#109 The UX Design Process in the Age of AI – Where Does AI Help (and Where Does It Hurt)?
22:07 Share

Hello friends and welcome to the future of UX, the podcast where we explore the trends, the challenges and everything that's shaping the future of design. I'm Patricia Reiners and in each episode we dive into the intersection of UX, technology and the future so you can stay ahead of the curve. Today we are tackling a question that's on, I would say, every designer's mind at the moment.

AI is changing how we work, but does it enhance or undermine the UX design process? UX is human-centered by nature, but AI isn't human. So how do these two things coexist?

We will talk about how or can AI truly understand users and their needs? Can AI generated designs replace creative thinking? And is AI making the UX process faster and better or less thoughtful?

In this episode, we will break down the five phases of the UX design process, emphasize, define, ideate, prototype and test and look at each one of them and do a little deep dive in how AI is or can be used in each state, where AI is actually useful and where it's risky and how designers should adapt to keep UX truly human-centered. I would say let's dive right in.

Let's get started with the very first phase: Emphasize. Can I really understand users? And maybe a quick disclaimer:

I think we all know that not every single design process consists of these five stages in the certain order. Design is messy, design is chaotic. But just to make it a little bit easier, we will go through those typical five stages. Although your design process for the project that you're currently working on might look a little bit different, this is totally normal. Just to make it a little bit easier,

We will focus on those five stages and we will get started with the emphasize phase. So the phase where we really want to understand. We go broad. We want to understand the user. We want to understand the problem. We want to see what this is all about. And imagine you're running a UX research study or a customer feedback for an app update.

You decide to use an AI-powered sentiment analysis tool to scan thousands of user reviews. AI tells you that 90% of the comments are positive. But when you read them yourself, you see messages like, Wow, love that this new update crashes every time I open it. Great job.

So AI saw love and great job classified as positive, but any human would know this was sarcasm. So how AI is being used in user research. AI powered tools like Dovetail, UX Tweak, ChatGPT can analyze massive amounts of data. It can be services, it can be interview data. This is already super helpful.

AI-driven sentiment analysis detects frustration, also satisfaction, and patterns in user feedback. And even chatbots can conduct automated user interviews to gather insights, although I wouldn't recommend this, so I think this is a bit tricky. So what we can already say is: User research and emphasizing is a lot about data, gathering a lot of information.

And what AI is really good at is structuring, analyzing, synthesizing the data and making sense of it, which takes a lot of time. So AI is wonderful at that. What is AI not so good at or what are the risks of AI in UX research? First of all, sometimes AI misinterprets contents and it can't fully grasp the tones and sarcasm or hidden emotions.

When you think about qualitative user interviews, there's a lot happening behind what's being said. For example, when someone, you know, just makes a certain impression and you see just like the way how they react, that they actually don't mean it that way. And there's a huge difference between what people are saying and what people are doing. And especially when it comes to like usability tests or where it's really important that you

have that you also see the interface so where the user is clicking on you have this qualitative interview and the user is clicking on a certain button or it does something wrong so for AI it's still very difficult to analyze basically the video because it can't analyze the video it can only analyze the transcript makes then become like the connection very difficult another risk is that AI is only good at it

as its training data. So if the training data is biased, the insights will be too. And this is the case for every AI product. And empathy isn't just about recognizing patterns, it's about human connection. So the key takeaway here for the emphasize phase is AI can process user data, but real empathy comes from human to human interaction.

AI is a wonderful companion when it comes to research, but it will never, never, never, never replace real user interaction and real user feedback. And I think this is super important and I can't stress that enough. Although there are tools like synthetic users or attention insights, that's all great. I mean, if you want to use them, use them if it's helpful, if you don't have access to user research, but use the insights there as hypothesis, as assumptions.

and not as fact-based information. So you need to validate these information at some point. Let's move to the next phase, define. Can AI help turn data into real insights? First of all, the define phase is all about synthesizing all the insights from the user research and really trying to make sense of all of that. At that point, you have a messy collection of user research, of feedback, of data, of usability tests.

And there are some tools that are super helpful. Thinking about Miro, thinking about FigJam or Notion AI or ChatGPT, Perplexity AI, Cloud AI will help you to summarize the data and help you to come up with suggestions, come up with summaries, help you to categorize the insights and then suggest something. For example, users find this checkout process confusing. Users want more personalization in the app.

That seems right and helpful, but sometimes there's a little bit of a problem because UX designers don't just define the problems. They also look deeper into the hidden needs. So for us also to use AI correctly, it's super important to not prompt the AI to say like, what are your suggestions? And then things come up as the example. You just find the checkout process confusing.

That's wonderful, but like why? We need to understand why. Not that they find it confusing, but what caused these confusions? What is the problem? Where did they get lost? So AI might summarize that users want personalization, but a human researcher, a human being needs to uncover that users actually feel overwhelmed by too many choices. So this basically critical thinking needs to be done by a human.

AI can also help you define the UX problems. AI can organize these large amounts of data into very clear themes. It can categorize it into pain points, into needs, into goals. Or you can use, as I already mentioned, AI-generated heat maps and predict where users will focus their intention. You can do that, but again, this is an assumption that you still need to validate at some point.

Or AI can process survey results and extract patterns at scale. This is super helpful, especially if you have large amounts of data. So where AI fails, especially in the defined phase, AI doesn't ask why. It only finds patterns in past behavior and it can't ask why. And AI might reinforce existing biases instead of uncovering new insights.

Some UX problems can be defined by data alone, but intuition plays a huge, huge role. So what is the key takeaway for the defined phase? AI is an amazing helper, an amazing assistant in organizing insights. But we still need this human being, and we will always do.

who needs to define the real problem, who have this critical oversight over everything and draws real conclusions that really go deep, that ask why and maybe, you know, sees where there are certain gaps in the research, where there are still certain questions that needs to be answered. Let's move to part number three, which is ideation. So coming up with features, with new ideas and brainstorming.

AI can be creative. AI actually really shines when it comes to brainstorming ideas and brainstorming features and coming up with new content. I mean, how new it is, is another question, but definitely coming up with ideas.

And this is super helpful and I can highly, highly recommend to use AI, especially in the brainstorming part where you have all the insights, where you have your problem statement defined, where you really know what this is all about, where you have your research insights, you understand the problem, where you get a sense of the solution already and then use AI to brainstorm and then also to prioritize the different ideas and then transfer this into features maybe.

And for this stage, I can highly recommend tools like ChatGPT, Perplexity AI, so all the large language models. Ideation is very much about

content. So it's not so much about like visualizing content. I mean, of course you could, but the large language models are the best ones for especially those faces. A few things that I think where I falls or fails a little bit in the entire ideation process

It can predict certain features or can recommend certain features based on past designs, based on other products.

And sometimes it doesn't create something entirely new or it goes too crazy. And my tip and recommendation is there to be also when you come to prompting the language model, be very, very precise. So you already mentioned the target group, the insights, the problem statement. So the more details you already give there, the better the features are. And then don't forget to iterate.

So don't use the first ideas, but iterate, iterate, iterate, prioritize them, categorize these ideas, then use this as a starting point. You know, so you don't need to start with a blank page. Okay, so now let's come to this episode's sponsor, Wix Studio. Web designers, let's talk about the C word, creative burnout.

Your client side has real portfolio potential, but between resourcing, feedback, end-tied budgets, and ever-tidier deadlines, it just doesn't make the cut. Wix Studio helps close that gap. Built for agencies and enterprises, you can bring your own vision to life and keep it alive with no-code animations, tons of AI tools, reusable design assets, and advanced layout tools. For your next project, check out wixstudio.com

That's wixstudio.com. Because a really big problem is that sometimes AI generated ideas lack the user context, so they need human refinement or really good information in the prompt. And yeah, this is definitely something that I would recommend. So the key takeaway for the ideation phase: ChatGPT, all loud language models are great when it comes to brainstorming about ideation.

But for you, it's super important to first of all prompt the language model the right way. So give adequate background information and then also double check the content, prioritize them and use this as a starting point.

Let's move to our step number four, which is prototyping. Does AI make designing much faster? I would say it definitely does. So how does prototyping look like? Generally, the prototyping phase is about visualizing your ideas or the concept that you came up with. Maybe you already got some feature ideas and you would like to prototype them to maybe even test it in the next phase or to present it to a stakeholder.

Prototyping is very important to help people to understand how the feature, how the product, how the idea looks like and also to test it. And AI can be very helpful. There are wonderful tools like Galileo AI or Vizilli, Relume. Many, many amazing tools that help you to basically to create a mini prototype

with just a simple prompt. My tip here is, and also what I'm seeing, you know, using these tools in workshops or with clients, is that they are not at a point where they can create visually perfect results.

So what AI tools will present you is a very first draft, a wonderful starting point that is, I think, the best starting point to start a discussion with your team, to share this with clients, do some first testings maybe even. But don't expect it to be perfect because it won't. It won't be. And I think it's good to have the right expectation when it comes to AI prototyping.

Another really important thing is when you prompt AI tools to create via frames, to create designs, don't be abstract because they can't, they have no idea. Before you prompt these tools, think about how the feature already looks like. Use another large language model like ChatGPT, for example, to brainstorm. Okay, I want to design a new search feature

search area for finding the perfect flight in a booking system what do I need what kind of functions do I need okay I need something like to put in the date probably I need something to enter the airport the arrival the departure airport maybe I need a filter maybe I need

Pads that I want to bring, people that I want to bring, any kind of like services I want to add. So you already think about like what are the different components that you would like to have included. So be very specific. Otherwise, you will get something very random and something that doesn't really make sense. And this is what I'm seeing what a lot of people unfortunately doing wrong and then they are super frustrated with the results.

Then this is also the problem where these tools fall a little bit short. They can feel very generic, they lack personality, but the better you know what you want, the better you can prompt the AI. Don't expect you to do the AI, like do AI or let AI do all the work for you. This is definitely not going to work. The key takeaway for the prototyping phase is AI can accelerate design, but designers must

First of all, define what they actually want in the prototype, how it would look like, and what they need to test ideas thoughtfully. So when you think that AI is going to take your job because there are so many tools that can already design and create visually beautiful things,

No, because they don't know what you actually want to get across with this prototype. What's your goal? You don't do this for fun. You do this to present a certain idea to a stakeholder, to do some testing. You have certain questions that you would like to have answers with the prototype.

Okay, now let's move to our last phase, the testing phase. Testing is using the prototype, using the assumption, the hypothesis and testing this with real users, also tested with AI. But a combination is usually the best.

There are AI powered usability testing tools like Amaze, like Lookback, and they predict user behavior before you do the real tests. This can be helpful. You can do that. And they offer certain AI based eye tracking tools, simulate where users will look first.

I already mentioned the tool Attention Insights, which create heat maps based on AI prediction of where the user will look at or AI can automate surveys and chatbot-based feedback. This is all wonderful. This can all be great, but still real testing is definitely needed.

Some areas where AI absolutely fails: AI testing is predictive and not real. So it doesn't account for human emotions. And maybe you have seen that when you had a shader design or when you did like a heatmap testing.

Sometimes I think also for me, it's fascinating to see where people are actually looking at. And this is not something that you can predict. It's not something about hierarchy. Sometimes people are looking at something because of a certain word, because of a certain sign, because of something. So you always need to test with real users. And AI can't really measure the subtle frustration or delight as well as human observations.

So the key takeaway for the testing phase is AI helps with preliminary testing, so something that you do before the real testing to basically have some thoughts, more hypothesis, but real-world usability testing is still necessary. So after going roughly through the design process and seeing how and where we can implement design, we are seeing that AI is not replacing your X design at all, but it's reshaping it.

And for us, it's so important to understand where exactly can we use AI? Where do we make the biggest impact? How and where do we actually speed up the process? Because there are so many ways where AI is actually much better than humans are. Like all the repetition, all the automation, all the data, generating ideas.

And so we can focus much more on like the human skills, real empathy, creativity, decision making. This still comes from human and it's very important to really dive deep into those topics.

By the way, if you would learn more about how to deeply incorporate AI in your design process, make sure to sign up for the AI for Designers 2.0 self-paced version and find the link in the description course. This course helps you to stay up to date, to really get you up to speed with the AI tools. So if you feel a bit overwhelmed, if you don't know how to use them, if

If you still feel like every day there's a new tool coming out, I don't know where to start. This is your course and highly recommend to participate. This is the best investment in your future. AI is here to stay and we can either boycott it or use it very strategically. So for us, it's important to really understand how and where AI can make a difference and how you can use it to become a better UX designer.

Thank you so much for listening. If you would like to connect, you can find us on Instagram. You can find us on LinkedIn. Shoot us a message. Please connect. Please say hi. And if you have any questions, always feel free to reach out. I'm super happy to get to know you, to have a little chat. And thank you so much for listening. I would say see you in the future.

you