← The AI in Business Podcast
The AI in Business Podcast
Why the Way AI Feels Is as Important as How It Works - with Carsten Wierwille of HTEC
00:25:16
Loading summary
Transcript37 lines
- [00:13]
A
Welcome everyone to the Emerge AI in Business podcast. Today's guest is Carsten Viravila, Chief Product and Design Officer at htech. Htech is a global engineering firm focused on AI centric software and hardware development, working across financial services, medtech, automotive, telecom and enterprise software from more than 20 engineering centers. In this conversation, Carson examines why design keeps arriving too late in enterprise AI initiatives and why that timing error is a strategic mistake rather than a workflow inconvenience. The conversation covers how AI should amplify human judgment rather than replace it, why the MVP framework breaks down for genuinely novel AI products, and how the role of design is shifting towards cognitive design thinking about how users will perceive, decide and trust before anyone writes a line of code or trains a model. Today's episode is sponsored by HTech. Emerge's editorial team has already earned the trust of some 85,000 business leaders and practitioners worldwide. Clients ranging from Fortune 500 enterprises to AI startups use Emerge's research based content to build that same trust with the executives and practitioners they're trying to reach. To learn how leading AI brands use Emerge to connect with enterprise buyers, visit emerge.com add1 that's emerj.com ad1 now the conversation with Karsten. Carsten welcome to the Emerge AI in Business podcast.
- [02:05]
B
Delighted to be here.
- [02:06]
A
That's great. I'm excited to pick your brain on design and AI today. You've been at this intersection of design, technology and business in general for over 25 years and something that I've heard from you, very specific frustration, is that design keeps arriving at the wrong point in the process. I think you've described it as there's a difference between building things because the technology enables it and building things because you understood that there was a problem that had to be solved. That framing feels especially relevant today, right now, when every enterprise is under pressure to just deliver on AI features and products as fast as they can. From what you've seen happen over the web era, the mobile era, and now, I guess, the AI era, where does design typically fit into an enterprise AI initiative? And what gets missed when it's thought of as an afterthought?
- [02:58]
B
Yeah, you're right. What often happens is when new technology introduces that we build something just because we can, and suddenly a web experience on a BlackBerry or worse on a feature phone back in the days, they're technically possible, but it doesn't really add a lot of value. I think this is very common. You have new technologies, there's a lot of excitement, there's venture Money things are built in the lab. And then when we're actually trying to figure out where's the real value here for users and for businesses, we have to tweak that. And that's typically when design gets involved. So not in the initial stages of the technology lifecycle, but at later stages. I think at the company level, when we build something just because we can, the problem is that it's often not solving the business end user problems that it's meant to solve, that the adoption is low, or that we're creating new products that no one buys, or even when we are building something with a focus on automation and efficiency gains, that that efficiency gain doesn't materialize because it sort of overlooks the human factor. Right. I think what you, what you read a lot about today is that we've invested a lot in AI, but we don't quite have the productivity savings or we don't quite have the return on investment. That's a more complicated problem. But part of that problem is also the human dimension. Are people working faster? Do they want to work faster? Are they working differently? These are all, broadly speaking, design problems. They deal with human motivation and interest. And if we don't understand them, if we don't know what is the right problem to solve, then we're bound to be disappointed.
- [04:21]
A
And it sounds that, I think we also have. I don't want to say we have leaders thinking this because leaders shouldn't be thinking this way, but I'm sure that there is an idea that we're thinking about tech and we're thinking about, oh, what can this do? And are thinking about our problems, but also mostly because we're not sure of the problems. Can you give us an example of what design as an afterthought actually produces in an AI context? Where does the user or the business feel at most?
- [04:46]
B
You mean, if design is an afterthought, what is the result that we're getting? I think you're getting that all the time. When things, whether they're built with AI or not built with AI, when the solution, the application, doesn't quite give us what we expected, it's sort of functionally there, but users are not adopting it. You are not getting the revenue, the sales that you're getting from, you know, an AI feature in the product. Or maybe you are implementing AI in the context of a customer service experience, which is very common right now, right. In call centers for chatbots, et cetera. And you know, it creates more problems than it's meant to solve. So I Think you're getting problems across the board if you don't have a proper integration of design and engineering, a proper understanding of both the technical problems and the user business problems.
- [05:35]
A
I spoke to one of your colleagues in this series, Ronnie, and he gave us a hint and he said, if you really want to know if it's working, take it away. Implement the new AI program and then take it away for a week. And if you feel it, if you feel that loss, then you know it's working. Is that also something that we can use as an indication of whether this design was effective and we, we thought about the right problem?
- [05:55]
B
Yeah, for sure. You can, you can take it away or I mean, you can just simply look at the adoption, you can look at what people are telling you about the solution. Absolutely. I think the adoption, the usage tells
- [06:06]
A
its own story, makes sense and it's practical. From what you're seeing currently in the environment, is this really still a problem? Are we still in the boat because we can phase for AI enterprise products?
- [06:17]
B
Yeah, I think so. I think we are definitely seeing it. I think there is a huge rush to get features out and there's a huge rush to demonstrate to investors and boards that the company is fully AI enabled. And so in that context there isn't often the time, or at least the questions aren't asked about what specifically is the opportunity. So it leads kind of to this rush. This is particularly dangerous now because the cost of developing something have dropped with AI. Right. So ideation is cheap, concepting is much faster. And so the bottleneck is no longer the creativity, the ideation of the bottleneck is now the review. And so you have senior people, senior engineers, senior product people, executives that are spending a lot of time reviewing work, reviewing in the worst case slop. And if you don't have the right criteria for that, if you don't know what exactly are we trying to do here, it just creates madness. And I think that's what a lot of people are experiencing right now. Right. I think a lot of, especially senior people and companies are quite tired because they're buried under the ocean of new ideas, new concepts, and at the same time there's pressure to demonstrate results.
- [07:19]
A
Yeah, and I think it's difficult to go back to a board and demonstrate results if you're not even sure where this is supposed to fit into the organization.
- [07:26]
B
Yes, yes. I mean, the easiest thing to do, I think a lot of the focus right now is on doing the old thing faster. I mean, it's on automation. Can we take documents and digest them and ingest them and review them faster. Right. I mean, we all do that in our personal lives and that's wonderful. I mean, the benefit is obvious. I think it's a very different task altogether to change a workflow, to cut some steps out, to give maybe people new sets of responsibilities. So I think that's really where the bigger opportunity is to use it, to use AI more truly as a creative tool in the broader sense. That allows us to rejigger and rethink workflows, that it allows us to change the way we deliver value to customers. Customers.
- [08:05]
A
Is that an integration strategy that you would recommend? Start by just using it to improve productivity, just making things faster, and then take the data from that to decide what to build onto it.
- [08:17]
B
Yeah, I mean, it's integration, but I think the core starting point is to think about how do we add more value, how do we improve the lives of the lives of users? Right. So maybe let's take a specific example. You think about the reality of a financial advisor, right? So they're proposed to manage the finances of their clients and they're expected to be always up to date in terms of opportunities and risks in the market. And yet often their day to day is filled with being a concierge to the bank, to do a lot of road tasks, to do paperwork, et cetera. So I think there, when you ask there, what's the opportunity that AI can provide, is you would allow that financial advisor to spend more time with their customers, spend more time really understanding the market rather than, you know, updating positions. Right. So you take in the core of what this person's supposed to do, which is help you manage your finances and you amplify that and you take away all of the tasks or a lot of the tasks that don't really directly add value.
- [09:17]
A
Yeah. How great is that? You can use AI to spend more time with your customers and that will help with retention, that will help with referrals. The opportunities are endless.
- [09:26]
B
Exactly, exactly. Because a lot of the things that these people used to be doing are still doing, can be automated. But the thing that's probably impossible to automate and that we don't want to automate is the human dimension. Right. It's the conversation about what do you really want to do with your money or, oh, so your daughter is going to college and how can we support that? Right. So you have a conversation about what, what are the things that matter to your client and how money can enrich someone's life.
- [09:54]
A
Yeah, that's crazy. So I know that we have a listener sitting there thinking, okay, this is great and it all makes sense, but how do I get to that? Where do I start? So what's the question that a product leader never asked at the start of an AI initiative, but absolutely should be asking this?
- [10:08]
B
I think product leaders typically are familiar with the basic idea of how to develop a product that the market that users are interested in. Right. They will do their research, whether that's desk research, market research, user research. So I think they will ask those types of questions. So I don't think that most people will know how to do their job. It's just that in the corporate context, in the pressure of bringing products to market, there may not be enough time to refine, to update, to polish. Because right now the pressure is that, oh, it's good enough, let's just get it out. So the pressure to shipping is overwhelming. And I think what people are underestimating is the time that it takes in this sort of ocean of concepts that I've described before, the cognitive and time toll that it takes to just review the mountain of ideas. So if you're not precise from the start in terms of what you're trying to build, then you will get overwhelmed.
- [11:01]
A
Is the secret to deploy smaller AI initiatives to. Obviously you said the pressure is there to do it as fast as possible. So if we take smaller workflows first, we'll get to results first or we'll be able to deploy faster. Is that a way that we can still keep everyone happy in terms of timelines and still get something done?
- [11:21]
B
Not necessarily. I think that's a good strategy when you're trying to incrementally improve something. So you're just trying to improve the acceptance of your credit card offer to take off them, or you're trying to make the customer experience in the call center just a little bit better. I think if you're trying to develop a completely new service or offering that didn't exist before, that wasn't possible before AI, I don't know, maybe like some kind of styling service, something that also has a more of a human and personal dimension to it. I think you need to think about this more broadly. You have to get the product, the idea, the let's say minimum viable product into a shape where the user can understand what you're after. So breaking that into. I don't think, I mean, this is sort of a much older discussion, right? The sort of 15 years ago we had the lean startup movement and this idea of minimum viable products and it Was this idea that you can test your riskiest assumption. So essentially what you're talking about, you break everything into small chunks, you test and you get feedback to improve. I think that works well when the core value proposition of a product is well understood. But if you're developing something completely new, if you developing a robo taxi service or a humanoid robot for senior care, I mean, these are completely different experiences that consumers don't have a reference for. And you're not going to. I don't think you're going to find product market fit by breaking it into small of a component. You have to have a sufficient enough artifact or application that people can respond to.
- [12:51]
A
I like that you're telling our listeners that we cannot be stuck in past strategies that MVPs are so last season. It's not for now.
- [12:59]
B
I wouldn't say that. I think it's a term that is just very often misunderstood and it has lost a lot of its meaning. The discipline that was meant to be in the idea of an MVP is often lost now. It just simply means it's our first product, our first release, and it's typically crammed full of features that makes sense.
- [13:17]
A
I want to get into something that I know you're also passionate about, and that is envision and realize that phrase is doing a lot of work. Most organizations can either do one or the other. It's very rare that we get them to understand how those two need to coordinate. And that's something that you can talk about from your experience, leading teams that have to hold both ends, both parts, and get everything done all at once. What actually changes about what gets built, when that integration happens early rather than later in the design process.
- [13:46]
B
Yeah. I think if you have that level of collaboration between engineers, product managers, designers, sales and marketing people, whoever involved needs to be involved. You can envision the right thing, but then also build the thing. Right. Right. So you can have that level of collaboration that's necessary. And I think AI provides a lot of opportunities to do this in a better way because everybody can now be in the same tool. You might have engineers and product managers, designers that are all in, you know, whatever, whatever your tool of choice is. They can all be in Claude code. Right. Their ways of working is much more similar than it was before and allows for new levels of collaboration. That's also sometimes uneasy now because it can feel that everybody is doing everybody else's job. So then people are asking a lot of questions. Are you eating my cheese? Are you stepping on my toes? Where does my work end where does yours start? Right. So we have to completely rethink collaboration and workflows, but I think this allows for a much greater level of collaboration. I think we can reach a greater level of fidelity as well. Over the history of digital design, there's always that problem of converting a great idea, a great design idea, a great product idea, into stable production code is not a trivial task. And now that translation, I think, is easier than ever before. It's still not as simple as some of the evangelists of AI want us to believe, but it's getting easier. The costs of implementation, the costs of translation from idea to execution is lower than it was before.
- [15:18]
A
And how is this affected by the reduced rework that actually needs to happen? What's the before and after after for a team that gets the integration right versus the one that doesn't?
- [15:29]
B
So I think what we're really trying to do is we're trying to replicate the entire process, the entire, let's say, software development life cycle in AI. So starting from a research document or a requirements document that is generated with AI assistance, people talk a lot about coding assistance, right? And that's where all, where all the discussion is. But really it's not just coding assistance. It's like assisting all of the human endeavors throughout the process. So from the conception to the product requirements, to the coding to the testing. And so all of that is replicated, all of that is manifested in the tool. And so that means that if I come up with a concept, if I can specify it, and there's sort of a framework called Speckit that we're using, where everything is basically stored in the same fashion, you can then go back, and if you make an update to the specifications, then making updates to the code can happen in a fairly seamless way.
- [16:25]
A
Carson, I'm going to tap on something that I know you're the perfect person to answer me on this one, because we're telling our listeners how important design is in this process. And early alignment is super important. But at the same time, we're also telling them that design is increasingly being automated by AI itself. Generative interfaces, AI assisted prototyping, auto layout tools. If AI can design its own interfaces, does the design layer become less strategically important?
- [16:52]
B
I think the role of design is changing and it's not new. It has changed over decades. Right. People that studied graphic design in the 1990s would still make their own type. And there'd be exercises where people cut type out of paper or cardboard. And of course, no one does that anymore. That's all automated away. It's similarly the task of creating a user interface, a mobile design or a web design has been even before AI, commoditized and standardized. And so what was craft becomes automated. And so then the craft does not go away, but it shifts. Today, craft is the way a design is expressed in a text file, or it's called a markdown file, which provides the specifications for AI to create an experience. So the tools are just changing where humans can add the most value. And that is true for engineering. I would argue the transition there is even more dramatic and it's also changing for design.
- [17:44]
A
And I feel that bridges us into the closing part of our conversation very nicely because what we're saying is design is not becoming less important and it's just the whole idea of what design is, is changing. So for a senior leader who is in the middle of scoping an AI initiative right now, figuring out what to build, who to get involved or in what order they should be doing it, how should they be thinking about design? Before anyone starts writing any code or trains any model, I think that they
- [18:11]
B
can turn to their design team to really think about both the functional and the subjective aspects of their users. The people who will in the end be using an application, whether those are auditors for finance applications or your end customers that are paying for a product. And there is sort of the more functional perspective to think about. What are the jobs to be done? How do we make people's lives easier on a functional level? And then the more subjective dimension of how does it make them feel? What are their motivations? And I think designers are not uniquely equipped, but particularly well equipped to address those questions. Often we think of this as something that is a creative task that is happening more in the beginning of a process, but it shouldn't be limited to that. I think one of the things that I find fascinating now is that design teams and user research teams and bigger companies are involved with defining the criteria, or what might be called the rubric for the AI evaluation. So they're specifying, sometimes even in quantitative form, how will we measure the output of the AI. So there you kind of see that the design perspective doesn't just end with the conception, but it goes all the way to how do we evaluate the AI output.
- [19:19]
A
I like that. And how. I also want us to just work on one more idea that's going to fit into that is designing for trust. I think that is a very specific challenge in AI enabled experiences. What makes trust a design problem rather than just a technology problem?
- [19:38]
B
Yeah, There are a couple of dimensions to this. There's do I trust the technology? And I think there's a number of strategies there to say, provide transparency about the output, how is the data created? What am I looking at? Why am I looking at this? So to be able to maybe double click to find out more about a particular result. Right. Maybe I use the AI to give a recommendation, a recommendation for a trip to Goa, or maybe it's something more functional, like your credit application is approved or rejected. And so what makes me trust this solution or answer more is if I can unpack it a little bit, if I can say why was this recommendation provided, who provided it, how can I change it? What if I change some of the parameters? How has this changed the result? Right. I think all that's conducive to increasing trust in the results and the technical efficacy and efficiency of the solution. There is the human dimension to this as well, which I think is often overlooked now, which is that in many large companies, employees are encouraged to use AI, but they feel that if they do that, they work themselves out of a job. So it's do I trust my leaders that they're doing the right things with this? Or. Or is this a trap, so to speak? That's maybe more of a leadership question than a design problem. But still, I think you want to approach this with this kind of human empathy to think about how do people think about it and what's holding them back and how do we set free their creativity, their ambition, their productivity.
- [21:06]
A
Yeah, I think that last point is definitely a powerful thought to leave our leaders with. It kind of combines in with culture as well. So it's not just trust means different things for different parts of this conversation, which I like. I'm going to try and capture the essence of our conversation. So for our leaders listening, the three main things that you should be taking away from our conversation with Carsten is that design arriving late in an AI initiative isn't a workflow problem, it is a strategic mistake. Definitely. And the organizations getting AI products right are the ones where we see design and engineering integrated from the start, not as an afterthought or not as a sequence. That one is more important than the others. And then for our leaders scoping an AI initiative, the most important design question isn't what it looks like, it is whether it solves your business problem. Carsten, anything else that you would like to leave our leaders with?
- [21:59]
B
Yeah, one thing that might help is a term that we use that may sound a bit abstract, which is we Talk about cognitive design. So it's not necessarily the creation of an interface, a ui, a UX type of experience, but really thinking about the how humans will think, perceive, remember, decide, solve problems so at a higher level. And I think that design because of the automation is shifting more towards thinking about those higher level tasks of how do we actually allow users humans to solve more of those higher level problems. I think that's what we are very focused on at this moment in time. Cognitive design.
- [22:36]
A
Yeah, something to think about and definitely something that will spark new conversations as well. Well Carsten, thank you for being so direct about this, for sharing your insights and giving us a perspective that can only come from someone that has lived this across every era of the digital product design. It's been great having you on the show.
- [22:53]
B
Thank you so much. My pleasure.
- [23:09]
A
Wrapping up today's episode, I think there are three key takeaways from our conversation with Carson. First, the reason enterprise AI investments underdeliver is not the model or the data. It is that the human dimension arrives too late. When design is brought in after the technology decisions have already been made. The result is AI that functions but does not get used. Adoption failure is a design failure and it is one that gets more expensive the later it shows up. Second, the shift to AI assisted ideation has moved the bottleneck in enterprise product development from creation to review. Ideation is now cheap, concepting is fast and the constraint is the judgment capacity of senior people who have to evaluate an ocean of concepts without clear criteria for what good looks like design. Clarity at the start of an initiative is not a luxury luxury. It is the only thing that makes the review process tractable. And finally, the role of design in AI is not shrinking as automation takes over visual and interface tasks. It is shifting toward what Carsten calls cognitive design. The question is no longer how does it look, but how will users perceive, remember, decide and trust? That is a harder question than interface design and it needs to be in the room before any model is trained or anything code is written. Emerge editorial team has already earned the trust of some 85,000 business leaders and practitioners worldwide. Clients ranging from Fortune 500 enterprises to AI startups use Emerge's research based content to build that same trust with the executives and practitioners they're trying to reach. To learn how leading brands use Emerge to connect with enterprise buyers, visit emerge.com add1 that's emj.com at number one. On behalf of the team at Emerge, we'll see you on the next episode.