happening. Um, one kind of specific chat example that we have is the sculptor. So something where you might take, okay, for example, an image right now, when we're using LLMs to change or manufacture images, it kind of has to redo the whole thing if you want to make any changes. Um, the sculptor would instead change a specific part of it. So you're sculpting out a specific of an artifact that you want to change and that becomes a chat based interaction because there is a back and forth between the user and the system. But it doesn't necessarily have to be completely regenerating or rather generating something new, uh, every turn of conversation, which is how LLMs work and is kind of how we think of chat right now. Um, the other postures that Josh is talking about, we have tools which are kind of the most controlled and precise. So rather than a back and forth, that's more give an input and get an output. Um, Shazam is like a really fun example of a tool, you know, old school, but it's there. You, you know, you have the system listens to a song and then it gives you the name of the song. Just super simple, just like that. And that's a fun one. And then agents, which are kind of, of a very hot topic right now and talked about a lot. And right now, marketing wise, everything is just called an agent. But for, but for us, agents are autonomous and proactive tasks that are happening that can be behind the scenes. Um, and then the last posture Josh mentioned was co pilots. And for us, co pilots means, uh, quiet and continuous collaborator that's kind of working on par with the user.

Speaker C: Yeah, always sort of behind the scenes. You know, it's like you use tools, you talk to, chat, you delegate to agents and you are supported by co pilots in the background.

Speaker A: Oh, I see. So you copilots, the things that are working in the background. Or is that agents? I'm slightly confused by the two. Sorry to be slow.

Speaker C: Yeah, no, I think that, uh, one thing that we sort of like think about for copilots is that they're kind of always listening and looking for an opportunity to help. So a long legacy and old school spell. Ch. Right, like here's something that is quietly watching. It doesn't interrupt you, it doesn't take over, you don't have to manage it. Just a little red squiggle is showing up for me to take action if I want. And so there's like another example of that is Google, uh, just came out with their sort of magic cursor. It's often called their AI pointer.

Speaker A: Yes.

Speaker C: The cursor is aware of what the context is of what you're pointing at. And when you want, you could say, take this recipe and add its ingredients to my shopping list. It knows what you're talking about in terms of which recipe. It's like, oh, I'm hovering over these ingredients. And you can put your cursor and move those over here. So it's this thing that's like this continuous intelligence that's just aware of the context. When you name the intent, it gets it.

Speaker A: That's interesting. What about, um, the kind of invisible tasks that are just where AI just gets on in the background? Because that's an area that. That particularly excites me. It makes me think of Christian, um, I can't remember his second name. It's a really memorable name, and it's just gone out of my head. Wrote book. The best interface is no interface.

Speaker C: All right. Um, Gordon Krishna.

Speaker A: That's it. Yeah. I mean, how can you forget that name? It's such a wonderful name. And that book, he's also, I just

Speaker C: want to say, a wonderful human being.

Speaker A: Oh, he is. He's such a nice chap.

Speaker D: Yeah.

Speaker A: Yeah.

Speaker C: Terrific.

Speaker A: Yeah, terrific. So that, um, that book talks a lot about the idea of having invisible interfaces and things, you know, where. Where you've got, um, you know, the things happening in the background and design doesn't just equate to a user interface. And that really resonated with me. And it feels like a. Opening up that. Huge possibilities in that area. But I'm guessing that is that mainly agents from your point of view, or is it also the. Cool.

Speaker C: Go ahead, Veronica.

Speaker D: Yeah, I think that just, um, to expand on what Josh was saying, while co pilots are kind of always there and in the back, agents are more at your beck and call. So, like, if there's a specific task that you might need help, but that's when you might call them in, they don't necessarily need an interface like you're talking about, but they're still more, um. There's still more something you're managing than a copilot.

Speaker C: Yeah, there's something. And I think that does bring up this thing, right? It's like agents, you give them a goal, they figure out a plan, they execute it, they decide when they're done, and they come back. And that could be in seconds or days, and you don't have to think about it anymore. I think that there is this assumption. We see a lot of loose talk about how agents could mean the end of ui, hey, we don't need an interface for this anymore. And yet we also see the early returns of people managing many agents as having this. Did you see that study about AI brain fry? This idea?

Speaker A: Marcus was talking about it in the last show.

Speaker C: Yes, right. And that thing where you just, you're getting a lot done but you're exhausted and you can't put a few sentences together after doing this. One of the things that we're seeing, we talk a lot about in the book around agents is this move from the user experience to the manager experience because we not need the UI for that task anymore. But there are these seams, these phases of delegation that are similar to what you have to do as a manager of people as well. It's just like establish the mission, that's just good leadership. Review the plan, uh, oversee the work, redirect or stop the work if necessary, and then review the results. For simple tasks, you might not need to have an interface for all of those rote repeating tasks that resemble more of an automation. But for things that are new or complex, we need interfaces for each of those phases. The work of design changes yet again, uh, along with the work of the user. I think that transition from, I mean, it's just like kind of your own career, one's own career of moving from, um, independent contributor to manager is a hard transition. How do we help our users make that transition with their agents?

Speaker A: Yeah, we were talking about exactly that in the last, in the last podcast actually, from was it doers to directors, I think I called it in the end. Um, so, yeah, that very much resonates me. And there's also humans are still humans as well and we carry with ourselves certain limitations and we don't change as fast as our technology changes. So things about, well, maybe we're not going to need a user interface. Um, not only is that flawed because of the fact that different information is best consumed, some information is best consumed visually rather than text based or whatever else. But then as well, well, I think there's the aspect of human nature and that sense of control and wanting sense of control over the process. One of the problems I see with agents a lot of the moment is that they go away and confidently say, I'm going to go away and sort this task for you. But you don't really get feedback about their progress in that task or what it is that they're doing or the decision points. Either that or you get the opposite where it tells you absolutely everything it's doing as it goes along and absolutely overwhelms you with all of this information. And working, working out that feedback process, I think is going to be an incredibly important part of the process that I don't feel we've quite nailed yet. So that's something else that excites me quite a lot. One of the things that kind of relates to that is the relationship of trust between us and AIs. Um, and if we're going to be working with AIs, we need to be confident in them. Um, so. And, and I think to some degree, that's going to be a generational change. Um, I can't see Marcus ever trusting an AI to book his holiday for him. While maybe there is a day where Marcus's kids would allow, um, AI to book a holiday for them. But setting that aside, there is this kind of fickle nature to AI, and in many ways it's like a human.

Speaker D: Right.

Speaker A: You know, everybody likes to think of AI. AI is a computer, therefore AI shouldn't make mistakes, right? Well, that's not the way AI works. People make mistakes. People claim all the time that they know about stuff that they don't. I can attest to that. I do it all the time. Um, so how do you design around that? How do you design around that kind of worry and insecurity and expectations around AI, you know, because that's a tricky one.

Speaker D: Absolutely, yeah. So in sentient design, we have the practice of defensive design, which is a huge part of sentient design, and kind of like how we work and how we think about incorporating these fickle creatures into our systems. Uh, as you were saying, LLMs are inherently probabilistic systems, right? Like they, they are made by putting together the statistically most likely string of words. And that's the result that we see, and that's the result that we trust and yet we so badly want them to do deterministic tasks. Like, a couple years ago, there was that meme that was going around on LinkedIn. Um, whereas, sorry, it's just like, crazy that memes are on LinkedIn now. But there's that meme that was going around on LinkedIn that was like someone asked, uh, ChatGPT how many Rs are in the word strawberry? And it said two. And the answer is three. Right? And everyone was like, it's, uh, wrong. It's wrong about something so simple. This is a stupid system and it's not right. Like, it's a very smart system, but we're asking it to do deterministic things when that is not at its core, what it's made for. So when we're incorporating this into systems that need to be reliable, if you're in a finance, uh, industry, healthcare industry, anything that is really important that the answer be right and that people be able to trust what's in front of them. Them, you know, how do we incorporate that into those systems? And I think the answer comes a lot in tone. You were talking about how people are wrong all the time, but. And we just accept that. The truth is if someone says something to me where, okay, say like, you have a dog that's covered in blue paint or something, and like the person. And a person says to you, like, this dog is blue, you understand what they're saying. And you can also, you have a lot of, of ways to figure out what they mean. Their language, their body language, their tone, maybe their phrase, their slang. Like, there's so many ways we're like, reading someone beyond the words that they're actually saying. And when it comes to AI systems, we really just get an answer and a disclaimer, which is totally the wrong way to go about it. Like, if you're returning into that the dog is blue, and then you have a little disclaimer that's like, yeah, things may not be factual. You better double check. Like, that doesn't mean anything for people. It also doesn't mean anything for people if you give confidence scores. Um, Netflix used to do this where it'd be like 73% chance you like this movie. And it's like, I don't, I don't know what that means, you know, and so if you're like, this dog is blue, there's a 10% chance this dog is actually blue. Like, that still doesn't really mean anything, you know, but. But if you're communicating in more ways through people, through design, then people can actually understand what they should or shouldn't be taking. If you're like, this dog is probably not actually blue, you understand what that means far more than like, this dog is blue. 10% chance that's true. Um, and there's also different ways through ui, you can express these things. So we have a thing called spaghetti scenarios, which comes from, um, weather mapping. You know, when whether people were. Are mapping hurricanes, they overlay all the probabilities on top of each other so they can see where the most likely paths actually are. Similarly, in ui, you can put a bunch of possibilities next to each other. And we see this in Google Search. So if you're asking a question like, um, are dogs good Pets. You used to get a really different answer if you ask are dogs good pets? Versus Our are dogs bad pets? Because you're obviously like looking for a particular answer with your question.

Speaker A: Yeah.

Speaker D: And now you see um, the little box, what's it called, Josh?

Speaker C: But yeah, like the, the feedback, the feedback summary.

Speaker D: Right. And you also have this design element where it's showing you similar questions and they're like little drop downs. So it's like if you ask are good our dogs good pets? You'll also see suggestions from Google that are like, are dogs bad pets? And you can see the answer to that question.

Speaker C: So it's showing you adjacent, adjacent paths. Right. Like adjacent paths is what is what.

Speaker A: That sounds good. That sounds better than um, the silly little box thing that we can remember the name of.

Speaker D: Yeah, that's it. That's why we need each other. Right. Adjacent path. So you're seeing not only what you ask, but also kind of similar questions and you can see how the answers differ from each other. So there are all these ways to, to show and to indicate confidence and build trust between users and systems. And almost none of it is by just being like the system's right. This system's totally right. Just trust.

Speaker C: It's this idea of presenting information as signals instead of facts and thinking about what is kind of the productive humility of the interface of this is an answer, it's not the answer. And that's in particular as we're thinking about it for fact based results. And I think one thing, and I think that's often what we've relied on competing for. It's sort of like the mainstream idea of like, they give me answers, they process a thing and give me a result, they give me an answer. And uh, I think like Veronica was suggesting, not every answer is of the same specificity. You know, there are some things where it's like there are a million different ways that you could write that sentence. And so there are a million good enough correct answers. But you know, if you want to know the minimum temperature to cook chicken, there's really only one answer. Please don't give me the wrong one. Right. Um, but so I, I think the, the as we think about the grain of this, um, as we were talking about earlier, I think, I think one of the elements that is really useful here is how fluent and intent LLMs can be. That used to be really hard, right. To understand what the user meant. These really, even just the first generation of Alexa, if you didn't know the exact incantation, if you said Turn on the lights instead of switch on the lights, you're going to stay in the dark. M LLMs, take that away. They can understand context not just from the prompt, but from a bunch of other inputs that you give it and to sort of understand what you want to do. But on the other side they're also expert at manner. We see that in the way that they can change tone or expertise, but also in the formats that they can deliver. Oh, wait a second, you want this PDF as a podcast format? No problem. You want this in JSON. Okay, so format. The way that they can return is super flexible. They're not great at facts, right? I mean that's something that, that uh, they are interested in continuing the conversation more than giving the right answer. But they do play well with others. So given tools like rag or MCPs or skills, they can talk to other systems that know what they're about. And so what all this means is like, uh, wait a second, they aren't as I've said a few times, they aren't answer machines, but they're excellent MCs of the experience. You know, the presentation layer is actually where they excel. How can we have them sort of understand what I want, return the result and the format that is most useful and talk to the systems that know what they're up to.

Speaker A: That's a degree. I kind of just. Lars, I know what you're driving at and yes, I agree but I do think where they tend to be quite weak at the moment and uh, it is only training. It's not the inherent problem with the AI, it's more how we're using it. But like Veronica is saying, they tend to be quite poor at uh, presenting back that information what that they've learned in a. To add the nuances of human communication in there. That would be helpful. So they will tend to say things like, um, the dog is blue. Um, while a human being would likely say, well from what I can gather, the dog is blue. Or they might say the consensus seems to be that the dog is blue. Or I think maybe the dog is blue. Um, but even that, that is a really. I understand why that hasn't been done as much because even that is very, very tricky. So just take the difference in communication between the US and the UK in that situation. In the UK we are much more likely to add doubt in there. Right? Or we're not sure. We think maybe this could be possibly, don't know. You better check it out for yourself. You know, we're much more hesitant in the way that we speak. So all of that needs to be taken into account as well. So part of the weird role of us as designers is almost a linguistic and a localization role as well in terms of how humans communicate and how um, they express different information in different contexts. And I don't feel that enough has been done in that field.

Speaker C: Yeah, right. And to your point, it's like these all speak like California Silicon Valley. Yeah, right. It's like, it's a vibe of it, but it's like I think what doing we're getting at all of us here is that presentation is a design challenge and it matters what the user is up to. So the raw capabilities of these tools for the flexibility that they have is remarkable out of the box. But the design challenge is how do we channel this for the context, either the user's specific context in the moment, but also the domain. I am not a believer that Claude or ChatGPT will swallow all software because I think that the design work is really needed to shape behavior around tasks and domains. I think we are going to continue to have a proliferation of software with intelligence embedded in the interface rather than a single do it all Oracle experience, Oracle small o, not big o 90s. But the um, although he's trying, he's coming back, back. We'll see. Um, we, um.

Speaker A: It's almost like the, the large language models, you know, the Claude and the Gemini and the, the Chat GPT, those are your, your operating system. Then on top of that you're building effectively your software applications that are really where all the, the, the, the kind of domain specific knowledge and communication and design elements sit on top of that and a, uh, an underlying L to be focused enough or trained enough in those specific domains to be able to do the job effectively, as far as I see it?

Speaker C: Yeah, well, and I mean like, like any good creative direction. Right. Constraints are helpful.

Speaker A: Yeah.

Speaker C: So as the designer you take on this more kind of creative director role of saying like, here's the language to use, here are the design components that you use, here's the design system that you may use and how to use it. And so we're able to, instead of having this wide open, ask me anything experience, which is really useful in some context, but not when you're trying to get a specific task done. How do we have AI support with essentially making design decisions in the moment? You as a designer can't be there for every second, for every user. What if you had a capable uh, designer making decisions in the moment to give you what you need based on the rules of the system and what you've asked for. That starts to then be like, oh, we've got that flexibility of input and output and of making friends with other systems, but now with constraints and guidance and specific rules, and you start to see the shape of that. Designers using design MD files with agents. But that is a clumsy engineering close to the metal system. That is great for certain types of people who like to make their own tools. But as the designer, providing solutions to people who don't like to make their own tools, most people. What a great opportunity and a collaborator that AI can be in that context.

Speaker B: One thing that I'm, um. And maybe it is because I'm a bit of a Luddite, but I'm feeling this is all a bit theoretical and a bit abstract. And you started off earlier on, uh, Josh, saying that you've seen things that make you genuinely excited and I wonder what they are. Can you tell me a bit more in sort of in reality of what's happening out there that I might get excited about?

Speaker D: Yeah.

Speaker C: Um, Veronica, you have some examples you want to share? I've got a bunch too.

Speaker D: Okay. Yes, we have seen things that we're excited about. Um, one thing that I thought was cool and is not super new now is Salesforce's generative canvas. So that's Salesforce, the sales platform, built a generative canvas which uses the Salesforce design system and then assembles components on the fly, such that the system is bringing to the user relevant things according to the user's context. So the system can see the user's calendar and says, oh, I see you have a meeting, uh, with Client X coming up. Let me go ahead and bring to you the link to the meeting, as well as notes from the last meeting, as well as, uh, all of your notes about this client. So the system is aware of what's going on in the user's life and is taking it upon itself to bring forward that information. I think what's great about this is, uh, the intelligent part is the assembly, but all of the actual content is not made up at all. It's all referencing real things that were touched by a person and that have been vetted by a person. So it's not making up information, but it's making decisions about what's important in that moment for that user.

Speaker C: We call this the bespoke UI experience pattern. We have 14 different experience patterns that are sort of archetypes to follow, and one of them is like a symbol of an interface on the fly Based on a small design system. Um, we've also got things that are uh, what we call NPC agents, non player characters, like from games where in a multiplayer environment the agents participate in the interface itself. So an example of this is um, in uh, Miro they have these things called sidekicks, where the agent comes in driving its own little, little cursor to do some things in the canvas itself based on your request. And so it's this thing that is participating, not off to the side in a chatbot, in a chat sidebar, uh, or in some other application, but it's woven into the interaction paradigm of the system itself. There's this uh, great plugin for um, Google Docs called Pointer AI, where AI participates as an editor in your document document to make suggestions and comments in the side of the text like another user, instead of pasting the whole thing or giving Claude or ChatGPT or Gemini a link to your Google Doc and getting all of the suggestions over here. It's something that is like, oh, we have a pattern for collaboration already. So there's this big just shift of being like, what if AI is a collaborator? That's sort of that image NPC pattern. How do we bring them into the interface?

Speaker A: I'll tell you one example, Marcus, that uh, I did recently. You know, I do a lot of conversion rate optimization work and I was hired by some company that produced landing pages at a rate of knots for every. Every campaign had a landing page, which is good, good best practice. But unfortunately none of the people creating these landing pages knew anything about creating landing pages. And they get a conversion rate of something like 0.4%. It was terrible, durable. So they wanted me to go in and create a template for these landing pages, which really felt like quite a weak starting point when all of these different landing pages have many different audiences, were focusing on many m, many different subjects. So instead what I did is I built out a design system of components. But alongside each of those components I provided very detailed advice about when this component should be used, where on the page it should appear, what elements should be in it, et cetera, et cetera. I also provided documentation about how the content should be written. You know, in terms of best practice for writing for the web, readability, accessibility, all of that kind of stuff. Also provided documentation for um, uh, the, the layout of the page, whether it's top of funnel or bottom of funnel, all of that kind of stuff. Stuff not with the intention of any of them ever reading it, but simply as resources for the AI so that the they could go along and say, I need to create a landing page for this audience on this subject, covering these, these talking points. And it had enough information to be able to do that from a framework without the need for a designer to um, come into the process. Now of course, that now freaks out every designer in the room going, well, there's my job gone. Right. But in actual fact, what, what that's doing is now opening up the designer to work on much more strategic work. So for example, they did have one loan designer which was nowhere near enough, and that designer now is in a position to do more generic user research around their product suite and that kind of stuff and feed that into the documentation, which empowers the AI to be creating even better landing pages based on the data and everything else. So you get into these really good virtuous cycles if it's set up well.

Speaker C: Yeah, I mean, I would argue that design systems should have been doing that all along.

Speaker A: Yeah, they should.

Speaker C: A lot of people settle for like little micro building blocks of like, here's your accordion and here's your input field and like, uh, whatever. I m mean, it's like so much of that is generic. So much of that is like, why are individual organizations doing that when it's really about, oh, here is the recipe of ingredients that solve a problem.

Speaker B: Ah.

Speaker C: And AI thrives with those kinds of, with that guidance. But also so do people. And so I think that there's this wake up call, I would say, to the design system community to up your game and connect it more directly with outcomes than with little building blocks. I think the last decade of design, the innovation has been happening in operations and process and in design systems, which was important as large companies brought design in house. It used to all be done by agencies. All right, how do we design at scale? But two things I think we had to sort of unintended effect of that of turning the design function into a production process of assembly, of just pull these things together. That wasn't the intent of design systems. It was like, hey, we've got a library of solved problems that will help you use them if you need them. And it turned into sort of a dreary compliance industry of you must use these and reduced the effect of design. I would say design systems can help to change that around by kind of connecting it to outcomes, like I said. And I think we've also got this moment where design has the opportunity to shift innovation into product.

Speaker A: What you mean by that?

Speaker C: We're rusty? Well, I just mean that we've Been focused so much on process and production. And like you were saying earlier, Paul, a lot of settled best practices. And so we've been consolidating and best practices, uh, inventing and exploring new things, but we have new interaction paradigms, including, you know, sort of like these. These 14 experience patterns that we mentioned, of which the book Marcus has is thick with examples. So this isn't all sort of like, coming soon. It's not just sort of like, oh, theoretically we could do this. It is a. Here's. A lot of people have been experimenting and exploring with this, but it has not been brought together or shared in a common space, which is what our effort is like. Oh, hang on a second. This may feel very chaotic, but there are patterns emerging. There's a foundation of a new practice here, and that's what we're trying to share, is to give everybody that grounding for what we might, might do individually and as a craft and industry.

Speaker A: So, I mean, where does this leave designers? Um, we've already talked about a kind of shift from doer to director, which is fine for old curmudgeons like us. Josh, sorry, I've just lumped you in with me, which is probably unfair, but, um, because, you know, we made that transition a long time ago. Um, but for many designers, and we were talking about this earlier on the show as well, that for many designers, the joy of their job comes from the doing, not just the outcome. Right. And if you're fundamentally changing the doing and how the doing happens, you're changing the nature of the job for many people. And not everybody wants to end up being a manager. Is that an inevitable, um, situation? And, and I, personally, I think it probably is for better or worse. But what I'm, I'm just interested in your opinion on that, what all of this M means for, for our future. And Veronica, how it, how it feels for you as well as someone nearer the beginning of the. The process. Because let's be honest, you know, at the moment, AI is being used to cut off off more people earlier in their career, uh, and going well. We don't need those people as much now. We can just, uh, use AI and that'll magically fix it, which I don't believe, but that's the, the attitude out there. And so I don't know where. I don't know what my question is there.

Speaker B: I think you've already asked.

Speaker A: Let people talk. Yeah, yeah.

Speaker D: Um, yeah, I, I think, think it will probably be a combination of all of those things. I think earlier we were talking about, um, you know, a painter in their medium and how it requires mastery of many tools, um, in order to be a great painter. And I think that that's an interesting thing to call to because obviously painting is not a very popular or commercial medium these days. But there are still people who are painters and I think, think something true, something similar will be true for design where there are still people who are just like, amazing at, uh, figma and who can do incredible things. Um, I also think there's been a shift in the industry. I know when Josh got into web design, um, many decades ago. Sorry.

Speaker C: We should say Veronica is not only my colleague, she's also my daughter. Daughter. I'm very proud of her. But you can see how this works out.

Speaker A: Yeah, yeah. This interview could have gone in a very different direction if we'd owned up to that at the beginning.

Speaker D: Right. I, um, think the beginning of this industry, like the beginning of web design, it was kind of the wild frontier and things like freelancing were super popular and kind of like the bread and brother, the bread and butter of many designers. And now, as Josh was saying, a lot of design has moved in house and I think a lot of younger people, like, a lot of younger designers are looking for those nine to five jobs within design and that, that has kind of become. One of the great things about this industry is that it has been able to provide that for many people. I think most of the younger designers I know are in fact at a company and not with an agency or freelance. Um, younger, meaning, like very young. Like, like post grad within the past couple of years. Um, so I. Which isn't to say that they're not in it for the art of it too, and that they don't love the craft of it. Um, I just wonder if they're maybe willing to get different things from this career. I don't want to be so, like, cutthroat about it. Like, it is still like a great way to spend time too.

Speaker A: Yeah, yeah. I don't. I think whether you're in house or external, I don't think should make much difference in terms of your ability to express yourself and be the designer. My concern, if I'm honest with you, Veronica, is that I'm seeing those in house roles beginning to disappear because there's this false presumption that, uh, oh, uh, we only need one senior person to manage a load of agents and we're sorted. Which I think is very, very naive. Not only from the fact that I don't think AI is at the position to be able to do that yet I also believe that we are shooting yourself in the foot for the next gener of senior people. So that uh, but, but am I wrong with that? I could well be out of touch. Uh, you know, but that's what I'm feeling.

Speaker C: I think that you're right and that there are new opportunities, you know, which I think is, is how technology works, right? Like it's like, oh, it closes some doors and opens new ones. And I think it's clear that the trajectory so far that we've seen, we certainly have seen it in software development and we're beginning to see it in design. We'll see how far it goes. But the trajectory suggests that a lot of production tasks are likely to be swallowed up. And I would say for a huge number of companies and individuals who want a good enough website, that work for web designers is going to be swallowed up by AI will help them get something up quickly. We've seen that with illustration, with other generatives stuff. So that's a drag. Um, but I also think that even as some of those tasks go away, the job doesn't go away and that a lot of the things with. So I think thing number one is it's going to be more and more important for designers to be behavior designers. Interaction design is all about the design of behavior, often to shape the behavior of the user. User, um, which will remain as true as ever of how do I kind of create the guardrails to help them get what they're trying to do done. But also now another collaborator in this of AI. How do we shape the AI, the design of the system in ways that we've been talking about already. Like you gave in your example of here's the material to use AI and here's how to use it. Um, go ahead. Um, the second bit that I'll say, and we're beginning to see this now kind of in some of the more forward looking teams is designers coming back again like in the old days, to working in code. And I think that seems a bit daunting. In fact it's AI working in code, but we're committing our work to the same common space as developers and product is beginning to put their documentation and requirements into that space too. And so at the moment that common space is often a git repo. And the thing that we've seen that that enables is really compressing handoffs and cycles. It's like, oh, if we're all working in the same place, development can come forward into the design process. We can participate in the development process as designers. Product can contribute to that. And so what we're seeing is like this really creative opportunity where there's more overlap in roles, which is challenging. Right? I mean, that's like a new management and role and career challenge. But there's also like, wow, with teams that get it. We've seen some, uh, incredibly powerful and creative results come from this. It is not just about speed, but is about all of these disciplines participating and collaborating in a richer way to create what ultimately matters most, which is a great experience for the customers and a powerful result for the company.

Speaker A: And I think, yeah, um, I would agree with that. I think there's a big democratization going on, um, in a lot of different areas. So, you know, back when I started, obviously I did code as well as design and, you know, you had to be something of everything. And then we, we moved into this world of huge specialization where we all had to specialize in different areas. Now I, I feel like the edges between those specializations are bleeding into one another because, you know, you can democratize the basic UI design process to product owners, for example, you know, with, with adequate AI guardrails and all of the rest of it set by the designer. They can be in a position to do prototyping, they can be in a position to do some of their own, but equally we can start, like you say, producing stuff in code. Now I find that very exciting and I love that idea. But there will be people out there that are going to struggle with that. And I have to say I kind of agree with Veronica that I think if you can get to, if what you love is sitting down and pushing pixels around, um, and creating a design like this, that you are going to struggle, um, unless you're exceptional. Right. The person that always springs to my mind in this situation is people like Mike Kuss or, um, Andy Clark. Right. Both of those people, ah, have got an exceptional design style. Um, and, and you will go to them if you want a style and an aesthetic. They're, they're going to be the, you know, they're going to appeal to the people who still like Vibe Final. Right. You know, the, the people that don't want the mainstream thing, that don't want, you know, the AI generated thing. But it's going to be a tough environment for a lot of people if you're not truly exceptional in that mechanics of building out a design.

Speaker C: Yeah, as always, will kind of, uh, continue to do well and, and adapt, but sort of the commodity Tier the production team. And this is a real worry about what has happened to design, like I mentioned earlier, of becoming kind of understood as a production capability.

Speaker D: Mhm.

Speaker C: It's like for the teams and organizations that have allowed that to happen, that will be replaced by AI. And the opportunity for organizations, design and designers is to return to really that more important um, piece of what design is, of understanding the problem of exploring the possibilities of collaborating with AI to deliver radically adaptive, individualized experiences where intelligence is woven into the interface. And that's like new stuff. AI can't do that on its own. That requires guidance of design around presentation issues that we've been talking about, about what the domain knowledge and tasks are, the expectations of the users. That's all stuff that for the moment is really uniquely human. And I mean I think this is a moving target. I think tasks are going to continue to be changing, but I'm confident that the job will remain. And like I said, I am a booster of what is possible here in the best possible way. Like really optimistic. And I also think that this is an exciting time to come into. Oh yeah. If the institutions can get out of the way of the young designers, um, people like Veronica, I think it's like, look out, go. Let's go. Like let's go.

Speaker A: Totally, totally agree with that. I, I'm, I'm excited too. Um, but I can understand people's concerns, I can understand people's worries. I think, you know, it is about kind of uh, being um, redefining maybe what you consider design to be in a lot of ways that you know, design is a lot more about, you know, than just pushing pixels around in figma. I know I keep using that line, but to, to think about human behavior, about you know, um, organizational benefits and, and work within the constraints of, of the technologies that we have and all of the rest of it. If you can get your head into that space, if you could get yourself seeing design in its broadest sense, then there is real, it is incredibly exciting. And I look back with huge um, gratitude that I happened to hit the web just as it come come came along. And so if you've just graduated now, you're getting that. If you can ride this wave because it will be an incredible ride. And I think it's hugely exciting.

Speaker C: I think that it's like, I think there's that opportunity for the, for the young ones coming up. I think the old heads have an opportunity here too. And one of the reasons that we

Speaker A: definitely, we're really important.

Speaker C: That's right.

Speaker A: Really Important. None of it will happen without us.

Speaker B: That's right.

Speaker A: Yeah.

Speaker C: Yeah, that's right. And we'll whisper it to you in your ear, in your sleep. But I think one of the big reasons that we wrote this book is we came into this with some fear and anxiety. And the more that we explored this, the more that we met, uh, the opportunity with a sense of not only enthusiasm, but relief that there is a chance here this could go horribly wrong. Right. So what we're advocating for is something is what we consider to be a good outcome. How can AI elevate design? Design instead of replace it? And some tasks will be replaced. But I think there's a bigger opportunity for us to embrace, and I think we have a lot of examples of how to do that.

Speaker A: Yeah. And, uh, to get back to the heart of what design is really all about in many ways, because, like you say, it's been processized and productized and reduced to pussing pixels. And so this is an incredible opportunity that I'm very excited about. If you want to learn more about, um, sentient design, you can do so by going to Boag World 3036, um, which will get you the show notes, and you can find a link to the book in there. You've also. Yeah, I was about to say. I was about to say you also could get 20% off of the book through to the August 31st with the very disturbing coupon code Sentient Dashboag, which makes it sound like I'm not currently sentient, but I am going to come alive in this kind of Frankenstein, Stein, sentient, monster way.

Speaker C: So we believe in you, Paul. We believe that you can do it.

Speaker A: I can make it. I, uh. Yeah, I will no longer be an amoeba. So, yes, there's that. Now, one of the things that Josh will know, but Veronica, uh, won't, is that at the end of all of our episodes, since night, since 2019, I nearly went 1990. Two. Thousand five. Five. Please, please, Veronica, tell me you were at least alive in 2005. You must be. It's fine. Yeah. So since 2005, Marcus has been telling jokes at the end of every podcast. We have tried. I've tried desperately to get rid of them, but we always get in trouble if we do stop them. So, Marcus, get it.

Speaker B: Not my fault, is it? You know?

Speaker A: No, it's not. I, um, mean, I blame you. I'm blaming our listeners.

Speaker B: Well, whether these are or not is a matter of opinion, but these are from a quite famous comedian called Tim vine, who's famous for doing one liner, dreadful puns, but he's also, he's also

Speaker A: famous for being funny. I'm just saying that m. He's got that advantage.

Speaker B: So this is, this is the delivery. Okay. This is the delivery that's going to let it down. So I'm preparing you for that.

Speaker A: Okay?

Speaker B: Velcro. What a ripoff.

Speaker C: I love it.

Speaker A: One more.

Speaker B: One more.

Speaker C: Yeah.

Speaker A: Cool. I'll do another one.

Speaker B: Exit signs. They're on the way out.

Speaker A: Once you've had one, the second one doesn't land as well. No, uh, that's good.

Speaker C: That's good.

Speaker A: That's all you get.

Speaker B: No more.

Speaker A: Yeah, I like it. Well done, Marcus. All right, well, thank you guys for coming on the show. Hugely appreciated. Um, and uh, yeah, where can people find out more about you both?

Speaker C: You can find us@bigmedium.com, uh, which is the agency that I lead and the agent and where Veronica is my colleague. Uh, we publish a lot there and we have a newsletter. And of course you can find out more about the book and our publisher, Rosenfeld Media. Dot com. And I guess with that, Veronica and Josh. On the way out.

Speaker D: On the way out. Thank you guys so much for having us. Thank you.

Speaker A: Oh, it was a pleasure. Sam. Mhm.