#102 AI, Ethics & Innovation: How Companies Can Integrate AI the Right Way with Zamina Ahmad
57:13 Share

Welcome to the future of UX. I'm Patricia Reiners and today's episode is all about one of the most exciting and complex topics in tech right now, which is artificial intelligence.

To help us navigate these ever-evolving landscapes, I'm joined by Samina Ahmad, the founder of Shades & Contrast, a consultancy specializing in fair and responsible AI solutions. With over 15 years of experience in data analysis, data science and machine learning, she has a unique perspective on how companies can integrate AI in a way that is both innovative but also ethical.

And in this episode, we will dive into the AI hype train, what's real and what's just noise, how companies should approach integration without falling into common traps. We're also going to talk about the three level integration framework that helps businesses scale AI effectively.

And we will talk about the challenges of responsible AI and how to ensure fairness in AI-driven decisions. Plus, we will explore the future of AI from automation to AI-driven innovation and what it all means for UX designers, for product teams and for businesses. So if you've been wondering how AI will shape the future of work and design and business, this episode is for you. I would say let's dive right in.

So nice to have you and thank you so much for taking the time. Welcome to the future of UX. Hi, Patricia. Thank you for the invitation. Looking forward to our conversation. Me too. And we will dive into, I think, our all favorite topic right now, which is AI. AI!

I think it's our favorite topic since two years or three years. I don't know. 100%. Everyone's talking about it because it's also fascinating. I am still fascinated by all the things that are the Chechi Petitas for us or Midjourney, all the videos that you can create with AI. I'm still feeling a little bit like a child. It's crazy. Huge fascination. But also what I saw last week, there was a GIF

the AI hype train is rolling again and it was about DeepSeek and Stargate and blah blah blah whatever is in the news again so there are always rolling news about AI and it keeps rolling the hype train. Yeah DeepSeek too this was like a really this really is a big thing right and it's really big pushing

Also, like all the other large language models to get even better. And even all the features that have come out since then from TGPT, like the operator and the deep research. Okay, I don't want to dive too deep into the content yet. Before we do that, I would love you to do a super quick intro so people know who you are and what you do.

Okay, yeah. So my name is Samina and I'm the founder of Shades and Contrast. That's a consultancy for fair AI solution, responsible AI solutions. And my background, my personal background, so that why do I do it, is data analytics, data science and machine learning product manager. So over the last 15 years, I

crunch data at first, consumer data. Then I developed algorithms out of it as a data scientist. And then I moved to product manager role where I developed solutions. And that brought me when I founded my own company to this AI stuff. Yeah.

So cool. I think especially focusing on the responsible part and the ethical part, this is so much needed because a lot of solutions, unfortunately, don't really know how to tackle the problems of responsibility solutions and ethical parts. So we will definitely dive a little bit deeper into that.

But we already touched a little bit on the AI hype train. So all the companies right now, not all, but like most companies right now are trying to integrate AI in their products. And I know that you are doing a lot of like workshops and study sessions with different clients on how to do that. I would like to jump right in with you and ask you, how do companies do that? Do you have some tips and tricks?

So most of the companies, they are sure about that they want to implement AI, but they lack a specific approach or specific topic or specific use case. So they are hyped by the market and say, okay, we need to do that as well. What they lack is specific knowledge.

solutions. So do they want to integrate ChatGPT as company tool? Do they want to automatize any process? Or do they want to develop an algorithm or model for, I don't know, predictions and marketing for budgets or something else? So there are very different use cases out there and also different approaches. The first

tip to be clear on what do you want to tackle first, AI tooling, JTBT, or modeling, developing machine learning models with your own data. That's a totally different stuff. But I think most companies don't know yet. They feel like let's integrate AI, but they're like, okay, what are the advantages and disadvantages of both? How do I decide what is the right fit for me as a company? Yeah.

And I think since chatGPT, everyone is thinking about AI that chatGPT is AI or cloud is AI. So any other tool of AI tools out there is AI. But actually you have so many different possibilities. So if you look at how we developed or the natural evolution of AI is actually machine learning and many machine learning systems can solve problems for you. What you need to understand is that something you want to invest

because you have data, you have structured data and you can build an innovation inside your company with machine learning systems? Or is the question, how can we integrate ChatGBT and improve our productivity? So are we talking about cost reduction? Are we talking about automatization? So there are so many different levels and on the market and discussion I observe, I see that people are not structuring this

So what is your goal? What is the technology? Is it a low cost technology like €20 for ChatGPT or is it a data science team developing something for you? So yeah, I think that makes total sense, right? Like productivity or is it like a tool that you want to build inside your product? Let's say it's not productivity, but it's a tool or it's like an AI integration.

What do you think? What would be a good process to get this started? So it's a tool, but it's not about productivity. You have an example? Yeah. You just mentioned that there are basically two ways to use AI. One would be productivity. So to use the GPT, for example, like to learn how to be more productive, replace a repetitive task or create your own data models.

So imagine like your client comes to you and wants to establish their own data models. So what could this be? Example, I don't know, automate certain. They're working on like a data project and they want to automatically show like the most relevant data and basically analyze that already for the user. So like a bit complex, like not sure if the example makes sense, but something like that. They have a rough idea.

What would be a good way to get started? Maybe they are not really sure about this idea, but they think it could be. I don't know. For some things like that, I would suggest to start with the impact effort matrix, right? So like you also know from product management tooling, that is the impact you expect worth the invest you need to take, especially in developing your own data models or machine learning systems.

let's say we want to build a prediction for marketing budget. So is your marketing budgeting very hard? Do you lose money here? Do you have very different customer groups and you want to optimize it on customer level? So maybe it makes sense. But on the other hand,

If you have only 500 clients, for example, they're on a local basis, you don't invest in paid ads, then it doesn't make sense at all. Then you can solve such marketing problems with AI tools, which are out there since two or three years, very low budget on low budget, but they can also help you

to be better in your marketing campaigns for, I don't know, local newspapers, generating text. You also know it, generating visuals is easier with AI tooling. I think it depends on company size and the problem, how big is the problem you have, and that determines the solution you want to develop.

People don't see that. That's a big issue. They have high expectations concerning AI. So you develop something maybe in six weeks and then it solves everything, replacing every human.

That is not the case, not for AI tooling and also not for machine learning systems or modeling. At the moment, we are not replacing humans at all. We are supporting their work and humans need to be involved to

evaluate the quality, the output. Can I trust that output? Is that high quality for our brand, that marketing output? Can we invest in that and something like that? And I think, yeah, high expectations are one of the big struggles and challenges I need to face and solve with clients.

I can imagine. Also, the expectation management is such a challenge because it takes so long to build those models. You need to iterate so much and you need to start with an MTP version that's so low. I think this is something a lot of people don't because they don't have experience. But this is a bit frustrating because it takes so long to train those models and it's so cost effective as well.

But on the long term, if it really makes sense, then it saves a lot of money and also brings you a lot of exposure. Maybe it depends on the product, but a lot of opportunities.

It's very interesting to see a different gap here in the industry. So data modeling and stuff like that are very a topic for corporates, and they invest in that since maybe 10 years when I started with data science departments. And I think we both know that from tech companies as well. We have data scientists and developing something out of the data, something very useful. But that's not something for the big markets, for

small, medium-sized companies, that's not a topic at all, most of the cases. And for startups and NGOs, they also don't have the expertise or the budget to invest in there. For these companies, we see a lot of this AI tooling and then

I also have different questions when I consult NGOs, for example. They believe that AI is something magic. So you have to explain, oh, it's mathematics and it's probability calculations and stuff like that. So questions you need to solve and explain are very different from different target groups. That's also something very interesting for me and also where I need to switch languages a lot.

I can imagine because there are a lot of things that are difficult to understand, especially when it comes to machine learning and

Even all the different words or explanations to get things right and also to help people to understand very complex scenarios is a big challenge, I feel. Also to align and collaborate with teams. How do you usually do that? Because I can assume that when you do workshops or work with clients,

it is a big challenge to get everything aligned and speak the same language data scientist and a designer or a data scientist and a product manager or someone from hr maybe so speak the same language we have some tips on how to do that yeah imagine or make research about your target group so i do that a lot about for my clients so i do research if i consult an

NGO, I can be very sure that they have no clue at all what is happening in a machine learning system or what is an LLM. This wording they are asking, what is LLM? I never heard about it. Or what is prompting? They never heard about that wording. I need to explain basic words and definitions, give a glossary, for example. So the language is different from clients who do your research beforehand, and

And then when I'm in a tech podcast like here, I can assume everyone knows what the data scientist is doing, some bit of it, and what machine learning models are. So yeah, that's one tip to

I think the hardest part is for all these big mass of companies which are in the middle, all these medium sized companies in Germany. They have no data science or tech experience, not so well, but you always have some people which are very early adopters in AI and they did some research and they are very

And then they ask suddenly questions I never thought about it. Last year, it was a CEO asking me, yeah, I did that with Notebook LLM. And then I had this bug. Can you help me to solve it? And I did not expect that he was doing something with Notebook LLM with his papers to make a podcast and then listened on it in his car. So that was quite funny.

Fascinating. Yeah, also to see what people are doing with the tools and I think super interesting. Very different. Yeah. That's a beautiful part actually, right? So AI is since chatGBD democratizing in a way, right? So the access to it is so low key, everyone can do something. It depends on your, I don't know, personal affinity to tech maybe? What would you say? What did you observe?

100%. And this is what I love about AI. It makes things so much easier for people. Because even if you don't code, if you don't understand AI, you can still use the large language models. You can build your own custom GPTs without writing a single line of code. Just by typing in. Just by using basic language that

that you send to your friends via WhatsApp. I think this is absolutely fascinating for me. And also with all the new tools that are coming in, like all ChhpT for example, like deep research, really diving deep into topics. When you're someone who's, I don't know, maybe you have ADHD or something and you really have a hard time focusing and concentrating, ChhpT does these things for you, present them to you so you can focus more on the creative part where you're really good at and shine.

and outsource things that you are not so good at that you don't enjoy even, right? So I think that's fascinating for me. And I really love that you mentioned democracy.

democratize democratizing i'm sometimes i'm so bad with what democratizing like knowledge and education and like doing things i don't know but i'm always also so fascinated by like those youtube videos where people share like their workflows that automized their whole workflows or stuff like that and it's so simple so i think ai is

fascinating and opens up so many opportunities for designers also for like self-employed people for basically everyone to build things that you couldn't have done before because like you didn't have the hard skills

But now you just need to know how and... Oh yeah, I totally agree. That's so true. You know what? When I had a project, I think two years ago, ChachiBT was on the rise and in that project I needed my Python skills and I'm not so good in Python, but ChachiBT was there and it helped me to improve my Python skills. So I was so happy to code better lines and

Then in that project, they integrated a data tool, Databricks, and this tooling has now an AI assistant. So I could also code in that tool with that AI assistant and it helped me to, okay, that's not logical, Tamina, or here's a bug, Tamina, and I loved it. I think that's a lovely example. That makes so much sense. Like you still have a skill somehow, right? You know how to code in Python.

But you're not so deep into it and you don't need to because you have tools to do that for you. You just need to have the critical thinking and the oversight of judging the result and then tweaking it a bit.

Yeah. I still want to do, you know, PyLadies, the event series that's for women in Python. And we all, yeah, we still need to do that event where we show how to improve your Python skills with prompting. And I think that's a beautiful skill and gives you a bit more access to programming.

Yeah, that's so cool. And this is really opening up so many opportunities for different people, especially for designers who are focused on like doing design work. We're not so good at coding or business. But now you can be right. Like you have everything at your fingertips.

That's absolutely fascinating. One thing that I definitely would like to talk about is your AI or the three level integration framework that you developed. Could we talk a little bit about what this is, how you came up with it and explain a little bit of how it works? Yes. So we touched a little bit on that when we discussed now.

And it was the observation when I observed my project I'm involved in that they are very different. So in one small company, I'm doing an AI essential training in AI tooling. And in another project, I'm again a data scientist crunching data and doing some modeling stuff. And I was thinking, okay, that's all about AI. So in one part, you are developing something and in the other part, you are just integrating tools for your productivity.

So I thought, how can we bring that into a framework and show that to companies when we consult so they also understand there are differences when you do something with AI.

So this three-level AI works provides a step-by-step approach for your own AI journey. And the first step is iTools and your individual productivity. So it's a starting point. It's a low budget and also low access starting point, right? You can just register at ChatGPT for free. You can start prompting.

You can generate text, you can add mid journey for visuals. There are so many AI tools out there. So prompting, I think prompting will be a skill, which is like Excel or Office or Outlook skills in the future. It will be, everyone will be doing it and we will use, everyone will use AI. And if you look at the market, I think the last

past two years were mainly characterized by that. Everyone individually were using tools, AI tools for their own use cases. And it generates quick wins and improved time management and yeah, increase your productivity. So that's the past two years. And it's, I think, a good step into the AI world for everyone.

Then I observed, okay, now people or companies want to go one step further, right? They don't want only people to have prompting here and there for their task. They want to automatize business processes. And if you look at the big tech players, you also see that with ChatGPT's Operator or Microsoft Co-Pilot.

They are now trying to let systems talk to each other, right? So co-pilot should read my emails, should analyze my Excel and then prioritize my task on my Trello or whatever. And operator should also do the same thing. So there are things in the market, features and solutions that are out there.

automatizing business and team processes and that's the point now and that's also the challenge we are now talking about not individual solutions we are talking about changing teams team works team processes and technology here at this point is the easiest part to be honest it's more now that

You cannot drive that with only having early adopters on board or enthusiasts. You need also the silent majority, which maybe not used AI so far or maybe is skeptical. You need to convince them now because you cannot change business or team processes if not everyone is involved.

I think the next two to three or four years will be very interesting to see which companies will make that change happen and which companies not. And yeah, and also I'm also interested if this big it's like a competition between this big tech players, right, with Copilot and

operator, which system will win, which will give the best features, which will give the best integration. I think Copilot has a big advantage because they are providing all these Office products for corporates with Microsoft. But that's the only advantage so far, not on feature level. I don't see anything so far, but yeah, it will be again a race of the big tech players here.

That's the second approach, second step in this AI framework. And the last step is totally different from the first two. It's not about cost reduction or automatization. It's about developing your own AI solution.

And that's about innovation. So maybe you have already data in your big company or small company, whatever, but maybe you want to develop something new, something which gives you access to a new market, gives you access to new clients, and then you invest in something. And that means you build expertise.

with your data, you build expertise and maybe connecting with LLMs, open source LLMs or whatever you want. It could be also an easy machine learning model. But what I'm talking about is building AI driven innovation so that you are at the forefront on the market and

At the moment, I don't see a lot of companies moving into that direction, but there are, for example, looking at computer vision systems with autonomous driving, that's also AI systems, or in the industry, looking at robotics systems. It's also very innovative things. Or maybe also I have big hopes in the health industry.

predicting diagnosis with computer vision data for cancer, for example. And that's a three level AI framework. And I'm just trying to navigate and make that visible to people so that they also understand differences in AI and tooling and also in machine learning systems.

Yeah, I think wonderful. That makes so much sense, especially all the examples that you mentioned with like health care and like cancer recognition and these things. So much potential. But why do you think are only like a few companies really like exploring or going really deep into these opportunities? What are the obstacles at the moment for companies?

So you mean the last part, right? The innovation part. Why is it? Yeah, it's actually you need investment. You need money for that. And you need a little bit lower expectation because why? When you develop models, we are actually talking about gathering first data. You need data, a lot of data, right? First thing, do you have data? Maybe that's the first obstacle you need to solve. Then second

The second one is you need to train and test modeling. So you need a person or a team who is able to look at the outputs and then adapt the parameters, the data, make it high quality output. So for example, in most of the models, 80% accuracy doesn't make sense. That's not good enough. You need 95% accuracy, for example, for autonomous driving, right?

You wouldn't send a car which is only adaptable for 80% on the streets because you wouldn't trust that car. You need a car which is safe for 95% at least, for example. But it depends on the industry, like you see. And that's the second obstacle, so train and test.

And that involves a lot of time, budget and patience. So I think you need to understand does it really make sense, especially with the big tech players pushing out tools for free or for low budget, then maybe you don't need, you shouldn't invest in that so much because there's a big probability that you fail.

Yeah, and maybe that you can use the resources that other companies are using for free, use an API and maybe integrate them in your own tool. That's not always possible, but if that's like a cheap version, you don't need to do much. Could be something, right? But when it comes to

a really big challenge. So imagine you decided to get the funding, get the investment to build your own AI models. You know, you got a little bit of money. You spend some time iterating and working on it.

And then the really big question pops up. Is this all like responsible AI? Are we going the right way? Is it like all ethical? And then the team is, we haven't really thought about that. Could happen, right? What should this team do? So...

at first relax because yeah because many systems are not you don't need to think about responsible ai right so for example let's talk about marketing algorithm yeah maybe there's bias inside but then that's it so why because it's not affecting my life so not so much maybe i'm

getting advertisement which is biased and stereotyped and I don't like it because that's my preference. Okay. But it's not that I'm not getting a false diagnosis. That could be something where you need to look at responsible AI. So are you in the health or medicine sector?

Thumbs up. That's something where you should look at. Another example I made already is autonomous driving. That's a lot about safety. So is the car recognizing all weather conditions?

Is the car stopping when it sees humans, for example? So responsibility is very much dependent on the case and the industry you are working for. If you work in the health or medicine sector, you already should have looked when you gather all the data into it, into bias, for example. So is your data representative enough to make a prediction for cancer, for example? Do you have enough

female data? Do you have enough person of color data? If not, how can we de-bias it? Can we weigh it just the way this and say women are 50/50, people of color are differently, maybe something like that. So there are different solutions for that. And it also depends on the market. So health is something where you should really think about

responsibility in advance. Autonomous driving is about safety and we have already regulation in place. So people working in an automotive sector are very much aware of it. Another industry is banking. So imagine we two go bank and we want a credit.

Maybe I get bad credit because I'm living in Hamburg and you live in Zurich and that's better paid. So we get different bank credit volumes. That's fair, maybe. But maybe I also get different

a little bit lower than the average person in Hamburg because I have a name which doesn't sound so German. So that's not fair. There are different things you need to ensure. But I always recommend, you maybe know that, Responsible Tech Playbook from the United Nations, developed with ThoughtWorks. I really like that.

If you do discovery and solutions, you should have a look into it. That's the best thing. But yeah, relax is my first point and make yourself aware which industry you are working on. So is it an online shop? It's a totally different topic than a medicine sector. Yeah.

100% and integrating that early on is helpful, like thinking about like the biases, potential pitfalls, the examples that you gave with the banking, getting a loan. Yeah, that could be a big problem and it depends on the data that you have. So some topics that you definitely need to tackle. And I also really love the resource that you mentioned. I will link it in the show notes so people can check it out.

and maybe use it for their own workshops, collaborate in their team and find really good solutions for their products that we don't have a lot of biases. There are still biases, right? Even we are all biased. Totally. But we need to try to reduce it as much as we can to have fair AIs, basically. Yeah, or unbiased AI is something very idealistic, right? It's already almost impossible, but...

we should think about the output and if we are not strengthen the bias. So we shouldn't create unfair results, for example, in the credit volume. So if we two earn the same and live in the same city, then we should get the same credit volumes and I shouldn't get be biased because of my name. And that's the thing people should test, for example, for their product.

I think even that we are living, that we are not living in the same city. I'm wondering if this should actually make a difference in the loan. I don't know, maybe because of tax reasons. Yes, then of course. But if we're making the same money, it doesn't change like our ways of paying the loan back, basically. So I'm wondering like all the questions that you need to think of, does it really make a fair difference if you live in Zurich and Switzerland and Hamburg?

Is it something where you need to differentiate? Or is it something where you feel like, no, it doesn't make a difference? Is it tax-wise? Okay, I also don't know. No, it's a really valid question. And fairness is not so easy answered, right? So yeah, I would say it doesn't make sense to have different... No, I think when we earn the same, then it's fair to get the same, right? But if we earn differently, but maybe the system doesn't know, let's assume that,

then it would assume probably that in Zurich the income is higher than in Hamburg. And that's valid, I think. But it depends on the information the system has. Yeah, yeah, yeah.

I think this is so fascinating because it really it also shows like how difficult it is also for us human to find good answers and that there's a lot of discussion needed in a team as well to align and to think about what are our values and it's not something where there's this like

10 rules to follow and then you're right, but find it out for your needs and what you mentioned, like your industry, it's very different, like marketing, healthcare, like someone dying from it, not like, can you make mistakes? How horrible would it be if something goes wrong? In like lots of lots of discussions and lots of things that companies need to think of. So when you think a little bit about the future, so let's say the next five, maybe 10 years or so,

Where do you see AI headed? I know that's a very difficult question because we all don't know. It's like looking into the glass ball. But from what you've seen in the past, how do you see everything progressing?

Last year, some months ago, if you would have asked me, I would have said something like, yeah, agentic AI, all these rack systems, for example. And I still believe that's one thing and one perspective, agentic AI, process automation, blah, blah, blah. But if you look at the news last six weeks, since we have this year, it feels like

totally strange and unpredictable. And if you look at, I didn't see coming the Stargate investment from Trump and this big investment, having all these tech people on his first day, what is it called again? Inauguration day or something. And you see that economic and AI will be a strategic advantage for Bitcoin.

big countries like US. That was the first thing. And then weeks after that, deep sea came around the corner. I didn't see that coming. And just proved that maybe AI doesn't need so big investment and you can do it cheaper and better. But it's also...

with political answers, right? So it's not answering anything about China, for example. And what about the European Union here and the role in the world? So I think there will be a lot of things which are unpredictable. And I think

I believe in the EU AI Act. I know that people in the tech sector are criticizing that. But actually, if you read and inform yourself that a lot of AI systems are not affected by the European AI Act, it's like I said, most of the people can relax about this regulation. But if you're

The regulation about is to regulate AI in kind of way. And people are criticizing that's stopping our innovation, that's stopping our speed and US doesn't have it.

But US has it in California and we all know that Silicon Valley is in California and they have a regulation there. And the regulation is not so strict like people are trying to picture it. It's just like I said, if you develop a marketing system, you are not affected by this regulation because it's not affecting humans' life. So it's about human lives and

how you affect with your AI system. If you're doing a credit volume score, then you need to be transparent. So why is the credit volume higher for you and me than for another person? So it must need to be explainable in kind of way, right? And that's a thing I actually like because it gives you the mindset that you need to think about what are you developing and what are you pushing out in the market and

I hope that Europe is taking the role of this responsible

AI leader, but of course I know that a lot of bureaucracy is going on here and regulation are far behind what is happening in AI. Even all the AI courses I was recommending some months ago, you cannot recommend any longer because they are outdated. So it's so fast. Like you said, you need to look into YouTube and what's the latest YouTube video from a person because you cannot

Yeah, there's so many updates. I was losing track on your question, but... You just asked a lot about what the future looks like. But yeah, I think it's interesting because it's not so easy to answer that. You need to give a lot of context and also think about what's happening at the moment. And there's so much change happening, right? What's your prediction?

I think a lot is going on, which we will see a lot of surprises because we have six weeks now in this year and we have already many surprises. I think it's continuing. What do you think about AI and the future?

I 100% agree. The year started really strong, a lot of surprises. I think DeepSeek now even pushed the race even further. I think it just started the really big race. Also for all the other large language models, OpenAI, they're seeing this, oh, there's an interesting competitor, there are new opportunities, let's put more money, more effort in it.

And quality is more important than money, I think, in this race, especially for the US, as we have seen in the last weeks. So I'm seeing a lot of interesting advancements coming. And I feel for us as designers, as people working in tech, we need to stay up to date. However we are doing that. If we are doing courses, if we are watching YouTube videos, if we're doing workshops, we need to stay up to date. Otherwise,

Other people are doing that for us and they are happily taking our opportunities. So I think staying up to date is essential. And of course, I totally agree. I have a course about AI for designers and it's crazy. I'm constantly updating it. It's something like so much effort goes into that because you can just create something and then leave it.

Every month there's something new coming and you need to update it. It's crazy. Yeah. And but that's the way how it is. And this will even get faster. So let's see how things are going. I have no idea how the next five to 10 years will look like. I assume a lot of automation, a lot of agents, a lot of automated workflow so that you can really automate whole things as we have seen with the operator. Fascinating to see that you basically watch it

book a flight for you, do groceries for you. So you need to, I think, also change your perspective on skills and expertise a little bit. Did you see Perplexity's approach of the future of e-commerce shopping? That you can say, I didn't test it, but I saw the video, that you can say, okay, I want to redesign my bedroom in Scandinavian style.

And then he curated, okay, Scandinavian looks like that or like that. And then looked for furniture, painting colors and everything. And it was totally curated and inspirational. I was totally amazed by that. Did you see that?

I think I saw it somewhere, but it was like a time ago, right? Like not too long ago. Let's link it in the show notes so people can check it out. Yeah. It's fascinating also to see what these tools are capable of doing. So this is, I think also a good example because it shows maybe the things that we are doing are not the right things and we need to think about even different workflows to get different results, even better results. Yeah.

It's some time ago, that's true. It was, I think, in November last year when I saw it. But why do I remember that? That's also a good point. All this productivity stuff is not amazing me so much. So I'm used already to it. And I think the same for you that, okay, I prompt, I do some AI supported stuff in prototyping or my reports, wherever I am, there is AI support in it.

But if you change the way you do something, like with the shopping, that's really something new and amazing. So I did not see anything better than that so far. Operator, yeah, it's by booking a flight, but...

Yeah. Let's see what else. It's nice with the flight and the shopping, but I need more. I need more. Let's see. But this is definitely coming. This is just like a little sneak peek of the future. Do you think... I have a thing. No, go ahead. Go ahead. Do you think, because I was thinking about AI tools and do you think that because you give an AI course for designers, you give it online live, right?

It's pre-recorded, actually. So it's a six-week basic bootcamp. So people can watch pre-recorded sessions. It's like five to six videos that they need to watch and they get a task. And they're basically building a project.

during those six weeks. And we have one live call each week where we do basically a workshop. So work together, iterate, learn together. It's always based on the topic for this module, basically. But the course is coming in a self-paced version, hopefully soon. Let's see. I'm still working on it because I need to add some content as well. So there will be live calls then, but only every month with Q&As or so. So this is like a little bit of a different concept.

yeah okay okay yeah because i was thinking did you test hey jen for your own oh yes yeah i did it's but the problem is i was thinking about because i want to redo like the intro part for it the problem is it still feels a little bit unauthentic it's fascinating it's great but it's not

Yeah. Super authentic. 100%. Yeah. Yeah. Unfortunately. Totally agree. Yeah. Yeah. I also try to test it and then I was showing it to my friends and then I was, did not tell them that it's an AI and then just testing them. And they was like, no, but you don't look like that. You don't do with your eyes all the time like that. Okay. Yeah. And also the way how we speak our mindsets.

Mother tongue is German. So when we speak English, it's of course you hear it. It's not a problem. But the avatars, they speak like such a nice English. It sounds super professional and so nice, the pronunciation and everything.

So for me, when I posted it on Instagram, everyone was like, "Oh, your English got so much better." I was like, "Thank you for nothing." I would love to speak like that, like the avatars. Yeah, me too. Thinking the same. Yeah. Oh, so nice British accent now. But this is also getting better. First version, they also improved it a little bit. So I also see there are a lot of opportunities where you can create courses

on the go where you can create content with your avatar. But then the authenticity lacks a little bit because what I think makes things so fun is for me at least where people are like, how do you say democratization or like things that are human where you do mistakes, where you say something wrong, makes us human and I don't know, gives us a feeling of belonging somehow. And when everything is so perfect, then

Things get a bit boring and a bit gray and a bit sad somehow, at least what I'm thinking. So I don't want that future. But I think it's still fascinating. But what I wanted to ask you is, do you have a favorite AI tool? Yeah, I was thinking about that. I think it's Claude. Because I had this moment with Claude. I know many people say, yeah, but JTBT is doing the job. But I had this moment with Claude where I asked him...

a question and the answer was there was a logical mistake in your questions are you sure you want to ask that and I was like oh you are so much more intelligent than ChatGPT because the same question to ChatGPT is just generating text and ChatGPT is trained to give me a yes or anything but no one is giving me feedback to my question and I was like

And since then, I think Claude is the intelligent version of chat GPT. Yeah. Super interesting that you're saying that because I actually made the same or I had the same thought. I'm currently tracking my blood sugar with the tracker in my arm because I'm doing like a Luven. It's called Luven diet. You do this during pregnancy and you try to eat no sugar so you don't have the glucose spikes.

They say it's super healthy for the baby. So I'm currently doing that and I'm like sharing my glucose curve with TGPT to give me feedback on the meals that I eat. And what I realized is I always get so much motivation from TGPT. Always like, you're doing great. You had a wonderful meal. Your glucose spike looks amazing.

And sometimes it gives me some feedback, but it's so positive all the time. And sometimes it's too positive because my glucose scope is not that great. And maybe I have pasta with something which is not so healthy because you get like a huge glucose bias, a lot of insulin and so on.

And I would have liked something what you mentioned. Actually, you need to think about your meal. This is not the best choice. You should maybe change that. And it doesn't do that. Chachaputry wants to be loved. I feel like it doesn't want to give you a lot of bad feedback.

Or if so, you really need to force it a lot. And Claude is very different. I think Claude wants to understand, also ask a lot of follow-up questions if it doesn't get it right. So it's more like the curious type of person who really wants to give you a good answer somehow. And there is actually approval of that. So I did last year a project with Bertelsmann Stiftung in Germany, and they wanted to test different foundation models in terms of

Quality of answers, bias, transparency, data security. And the best one was in all the criteria after our end-hugging phase. And so maybe we can also link that in the show notes because it's open that analyzes as a website and you can, as a tech person, understand different criteria and which foundation models you can use in your products.

Yeah, yeah. And perfect. If you use a tool like perplexity or so, you can basically switch the models and choose the models. True. Yeah. That works well for you and for the project that you're working on.

Super nice. Do you have some resources that you... You already mentioned a couple of resources, but is there something else that you would like to share with the audience where you feel like people should have a look at? We publish our free resources on our website, encontrous.com and then resources.

At the moment we have a prompting guide that's for basic, but we also have, because people think prompting is basic, but inside there you find tips how to reduce hallucination. For example, the best tip is ask for the confidence level. How sure are you with your information from 1 to 100%? And I have a lot of cases where GBT says, yeah, I only have 50% confidence level. And then, okay, thank you. Yeah.

Yeah, that's something we share. We also share our bias sheet where you can observe, understand bias and how to mitigate and reduce bias. And we also have an AI product canvas. That's a Miro template you can just download and then use in your product development. So that's also

very helpful. I'm just setting up my newsletter, to be honest, because people are asking me, how do you stay up to date, Samina? It's so fast-paced. And actually, I'm reading newsletters from the US to be up to date here, but I lack a

A European newsletter with a responsible take on it. I would like to give my perspective on it. That's something I would recommend. How do you stay up to date, Patricia, with AI? Do you also read newsletters or how do you do it?

Great question. Definitely. So I must say, I spend also a lot of time going through resources and watching videos. For me, it's newsletters. I have some like Ben Bites, for example, that I like. The one from Jacob Nielsen, he's a famous designer. I really like his newsletter. And a couple of others where I just basically go through. I don't really enjoy them super much, but sometimes I find something interesting.

Then LinkedIn, of course, follow the right people, then seeing what they post. Then I have actually a chat GPT scheduler or like a routine integrated. Send me the most important resources, check the web and I get this every evening. This is something that I... Oh, nice. That's really nice. Do you filter it on AI for designers? Exactly. AI for UX. So this is the topic. And then I get this sent actually to my email.

at 6 p.m so in six minutes okay yeah um i get it that's super helpful and besides that

I think Instagram is not a good resource for updates. Sometimes I see something, but not that much TikTok. Neither, but it depends. Like some people are sharing really cool, helpful content. Yeah, I'm not on TikTok, so I'm not sure. Yeah, okay. Don't. Don't start. Yeah, TikTok is fun, but I think not so much for keeping up to date with things. And besides that, YouTube usually. So when I find something interesting, I just Google it, see maybe someone did a tutorial or just like a deep dive or description, and then I have a look at it.

I can recommend, I think it's the AI Revolution on YouTube. They have nice videos. Cool. Let's also link it. Yeah.

So nice, Samina, thank you so much for being in the podcast. I really love the conversation with you. Lots of insights and especially interesting to get your take as like a data science person who's more into the programming part in the data part. Super inspiring thing not only for me, but for a lot of the listeners. So thank you so much for. Yeah, I really loved having you here. I'm going to link all info in the description box.

If people want to check you out, they can find you on Instagram, on LinkedIn. I'm also going to share that in the description box so people can connect. Besides that, thank you so much for being here and talk to you soon. Yeah, thank you so much, Patricia. It was nice talking with you and exchanging with you. Of course. Thank you. Bye.

you