Homepage / UX Research Geeks / Thorsten Jonas | The Hidden Costs of AI in UX Research 
small-flowers half-flower half-circle
 Back to All Episodes

Thorsten Jonas | The Hidden Costs of AI in UX Research 

half-circle publisher
Petra Rajkov Petra Rajkov
•  20.11.2026
Share on socials |

Thorsten Jonas, founder of the SUX Academy and co-author of the Web Sustainability Guidelines, reflects on the hidden costs of AI,  from “de-skilling”, homogenized outputs and hallucination risks to environmental and supply chain consequences. Thorsten also explains why we should always remain true to the “brain first, AI second” approach and shares his advice on how to use AI sustainably and ethically. 

Episode highlights

00:01:24 – Defining the place of AI in the future world 

00:04:07 – Best cases for using AI in UX research

00:07:18 – Alternative AI tools to the big tech AI models

00:10:43 – The hidden price of AI 

00:20:17 – What skills are UX researchers actively losing? 

00:26:46 – AI hallucinations 

00:30:12 – Summing up 

00:36:50 – Useful resources for UX researchers and UX designers

About our guest

Thorsten is a digital Sustainability Expert, Designer and Activist, responsible AI advocate and Sustainable UX Trailblazer. He is the founder of the non-profit SUX Network and SUX Academy, keynote-speaker, advisor and trainer for digital sustainability, sustainable UX and responsible AI, invited expert at the W3C for Web Sustainability, author and host of the SUX Podcast.

Quote:   We should carefully evaluate every use case, every scenario that we have, and evaluate if we should use AI for that or not. 

We should carefully evaluate every use case, every scenario that we have, and evaluate if we should use AI for that or not.

Thorsten Jonas , founder of the SUX Academy and co-author of the Web Sustainability Guidelines.
Thorsten Jonas , founder of the SUX Academy and co-author of the Web Sustainability Guidelines.

📍Useful links from the episode

SUX Network: https://sustainableuxnetwork.com

Sustainable UX and responsible AI courses: https://sustainableuxnetwork.com/academy

True Price of AI: https://truepriceofai.org

Thorsten’s linkedin: https://www.linkedin.com/in/thorstenjonas/

Thorsten’s personal website: https://thorstenjonas.com

Podcast summary 

In this episode of UXR Geeks, Petra Rajkov talks with Thorsten Jonas, founder of the SUX Network and SUX Academy, about what AI really costs and who pays the price.

Thorsten pushes back on the idea that professionals must adapt to AI. He argues that instead of adjusting, we should try to push back and actively shape our new reality and what place AI takes in it. 

Thorsten also argues that every AI use case needs a holistic evaluation of internal opportunities and risks alongside external impacts. He describes how the Responsible AI Canvas he developed with Anna Folati can help UX researchers and designers get a better overview of the impact AI will have during their projects. Thorsten also argues that most work does not need frontier models, and that smaller AI models not only cut compute and harm, but helps us stay less reliant on the big tech and saves companies money and time.

Finally, Thosten and Petra go through the examples of hidden costs of AI, from “deskilling” and the environmental impacts to physical and mental consequences for the population. 

Thorsten believes the answer is not to abandon AI but to “think first, brain first, AI second”. 

What this means for UX teams

For UX and product designers

Don’t use AI for discovery and definition. Handing AI the whole development process tends to yield mediocre results that look like everything else. Budget the time to identify and analyze the user problems and prepare original concepts before including AI in the process to retain design quality.

For UX researchers

Synthetic participants reproduce the most probable answer, which is also the most average one, so the non-verbal signals, such body language and the user’s tone, get lost. Use AI where appropriate, but do not substitute the real participants. 

For product and UX managers

Use the Responsible AI Canvas to weigh internal opportunities and risks against external negative and positive impacts of AI before committing to a use case. Consider which model size the task actually needs: smaller open-weight models help you cut compute and harm, and give you full AI cost control. 

For design and research leads 

The de-skilling risk is quite prominent right now, so make sure you protect the skills of both your team members and your users. For the teams, make sure all problem-framing and decision-making happens by humans, preferably with inter-departmental communication. As for the users, give them enough time to try to complete a task first rather than offering immediate AI assistance. 

Podcast transcript

[00:00:00] Petra Rajkov: Welcome to UXR Geeks, where we geek out with researchers from all around the world on topics they are passionate about. I’m your host, Petra, a senior UX researcher and a research advocate, and this podcast is brought to you by UXtweak, the UX research platform for recruiting, contacting, analyzing, and sharing insights all in one place.

Welcome to UXR Geeks. My guest today is Thorsten Jonas, founder of the SUX Network and the SUX Academy, and co-author of the Web Sustainability Guidelines. He has spent over 20 years in UX, and these days he helps companies see what their digital and AI decisions really cost. Most AI talk right now is about how it changes the way we work.

Thorsten looks on the other side. Who pays the price? The planet, people in the supply chain, and us. He’s not saying stop using AI. He’s saying: think first, brain first, AI second.

Thorsten, welcome to UXR Geeks Podcast. You are with us for another episode, and you are right now on the tour. You are visiting a lot of conferences, so thank you for taking the time to speak with us today.

[00:01:26] Thorsten Jonas: Hey, thank you so much for inviting me. It’s an honor and a pleasure to be here with you.

[00:01:31] Petra Rajkov: Amazing.

We will have an interesting discussion today because in my previous episodes, and also in every conversation that I’m having with other researchers, and also with people outside of the field, with designers, PMs, technology guys, everyone is talking about AI and what impact it has on the way we work. But there is more to the story, right?

So that’s what we’ll be discussing today. But I like to start with a hot take from our guests. Something that you feel passionate about, or maybe it’s something that you might find a lot of people disagreeing with. Do you have anything like that? Anything that really drives you?

[00:02:13] Thorsten Jonas: Honestly, there are so many things actually that, that are driving me, but I think what we see at the moment, the discussion about AI, right?

And how so many people are so afraid of that they have to adapt to AI, otherwise they will lose their job next year, et cetera, et cetera. And I think, hey- We should a little bit get away from this I have to adapt to this AI reality, and maybe rather get into something that is more about how can we shape how this AI reality should actually be.

Because what we should not forget is that there is very powerful people, right, big tech people that are shaping this reality at the moment through the narratives that they tell. And so I think my hot take or my passion is how can we get from adapting to a reality that is set by big tech people to how can we shape a tech and AI reality that is good for every being on this planet, and how can we use this fascinating and powerful technology actually for the good of the people and without destroying the planet.

[00:03:16] Petra Rajkov: I do agree with you, and that’s how we try to approach things in UXtweak. We are not against using AI, right? AI is not all bad, but you have to think where it makes sense to use it and where not. And there is a lot of companies that are implementing AI features just for the sake of having an AI feature, and then it ends up not being adopted, not being picked up by users, but it’s already there.

I do see that happening as well, and maybe another question connected to that, how do you decide whether having an AI feature makes sense?

[00:03:53] Thorsten Jonas: The typical UX answer is it depends. Um, and, and that’s- Of course … that’s the hard answer here, right? That it’s– there is no easy answers, but we should carefully evaluate every use case, every scenario that we have, and evaluate should we use AI for that or not, and that includes many aspects.

It includes the obvious ones. Does it help me? Is it a business case in terms of economical costs, right? But also what’s the external price actually? Who else is paying the price, right, being it- The planet being at other people somewhere on the world that have to pay the price, and we might get deeper into that.

So we always need to evaluate by ourself if we think whatever we create with AI is worth the price that needs to be paid and is worth what it destroys. One tool that we use a lot and that I developed with a friend of mine, Anna Forlati from Amsterdam, it’s called the Responsible AI Canvas. It’s a very simple canvas where we look at AI use cases and evaluate them.

We look at, okay, what are the internal opportunities but also risks, and also look at what are the external negative or also positive impacts of this AI use case or of the general use of AI, right? And by this we create and a holistic view on the AI use case and create a base layer for decision-making, which we lack very often, I think, when we talk about AI.

Then we can also look at, okay, I mean, AI is not AI. There are different tools, different models, different model sizes that we can use, and this is a very interesting field, right? When we talk about AI, people most often talk about Claude or ChatGPT, I would say, but we do not need to use the latest frontier models for most of the stuff that we are doing.

We can use much smaller models that need a lot less compute power, so a lot less server capacity and therefore making a lot less negative impact to the world, for example. So there is a lot of things that we can, um, that we can do, and I think it’s important to bring that into the, into our evaluation actually and think about these things.

On the same end, it’s also important to try things, right, and to learn how this works because that means also to understand what works good and what works not so good, right? We cannot only evaluate that from the ivory tower. We also need to dive into it and, and it’s what I’m also doing, right? To see what works and what does not work and to, to understand this.

So yeah, there is no easy answer. There is the UX answer. It depends, but diving deeper into this is a very good approach, I think, to think about AI, though it is the more difficult or more exhausting one, right? It’s not just hacking my prompt or starting my agent, but it’s about let’s think first and then decide if we really should use it or not.

[00:06:37] Petra Rajkov: And also think about those external consequences because we usually evaluate the internal ones within the company. So yeah, I think you already gave us a tip, but you mentioned ChatGPT and Claude, right? So you mentioned the big tech, the biggest players, and along with that is Gemini. So can we also give a shout-out maybe to those smaller models that you like to use, and maybe it depends on the use cases that you use those models for.

So let’s give a shout to those better players on the market.

[00:07:09] Thorsten Jonas: Yeah, absolutely. There are hundreds of models of different sizes of open source or open weight models that we can use, and so there is not just Just the one, I mean, what Mistral does, especially for people from Europe, is a good thing. There is a service called GreenPT.

They are sitting in Amsterdam actually providing like a tool suite and an online chatbot, right? But hosted also in Europe, actually. They are disclosing the carbon impacts of everything that I do there. But there is a bunch of, bunch of models, and what I find really interesting is, especially for us here in Europe, there is a bunch of European hosters actually that offer AI hosting, so they host open weight models in their data centers in Europe.

So fully sovereign, not depending on US big tech, not depending on US-controlled servers. Ready to use, or if you are more sophisticated, you can have your own GPUs actually, right? Your own server capacities to run your models, whatever you need. So, um, a lot more is possible already today. Also for people who are not these super-duper tech nerds, so I’m also not a super-duper tech nerd.

I’m a UX designer, right? That’s my passion and my profession. But there is lots of things that we can use and yes, it’s definitely worth looking into that. And again, what I said before, right? It’s important to dive into that and to try to understand at least a little bit to be able to say, “Ah, I don’t…

Maybe I don’t need to use Claude for everything.” There is alternatives that are causing much less harm, and if that’s not enough, that very often are also the better business case because I can control the costs, right? My passion is how can we create a better world for everyone? But the narrative that I use very often is, okay, but it’s also a business case to do the right thing, right?

Because costs of AI, they will grow exponentially in the next year. And so if you use a smaller model on a s- on servers that you control, you have full cost control. And that’s the argument that you can discuss with your business people actually.

[00:09:13] Petra Rajkov: We’ll be right back after a short break with a commercial message from our sponsors

Hello, UXR Geeks. This is Petra speaking. You might already know that this podcast is brought to you by UXtweak, a truly all-in-one UX research tool. UXtweak handles everything from concept testing to usability checks, both moderated and unmoderated. It also takes care of the analysis with no messy spreadsheets, no frustrating dashboards, or endless reports.

Just clear insights ready to share. So if you are curious, go to uxtweak.com website and start for free. No credit card and no strings attached.

[00:10:09] Petra Rajkov: If there is a person for who the argument is not enough the external consequences, we can also talk about the hidden price of AI, and you talk about it at your sessions as well. So let’s touch on that. If we are using these bigger models and we are using AI for everything they tell us to use it for, to be more effective, more productive, and it will solve this and that, what can happen?

What are the other impacts of going this way?

[00:10:37] Thorsten Jonas: You mean on the economical side or also on the external side? So the, there is different layers of the hidden price, right? Mm-hmm. So on the economical side, I read some weeks ago, I read an article about how highly subsidized computing power for heavy users of, for example, Claude are.

Mm-hmm. So Claude Max costs still, I think, $200 per month, so that’s $2,400 per year. And I read in an article that someone stated, okay, the compute costs for Anthropic could go up to high five digit dollar costs. So he was naming up to 90,000. I don’t know if that’s really true, but let it be 50,000, right?

Let it be even 20,000. You can see the misbalance actually in the calculation, and you can see where the price will go to in the next years. So that’s one hidden price actually, right? But there is many other interesting things. On the first-hand side, there is very interesting research about how the use of AI is changing ourselves.

There is a very interesting term, it’s called de-skilling, and if we use AI the wrong way, it will make us more stupid actually, right? We will lose skills that we already have, and there is research showing that. Doesn’t matter which profession, right? It’s not only a digital thing, it’s also doctors using AI support, et cetera, et cetera.

That does not mean that the use of AI in general is de-skilling us. It means How do we use it? Do I think first and then use AI? So a term that I use very often at the moment is brain first, AI second. Yeah, so if you would be here, I would give you one of our very nice stickers with this claim. But I think this is a very important mantra because if we don’t follow, AI will change us and make us more, more stupid.

And for us as designers, generative AI can create great things, but how or what’s necessary so that AI can create these things? It need to be trained with creative work of humans actually, otherwise it couldn’t produce things. And these humans who created these things that are part of the training sets, they are very often not compensated for their work that is used actually, right?

Last year we have seen this, um, there was a new ChatGPT model, and then there was, I think, Sam Altman in the promo video made a Studio Ghibli style image of himself, right? And then everybody was creating Studio Ghibli style images, and you could see the LinkedIn feed was flooded with these images.

ChatGPT would not be able to create these type of image if it would not have been trained with the creative work of Studio Ghibli, right? And they did not get any compensation. So the creative people who create the work that is used to train actually the models, they are paying a hidden price because they are not paid, that their work is actually used, right?

That could also happen with me, right? If I use Figma for my designs and all my designs are on Figma servers actually, and when Figma introduced two years ago their new Figma Make feature, by default they turned on the option that my designs can be used for training their AIs.

What could that mean? It could mean in worst case that my design is just the future result of the prompt or agent of someone else, right?

Then I am paying the price actually. And then there is another side of the price, and that’s the price other people are paying, right? AI needs, I use the term GPU, so special type of server component actually, right? So because it’s high intense, high performance, so you need this. To build all these things, we need rare raw minerals, and they are mined in big mines somewhere in the world, and this mining is pretty dirty actually.

It’s super bad for the people who have to do that. It’s super bad for the people who are living right next to that. There is super interesting research actually how bad that actually is. One of the biggest mines is called the Bayan Obo mine. It’s in China. It produces so much dirt, being it in the air, being it dirty water, radioactive stuff, right?

So that, for example, the rate of lung cancer is significantly increased in the region actually because of the mine. And once we have mined the stuff, the GPUs and the server chips need to be built. That’s done, for example, in big factories in Taiwan. These factories, again, need a lot of power. Taiwan is still super heavily powered on carbon-heavy power production, so gas and coal.

These factories need a lot of water because you need a lot of very clean water actually to produce the chips. And it happened some years ago that there was a water crisis in Taiwan actually, and then also the factories had to produce less, but rice farmers were forbidden to plant rice because there was not enough water.

So I could say the factory needs to run, but we can’t grow food anymore so that the factory can run That’s another price. And then when we have c-created these chips and we have put them in data centers, then the models need to be trained, and this training data needs to be overlooked by, again, by humans because there could be very bad things in the training data, pornographic stuff, violent stuff, et cetera.

And the people doing this are people in the third or second world. In rural areas in India, for example, right? They are paid quite well, but they have to look at really bad things, right? And they suffer from trauma actually because they have to see so much bad things, actually. So that we don’t see bad outputs of the, of the AI.

They are paying the price even before we start using that. And then when we have the model and we have this, the data center running, actually, these data centers, especially in the US, they are huge. And again, they need so much power. Uh, the Colossus data center of Elon Musk, so the SpaceX data center.

Anthropic is running their Claude, uh, models in these data centers as well. They needed so much power that the local power grid was not able to provide the power. So Elon Musk brought in mobile gas turbines to run the data center. Without permission of the local authorities, he just did it because he said, “Yeah, it’s necessary.”

When you live right next to a data center, there’s very interesting research about what really bad sounds you have to suffer from, right? And then there is infrasound sounds that you cannot really hear, but that your body can actually sense. So people living right next to the data centers, they suffer from these sound problems.

Or since these data centers need so much energy, they influence the price for energy that normal people have to pay. Not here in Europe, but in the US, this is actually happening. Research shows the closer you live to a large data center, the higher is the probability that you have to pay more on your power bill than other people living not so close to a data center.

So that was in very short amount of time, a lot of actors that have to pay a price so that I can use the AI model and do some text or create some nice image or whatever, right? And that does not mean that we must not use AI at all, but it means that we have to carefully think about when do I use it, what’s the true price, right?

The true price paid by others, but also paid by ourselves now or in the future. And what are the alternatives? And alternatives could be not using AI, but alternatives could be also using a different model, a smaller model that needs a lot less energy, et cetera. And in the end, the interesting thing about AI is we do not even know how to produce all the energy that we need to run all the data centers that we are going to build in the next years.

So this problem is not solved Yet, so even if I do not care for the environment, that’s a substantial economical and societal problem. And therefore, part of using AI needs to be how can we use it more responsibly, meaning when do I really need a frontier huge model, and when can I use smaller models, and when should I maybe not use it at all?

Because otherwise we are running into a huge crisis. That, that’s what the numbers actually show at the moment. Sorry to be so depressing. Yeah. I didn’t want to…

[00:18:35] Petra Rajkov: No, that’s totally okay, and that’s why we have you here today because with my previous guest we were talking about how can you use AI and what shift is happening right now in the product teams and how to cope with that, and where does the bottleneck move in the product development cycle and stuff like that.

But there is also the other layer to using AI, so I think it’s amazing that we are talking about it, and we want to give you the space. And it’s so important that you are going to the conferences and you are getting the invitations to talk about this stuff because we all have to be more aware about the impact on not only us and the people who are living close to these data centers, but also for the future generations, and that’s what we see happening when you said the term de-skilling.

We see it with juniors, right? We don’t have so many juniors anymore because they don’t go through the process of maybe learning to design, going from those low fidelity wireframes and then perfecting the design and getting the feedback, and it’s the same for researchers. I think it’s funny to see the evolution.

Maybe two years ago it was like a flex to say that I have created this AI, and now people are actually trying to hide that they are using AI for creating certain output or using it for, I don’t know, writing and all these, all these activities. So when you talk about de-skilling, what skills are we losing Right now when the AI does the work for us, have you seen anything like very specific happening?

[00:20:07] Thorsten Jonas: The most important thing, and that’s down to the core what especially we as UX people probably do is, what’s the problem is one of the most important questions that we usually should ask, right? And it’s a question that AI usually cannot answer. That’s something that we must not lose, I think, because before I can even start using AI, I need to understand what is the problem, what is the problem of the stakeholders, what’s the desires of the users, what’s the needs of the users, et cetera.

So drifting a little bit away from your question, I would like to look on what’s the skills that we need to keep and that also help us keep our jobs. Even with the most powerful tools, it cannot start from scratch. Someone needs to tell the machine what it should do actually, and this thing is much bigger than we think because it’s not only one problem, it’s understanding.

If you are in a bigger organization, it’s bringing people from different professions together, from different silos in the organization together, right? That’s the biggest struggle. AI cannot do that and will not do that. We should be very careful not to shift over to the AI I recently held a talk. I was speaking about the design process and how I think if we look at how does creative work actually happen, the process is super important because it means I start with a problem, right?

And I work with the problem. It’s not about finding a solution immediately. It’s diving deeper into the problem and then thinking about possibilities, opportunities, going in one direction, then maybe changing, going into another direction. So the process is important to get to a better output. When we use AI from the very beginning, we cut out the process and focus on the output only, and that’s the big problem with AI in general, I think.

AI totally focuses on the output, right? You can get your output even faster, but does that really mean that the output gets better? And I’m convinced, no, it does not mean that. Could be in some use cases, but for very good results, you need the process, and AI can be part of my process for sure, but it cannot take over the whole process.

Or if I let it take over the whole process, the result will be mediocre and look like anything else. Because that’s another very interesting thing with AI. It makes things get more and more similar because what AI is in the end, it’s a brute force calculating the highest probability machine, right? I’m asking an LLM something, and then it’s calculating what’s the most probable answer token by token.

And that means is the, the most obvious answers will be reproduced again and again, right? And we will get more and more similar outputs. The strong signals in the whole dataset, they will be repeated all the time. The weak signals, they will vanish over time. So that means everything will more and more look the same, be the same, et cetera, et cetera.

So using a little bit less of AI could mean From a brand side, okay, it could mean that I make my brand a little bit more unique because it does not feels and looks the same like others. Again, does not mean not to use AI at all, but to use it as part of the process, but not letting it take over the whole, the whole process, right?

So besides the de-skilling part that we are losing certain design skills, actually we are also losing quality, and that’s also the super important argument. And coming back to your questions, it could be every skill that, that we lose. When we let AI take over something, then we lose it. That’s what the research actually shows.

And I personally think, but that’s rather a gut feeling, what’s important in design is also this feeling. Does this feel right? And I know we are talking so much about numbers, right? We always want to prove something. But in the end, it’s also about this feels like a good direction, and this is very often an important part of our process, and we cut this out if we let AI do that.

So yeah, there is not a clear line where to use AI and where not. It’s rather I need to decide carefully actually where is it helpful in my process without cutting the process at all. That’s how I actually try to decide when to use AI and when, when not.

[00:24:23] Petra Rajkov: Yeah. I was part of a conversation where one product team was saying that they cut out the brainstorming at the beginning of the product development cycle, the ideation, and they were like, “We ideate with AI now.”

And to me, it’s crazy because the power that I feel when I’m in the room with the people from other departments and we look at the problem, user problem, from different perspectives, and we also talk about risks. Maybe we do not come up with the most polished solution, but it’s solution that works for our user, for our product, for our company, and for our team.

So it’s very different. And we, in UXtweak, we did also a meta-analysis on synthetic participants, and we were going through all of the studies that are out there on the market from different fields. And I want to use an example of when you were talking about the homogenization of results. Mm-hmm. There was a study where they asked users which game would they describe as art, and they ran it with regular participants and then with synthetic participants.

And with synthetic participants, it was like, I think, 60% of people or synthetic people that they said a game called Journey. And then when you looked at the results from real participants, the game Journey was mentioned few times. It was like, I don’t know, 7% of people. I’m making up these numbers, but- Mm-hmm Contrast was crazy.

And it’s just because this game was talked about online. It was very popular. But people, they think in time, right? They remember the game they played most recently. Maybe they were hanging out with friends, and we are shaped by these experiences. And to me, when I saw that study, that was crazy to me. So I think that’s when you were talking about the most probable answer, it’s also the most average answer.

You don’t get these wow moments that are happening at the interviews and at places like that. So I think the power of people is so strong, but sometimes we forget about it because AI is just out there. It’s right here on your computer, and maybe it’s, it looks too tempting. So, yeah.

[00:26:36] Thorsten Jonas: Yeah. There is a nice quote, and that’s already two or three years old, by I think it’s John Herrman, and he said that in New York Magazine.

Don’t know. He said something like, “ChatGPT is now better than ever in pretending it’s something that it’s not.” And that’s what AI is very good at, right? It pretends to deliver very good answers to whatever question we ask, and it needs time actually to look into to discover if these answers are really good or if they are not.

And there is a pretty significant probability that they are not so good because in the end they are very often average because it’s reproducing the most probable answer to whatever we are asking. And especially in user research, I think if you do a user interview, it’s not only what the people say, it’s what’s the body language when the person’s doing something, right?

How is the look? It’s these nuances.

[00:27:23] Petra Rajkov: Nuances.

[00:27:24] Thorsten Jonas: Yeah … right? What a friend of mine called the weak signals actually that are especially interesting in research, right? Again, that does not mean that virtual or synthetic personas cannot be helpful in certain cases, but they cannot replace the other thing.

That the dangerous thing there, right? That when people think, yeah, we can replace it because as you said, in best case, we just get average answers and it can get even worse, right? We did not speak about the whole hallucination problem. Mm. And that is a significant problem, and it’s not getting better with bigger models.

The opposite is the case. The bigger the model, the higher the probability for hallucinations. That does not mean that it always hallucinates, but we can see that. I just had one very famous case that I use in my talks at the moment is from KPMG consulting company, and they did a report. Yeah. It was called Redefining Agentic Experiences, something like that

And so it was a report showing 16, 17, 15, something like this, agentic AI use case in the real world of- clients of KPMG. In every of these cases were mistakes and errors, and two of them were totally made up. They were totally hallucinated, right? That’s what you also can get when you just let AI do things and do not check the stuff actually, right?

And there again, it’s not only that you get average, it’s worse. It’s becoming a business risk if we use it the wrong way, and that’s also important part of the whole equation, I think, that is very often not part of the discussion, right? It’s always, yes, I can let AI take over the process, and then this process is much faster, but how good is the result, and how much time do I need before I let AI do something to prepare properly that it can work, and how much time do I need afterwards to check all the results?

And that’s n- very often not put into the equation, and that’s not fair if we discuss the true price, economical price. It’s also important to create much more transparency about these things. That’s what I experience, right? Because most people have no idea, and I’m not blaming anybody for that. I can totally understand that because the overall narrative is so much about how great AI actually is that even people who should know better do not look behind the curtain, do not learn and understand how does it really work and what does it really mean to look at the results that I actually, actually get.

[00:29:41] Petra Rajkov: Okay, so let’s summarize a little bit what we were discussing or what is the takeaway that we can share with the listeners. What can a researcher or a designer do tomorrow when someone says, “Let’s add AI to the process, to the product”? You already mentioned the canvas, right? What else? Yeah.

[00:30:03] Thorsten Jonas: I think at first it’s a mindset, and the mindset is not only about AI, it’s about for us as designers, and that’s how my journey, and I don’t know if people see the video, they can see behind me our SUXl logo here and how the SUXl network actually- started with the idea that whatever we create, whatever we do has an impact to the surrounding ecosystem, to the surrounding world, and that’s very often not part of whatever we do.

If we look at UX tools, as long as I do UX, and I do that for 20 years, more than 20 years. Damn, I’m old. So I do it for a long time.

[00:30:37] Petra Rajkov: You don’t look it, so that’s fine.

[00:30:40] Thorsten Jonas: Thank you so much. So as long as I do this, I remember we as designers, we always fight for the user, right? Yeah, let’s consider mo- the user more.

Let’s bring the user into the center of our doing, and we, we were fighting these battles against, I don’t know, marketing, business people, developers, everybody. But if we are honest, this is also a pretty exclusive view because if we put the user in the center, what about all the other actors in the system?

Mm. And that’s what we talked about when we spoke about the hidden price of AI, right? And it’s not only AI, it’s with everything that we create. When we center around the user, that’s a pretty exclusive view actually, and it’s very often harming other actors in the system. Very often someone else needs to pay the price for the great user experience, for the great convenience that I create for my user.

So the first thing is this mindset shift. Thinking about having this systemic view on whatever we build, whatever we do, who else is impacted by what I am doing. That’s important. The responsible AI canvas, especially for AI, is what I find, and I did a lot of workshops with different people from different professions, super helpful to create this holistic view on, okay, how good or bad is this certain use case?

Or also let people understand, hey, I have no idea. I do not really know. Being it also internal risks or opportunities, right? Is it a feeling that it’s good, or can I really prove that it’s good, whatever comes out of the machine. So that’s a pretty helpful tool. We have a bunch more tools actually in the SUX Network.

They are available for free. You can just download them, especially for designers where you can get more understanding about who is impacted by a system, user journey mapping, actor mapping, needs to consequences, right? So how are user needs creating negative consequences on other ends? So that’s all– that’s not AI only, but that’s important to, um, before we design everything to create this understanding.

Then what tools do we actually use? I think we as informed designers need to be much more careful, especially which tools do we use. At the moment, everybody is using the US big tech tools. I think this is super dangerous, and this is also, again, it’s a business risk because we are highly depending on them, and if someone says, “Oh, the Europeans did not behave well, let’s switch this stuff off.”

Right? And that could theoretically happen actually. Or if my data is stored on servers of these people, it’s not safe because if the US administration says, “I don’t want to have this data,” and it’s a US company, they have to hand over the data, no matter if the server is located in the EU or not, right? So we need to, um- Be aware of these things and make decisions of, “Hey, is there another tool?”

As we said, can I do it with another, with another model, for example, right? Can we use our own model at a European hoster that does the same stuff? Is there an alternative for Claude Design? Have I tried this? There are alternatives. They do not work this well, but maybe they are good enough for what I want to do.

So being careful with what tools do we actually use. Then in the process, how do I use AI in my process? I can totally understand that it feels so nice that I can say, “Okay, I just let AI… Yeah, one prompt, one agent, and now here is a design system. Do whatever what I tell you you should do,” right? And yeah, I think we should not do that because as we said before, we are losing our skill to design, and that does not mean we have to design everything on our own, but we still have to understand what is going on actually, and that includes what’s super important, and that’s not new.

The work always starts with, with pen and paper, I would say, right? With bringing people together on the table, as you said. Very often it’s about bringing different people from different professions in an organization into a room because they never talked with each other before, and it’s people like us who say, “Hey, if we want to understand the problems, the needs, we need to bring these people together,” and then we can use AI afterwards to create a prototype, right?

If we have understood the problem. So embrace the process and embrace the problem understanding is super, super important and crucial, I think, for the future, and also crucial for good results and for creating great products, great designs, great experiences. And then if we use AI in our process, we need to calculate the time that is necessary to check Whatever we have created with AI, and I have seen that by myself.

How often have I got, I don’t know, summaries from online meetings, and I look at the summary and I can see that two very important topics were not part of the summary because the AI missed it. Again, right, there’s the interesting weak signals. Being it on our own or being it in discussions with other people, we need to consider this as well, and we need to put this on the table as well and say, “Yeah, we can use AI, but then let’s be honest about what’s the true time savings that we have if we bring into the calculation the time that’s necessary before we can start using AI and the time that’s necessary to check whatever my prompt or my agent has created in the end.”

And one thing not to forget, when we design for users using agentic AI, we also should not de-skill our users, right? So very good mantra there is let the user try to do something and then offer AI assist afterwards instead of offering AI assistance from the, from the beginning. We let people still be in control of the process.

So yeah, everything that is important for us is also important for our users. Every harm that we could do ourselves, we could also do our users, and we should be very careful there as well. And I think, again, the mantra of we should not cut out the process But we should embrace the process and see, okay, where can we help with AI in the process?

That’s the right mindset. Maybe the most important things to do. Yeah.

[00:36:41] Petra Rajkov: Okay. That’s, I think, very useful, and I’m really glad that we had this conversation. And you mentioned the responsible AI canvas. We will also link up the resources that you have from SUX Academy, and maybe just to also share where else can our listeners learn from you?

Where else can they find you?

[00:37:02] Thorsten Jonas: Ah. Thank you so much for mentioning that. I’m so glad. Yeah, that’s what we do at the SUX Network. We have the SUX Academy. That’s where you can learn sustainable UX, the SUX tools, but also there is a course about responsible AI, sustainability, and ethics in AI. There are courses for individuals, so I usually do that every few months.

Next courses will be probably in November, I think. We also offer that for organizations, so, and go into organizations doing that. I have to say that’s the only thing that’s not for free. So everything we do at the SUX Network, all of our events, everything is for free. The courses are the only thing that are not.

They are also helping to finance, actually, all the work that me and many other wonderful people do there. Yeah, and there you can learn to be a responsible designer with the tools that we have, and also to use AI more responsibly and to understand the negative impacts and how to use AI more responsible.

Would be happy to welcome some of you there.

[00:38:00] Petra Rajkov: Amazing. Thank you so much, Thorsten, and thank you for raising awareness around using AI and what consequences are connected to that. And yeah, thank you for being here.

[00:38:11] Thorsten Jonas: Thank you so much for having me. It was a pleasure and a real honor to be here with you.

[00:38:20] Petra Rajkov: Thank you for listening to UXR Geeks. If you enjoyed this episode, please follow our podcast and share it with your friends and colleagues. Your support is really what keeps us going. 

If you have any tips on fantastic speakers from across the globe, feedback, or any questions, we would love to hear from you, so reach out to geekspodcast@uxtweak.com.

Special thanks goes to my colleagues, to our podcast producer, Ekaterina Novikova, our social media specialist, Daria Krasovskaya, and our audio specialist, Melissa Danisova.

And to all of you, thank you for tuning in.

💡 This podcast was brought to you by UXtweak, an all-in-one UX research tool.

 

Read More

uxcon special with Aurelia Alston

Aurelia Alston leverages her experience as a Product Researcher to drive impact in both large corporations and startups, emphasizing scalable research practices.

Noam Segal | Garbage metrics: the case against NPS

Noam Segal discusses why NPS is a flawed metric and why companies should stop using it. He explains the issues with NPS from a survey science perspective and suggests better alternatives for measuring customer satisfaction and engagement.

Improve UX with product experience insights from UXtweak

Test your assumptions quickly, access broad and qualified audiences worldwide, and receive clear reporting of findings - all with the most competitive pricing on the market.

Try UXtweak for Free
Improve UX with product experience insights from UXtweak