Deep Dive
Shownotes Transcript
The realm of possibilities is expanding. You can use AI as a support in your research process for processing data. Take everything it does with a grain of salt. So make sure to really think critically about what AI does.
and what you have to ask from an AI system and never eliminate real people from your research efforts because you will end up with a very stereotypical, limited solution and view of the problem that you're solving. And so the AI-generated insights would be seen as pointers or starting points, but they're not replacements for authentic human conversations or for a proper design and a design system philosophy that translates into your product. ♪
Hello designers and welcome to a new episode of Honest UX Talks. This is the second episode I'm doing on my own because it turns out that I have a lot of things I want to talk about and they're mostly about, I wouldn't call it futurism, but they're around capturing the present and encapsulating some directions about where the future of design is heading towards and the future of our careers.
This time we will be discussing about how the design process is transforming and I will be unpacking some of the most common areas of the design process where AI can support us, where we can be mindful, where can we trust it and the things that it can't really do well and the parts that will forever be in our hands.
So, with this intro, I want to just take a moment to thank our sponsor. Since we're talking about AI, I just want to mention that Wix Studio offers very interesting AI capabilities. If you haven't played around with them, you should definitely explore.
One of the things that I found super useful is the object eraser. You've probably been in situations when even the most minor design change feels like a huge hassle. So let's say you want to edit a photo, it requires you to often open a new software, go to Photoshop or wherever you're using, you're editing that photo, edit the image, the image is exported, uploaded,
uploaded and so on. So this requires a lot of time, but you can now do all of that inside the editor with Wix Studio. So with this tool, you can now erase certain elements from an image or erase an area of the image and make it transparent. So converting it to PNG. Also, you are given access to the AI Expand.
so generative feel, very similar to the Adobe capabilities. So imagine your clients want you to add an image to your design, but it doesn't fit the layout. Resizing it in your design software often leads to this frustrating choice, distort the image and compromise quality or awkwardly crop out key elements. And so this disrupts your workflow, forcing you to either find a new image or spend extra time in a separate place.
photo program. So you are now able to edit image in a way that it adapts it to your design with AI and not the other way around. So AI is already a part of our design process with tools such as Wix Studio, and it's going to be a part of our process even more moving forward. But I would love to explore some of the things that we need to keep in mind, both from an ethical mindset, even philosophical perspective.
and then dive deeper into how we can incorporate AI in our work without compromising the quality, without compromising the depth of the work that we're doing and the value that we're putting forward when we're releasing outputs in the world and outcomes. Yeah, so with that being said, let's start. I think we're at a stage in the design industry where we're all feeling a mix of excitement, maybe we're a little intimidated, but
Most of us are curious about what the tools can do for us, how the work is transforming, will we all become builders from that point on? We have a recent past episode where we discuss the emerging roles and
You know that my prediction is that eventually the design, engineering and product role will converge and there will be this unique role where we're all just building thinkers, architects, makers, right? So that's where I feel the future stands. So this is in a way frightening.
but it's also exciting and at the same time I feel that we feel a tremendous amount of noise and maybe we feel what I like to call the AI fatigue so we're all a bit tired by these topics everybody's posting about AI recipes and AI tools and how to use them and so on I was recently talking to Rit Michael Rittering who runs Figma Academy and he's a great designer he has always inspired me and actually asked a really interesting question in our one-on-one where he said but
you really use AI tools in your process? And at that moment, I realized that my answer for the past three years has mostly been not really not that much. I do use ChatGPT. At that point, across the years, I was playing around with different image generators, maybe, but they didn't really translate into UI work at any point. Maybe I could do some, I don't know.
content support for my UX goodie stuff, but I never really used UI generators, image generators deeply in my design work. So the future is not really here, partly because these tools don't produce the level of quality that we're expecting just yet when it comes to mock-up generation and design generation. They're still evolving, but I think they will get there.
But the truth is that we don't really need to use 30 AI tools at this moment, right? So there's this pressure on designers, try this tool, try this. And maybe I'm partly guilty of that, right? I talk about a lot of tools because I'm passionate about freely, loosely experimenting with them, see what they can do, what are the possibilities.
abilities of these new technologies? What are the limitations? What are the dangers? And so I play around, I experiment, I share that. But you don't have to use 30 AI tools in your work to be productive or to even benefit from the power that AI is now presenting us with. And so I think that should take a bit out of the pressure. I know that Reid also confirmed that most of the senior designers he asks this question to have a similar response. Well, I
not that much. I don't really use AI tools. Maybe, I don't know, here and there, chat GPT, you know. And so this is what the reality of designing in 2025 looks like. Of course, it's really interesting to build with tools like Lovable, Cursor AI, play around with the things that we weren't able to do before, like coding. Now we have expanded our
our powers, if you want. We have expanded the possibilities. And I think it's really interesting to explore beyond our design role. But in our design role, we don't really use AI that much, right? Still, I feel that it's a very interesting exercise to, right now, while you're listening to me, take a piece of paper and a pen...
If you have them at hand, then quickly sketch your typical design process. And don't go for the checklist necessarily, but like on your current design challenge. What is the process you have chosen to orchestrate, to go for, right? So you've probably done some research and then you have done
I don't know, stakeholder interviews and you've captured the business requirements, you've captured the users, the problem space, right? Unpacked it, synthesized and processed and extracted insights. And then maybe you went into solutioning and ideating and transforming all those insights into solutions and potential features and then refined those features, maybe loosely tested them in a very rough state and then increased the
fidelity of the design you're testing, and so on, right? So it's, in a way, design goes on a pretty typical process. Of course, the more complex the challenge, the team, the context, the more complex the process is going to be. And we've talked about it many times that the design process is nonlinear. So you don't start from A and go through 10 stages, and then you arrive at the destination. It's messy, it's back and forthy, you keep
Going around sometimes in circles, there's a lot of ambiguity, there's a lot of pausing, there's a lot of getting stuck along the way, revisiting the design questions, revisiting the research questions, realizing you don't have enough confidence and then you have to go back and build more confidence before you move forward and so on. But nevertheless,
We can pretty much describe the design process as a set of activities that we typically do in whatever order, but in order that makes sense to our project. And then we arrive at this kind of solution that we put in front of the users and then the process continues. We see how the users respond, we measure, we iterate and so on.
So if we go through these typical stages, and by now I hope that you paused for a second and just drafted your typical design process, and it's probably going to look quite similar among us. Most of the times we start with research. The question here is can AI augment or supplement, or I would never dare to say AI can ever replace research, but can it support us in the process? For once, one of the things that AI is good at is processing large volumes of data and
we are not very good at that. And this is a place where AI could actually come in handy. So producing research data is not the hardest part. We can have 10 interviews in a day and produce 10 interviews of one hour. And that's going to be a lot of content, a lot of text, a lot of data to process. And sometimes we don't have the
time. So I've seen many companies and I've personally worked on projects and on teams where we would produce data, we would produce research, but we never set aside enough time to go through it, to interpret it, to review it, to re-watch it, right? So a lot of that was just wasteful. Like we got a sense of what our users want, we got a vibe from them, we built an understanding, we advanced in our understanding of the problem space, but a lot of that was just wasted.
And so AI can help with that. And I think it's a very nice opportunity. For once, AI can help us process all the data that comes from research. Of course, there was a very early article. It's from 2023 by Nielsen Norman. They're essentially spreadsheets.
splitting the research tools into categories, insight generators and collaborators. So essentially, insight generators are the ones that are accomplishing a similar scope to what I was saying. They summarize user research sessions based on the transcripts, but
At that point, they had some limitations. Of course, with Notebook LM, for example, which I sometimes use to document my podcast episodes, you can give them now more context. You can give them other types of formats, PDFs, a list of the customer complaints from your call center. The thing is that a research tool is as good as the context understanding it has. So to be productive, a tool should really grasp a complex context.
If it's unable to capture that and accept different types of materials to build context, you can't really rely on its insights without the background information that it needs, right? Yeah, that's one of the problems. AI collaborators are similar to insight generators.
but they can accept some contextual information. Researchers can, for example, show the AI some human-generated codes to train it. The tool can then recommend tags for the thematic analysis of the data. In addition to session transcripts, it can analyze the researcher's notes. So collaborators can create themes and insights based on input from multiple sources. And so they're more sophisticated. And I think even from
The study from 2023, the analysis that Nielsen and Raman performed, I think these tools got better. There's also a very controversial thing happening in the industry. There is a product called Synthetic Users. I'm a big fan of the founder of this startup. I think he's a brilliant guy. Synthetic Users is, as its name says, Synthetic Users.
So you can describe your persona and then it's going to produce with AI versions of that persona. Like you would do research with people who fit a certain demographic or description. It's, of course, based on statistical information, their reflections of an archetype.
Of course, every research tool will get some level of controversy. I recently discovered a very interesting Figma plugin. It's called Velocity. And what it does is that it analyzes your mockups and your prototypes as a person would.
and it looks at design coherence and also accessibility. It goes through the flow and then gives you insights like you would get in a usability testing session. So that's also a very interesting use case for researching because you get answers much quicker. Of course, there is a limitation that you need to actually show those to real users at some point. This is, let's say, an initial step
that helps you refine the prototype as much as possible before you show it to other stakeholders. You have some answers when you go in stakeholder meetings, right? So look, I run it through this AI simulation and it performed like this and that. And now let's show it to users to see
learn more so none of these should really replace talking to real people because real people are quirky they're unpredictable they have nuances they're sophisticated organisms right so there's no replacing that and
the level of learning that happens when you talk to a person can never be achieved by just doing an AI simulation, but they really can come in handy to help you speed up decision-making. We all get stuck in moving forward because we don't know what the next best step is. And so AI can support you in jumping from one step to another, but then at some point you really have to do the extended classical research effort, right?
Getting back to the imagined process, I've mentioned two tools that can help you with analyzing. So one of them is Synthetic Users. It acts like generating personas for you. And then Velocity. And then I just want to mention that none of these are ads, okay? But you can use Zoom, Zoom AI to generate, use their AI companion to generate insights from the interviews that you do with people. Or you can run interviews through other AI companies.
and they will also generate transcripts and real-time text from the conversations you're having. And then you can go into tools like Dovetail, where they also offer AI capabilities for tagging, for identifying themes and sentiment and so on. And you can analyze those interviews and you can further refine the research, like the raw data that you get from interviews and turn it into something that's easier to interpret. And once you have that, you can go on the next step of
of the design process, which is to gather data from multiple sources and start making sense out of that data. And you can do that in a place like Miro, for example. It's a whiteboard, you bring all sorts of formats and elements and you have sticky notes, maybe you have links, you have PDFs, you have all sorts of things in there. And then you can start playing around with different AI capabilities available to navigate the synthetis of that data faster.
And then once you've done all the research efforts, and again, bear in mind that you really should be very careful, right? AI tools in the context of research have limitations, right? So their recommendations are mostly vague, not very reliable. They, again, have a limited understanding of the context. Many times they are unable to show citation, the sources of their thinking, let's say. They also probably have a level of bias. But again, these are evolving very quickly.
and you will probably see that you now have an emerging research companion. When you move out of the problem unpacking space, you can start diving deeper into GPT. So the next kind of stage of the design process is mostly about structuring your thinking. And GPT is pretty good at structuring random, unstructured, raw ideas you can just
brain dump everything you've learned, all the insights, all the knowledge that you have, all the ideas that you have in mind, you can just feed it. And you can also do that in Notebook LM in a bit more of an organized structure. And then GPT can help you draft the problem statements. You can use it as a brainstorming partner for different solutions. It can also complement your research insights and your research thinking with a bunch of market research.
now with its deep research capability it can actually take time perform more advanced reasoning read multiple websites so it can come up with much more relevant and not hallucinated made up context for the market you're building launching that solution in but then yeah you can brainstorm things and then it can help you draft and refine the feature list and maybe
explain some of those features, maybe write user tasks, user scenarios, and so you can use it as a brainstorming partner. Again, as you see, I'm always talking about what you can do with these tools. So you can't really rely on these tools to have agency and understand what they're supposed to do. You still need to do the thinking, but you're kind of having this external assistant that helps you in a way move faster, unblock, expand,
pan your thinking, clarify your thinking, structure your thinking and so on. Then we're entering the era when we're starting to visualize this thinking. We're starting to visualize design, visualize the solutions. And you can use AI to draft wireframes. I think you can do that in Figma with some plugins. I think even make design or first draft
and Figma supports that. You can use tools. In the past, there was Wizard, which was in the meantime acquired by Miro, but they were turning rough paper sketches into UI. That's a technology that's increasingly prominent. I know of a tool called Designverse that solves for the same problem space. And so you can now start from paper sketches or just text to design. And now you're gonna have some very basic mockups
Not very basic, but you do have to spend some time to train the AI generator you're using to train it on the brand style, maybe on the design system, on your requirements, on the preferred imagery. So it really has to be educated in the sense of your aesthetic and direction and problem space. And then it can produce more results.
relevant designs. That's also, I think, a transient problem. It's because of the infancy of these tools, but they will definitely get better. So we will not be generators of UI anymore. We will not be pixel pushers in the future. We will just be curating the things that AI is producing from a visual UI standpoint, right? So AI will produce and we will decide, and this is where a very new, important, very circulated skill comes in based on our taste, our
We have to educate our taste, expand our taste, invest in formulating and building a taste. That's where we will be able to make an impact by co-working with what AI produces to make it beautiful and pleasant and useful and relevant.
Of course, you can use AI now to generate color palettes. There was a tool called Chroma. I think in the meantime, you have multiple tools that do that. Kitol is another tool that I've been playing around with to generate product imagery. I am very excited about a tool called ReCraft.
Recraft is absolutely great at producing images and set of images that help you kind of stay on brand and create a sense of unity by producing multiple images in the same vein. And of course, you can define also custom style. So you have more control now in 2025 on what the...
input the AI system gets in order to produce a relevant, useful output that you can actually use in the wild. Because many times they produce things, but it's not there. It's not okay. Yeah, I can't really use this. I have to remake it. You can continue to create product imagery with AI, like these nice mock-ups of people holding your app and enjoying the time there and so on. So the realm of possibilities is expanding and
And yes, definitely you can use AI as a support in your research process for processing data. Take everything it does with a grain of salt. So make sure to really think critically about what AI does and what you have to ask from an AI system and never eliminate real people from your research efforts because you will end up with a very stereotypical limited solution and view of the
problem that you're solving. And so the AI-generated insights or the AI-generated output for UI, all of these should be seen as pointers or starting points, but they're not replacements for authentic human conversations or for a proper design and a design system philosophy that translates into your product.
And more and more since AI will support with a lot of these functions that again, you're at the center of the process and you decide what the process is. You decide what are the stages that you need to go through and you adjust them as you go. So you course correct at different points and you go back when you need to go back. So you're constantly making decisions and sitting at the center of this process. You're orchestrating solving these problems.
You're in the driver's seat. So it's not AI. AI is just complementing and helping you structure and make progress at the points where it feels difficult or more friction-y. Yeah. So I would say embrace these opportunities, but keep your mind very active as you do everything and be very intentional when you're designing things.
and constructing things with the help of AI. And always think about what is the sentiment, the feeling, the experience that you want your end user to have. Because in the future, I think things will start looking increasingly well. So if you think about the internet in its early days, it was so ugly. And now it's a pretty beautiful
place. And we're going to see the same thing accelerated now that AI can produce so beautiful imagery, right? We can just do more of the things that we find beautiful. So in a world of beauty, you can stand out through amazing storytelling, a strong brand, a strong value proposition, a strong experience, right? So beauty will be very available with the help of AI. And I think that we need to become storytellers and we need to
be very intentional about solving real problems in real ways and useful ways. So I think that that's my last invitation to you. Yeah, embrace AI as a partner, but take it with a very strong grain of salt. Think about the limitations and keep your mind very active. Use it as an opportunity to spend more time doing creative things as opposed to doing tedious things like pushing pixels around or responsive design or stuff like that.
AI is an amplifier, it augments our human skills, but our ethical compass, our critical thinking, our systems thinking, our intentionality will always remain indispensable. I think we need curiosity, we need caution, we need a commitment to create valuable and responsible outcomes.
experiences and ethical inclusive experiences and i hope that we're all gonna be promoters of good in the age of ai so i hope this episode was useful please send us ideas for future episodes submit your topics you can text me directly on ux goodies maybe rate these episodes if you like them rate our show to support us and thank you for tuning in bye everyone