How to become AI First? - Part 1

00:00:00: Welcome to Designing AI Heroes, where AI and people align to drive productivity and innovation.

00:00:10: This is the podcast that empowers businesses and individuals to integrate AI into their

00:00:15: workflows and workplace, unlocking their full potential in the digital age.

00:00:21: We bring you insights, strategies, and real-world applications to help you be an AI hero and

00:00:27: stay ahead.

00:00:29: Let's dive in.

00:00:34: Welcome to another episode in the Designing AI Hero podcast, and I'm very happy to welcome

00:00:41: Sofia Raffa, or Sophie Raffa today, and we're talking about AI in Qatar, in Europe, the

00:00:49: AI markets, what are the differences about AI technology, and to announce, to get some

00:00:55: former in it.

00:00:56: This will be part one of a small podcast series we will do together to cover some very interesting

00:01:03: and nice topics about AI.

00:01:05: Hi Nadine, hi everyone, I'm Sofia Raffa, digital marketing and AI consultant and freelancer

00:01:12: blogger.

00:01:13: So I'm super happy to take part at this conversation because currently AI is a really hot topic,

00:01:19: and as you mentioned, not only in Europe, but also worldwide.

00:01:22: And I also do believe that we should also talk about a couple of points and also to make

00:01:27: clear a couple of definitions before we really start implementing AI.

00:01:34: So a little bit background about me, with more than 15 years experience on digital marketing

00:01:40: and more than 10 years experience with technology marketing, I'm more than excited to take part

00:01:47: in this conversation and hopefully with all my insights from Europe, from Middle East

00:01:52: as well as global, can help you move forward with your AI journey.

00:01:57: Yeah, thanks.

00:01:58: So there is a lot of hype around AI and out there, and it's really important to clarify

00:02:04: things and make things understandable, and we both work very internationally, I think.

00:02:14: So you live in Germany and you travel the world for conferences.

00:02:19: And I have my office also in Germany, but I live in Belgium and I work a bit around

00:02:24: with UK, Benelux and worldwide, and not worldwide, but mostly European based companies.

00:02:32: So the first question would be, are there any difference between markets, worldwide regions,

00:02:41: countries when it comes to AI?

00:02:44: Oh, okay, so let's get start with Qatar.

00:02:48: So there was a really incredible experience to take part at the web summit, Qatar as a

00:02:52: freelancer journalist and blogger, and I also had the opportunity to talk some of the representatives

00:02:59: from the Qatar business environment.

00:03:02: So recently I just published a really nice article with Hamad al-Hayri, who is the CEO

00:03:08: of Sununu, which is currently a unicorn in this region.

00:03:15: But there were also a couple of really interesting insights in Qatar, what I learned and what

00:03:21: we could also implement here in Europe.

00:03:23: So starting with the government initiatives.

00:03:27: So I think not too many people know that, not only Qatar, the Qatar government supports

00:03:36: the startups, but also the EU.

00:03:38: So currently there are initiatives such as the Horizon Europe, which is an EU flagship

00:03:43: for R&D programs, and this is actually a co-financing program for innovation and data management.

00:03:52: Also we have here in Europe, the Digital Europe program, which is an AI found for data spaces

00:03:59: and digital skill buildings, but then we can also have the list over and over again, and

00:04:07: then we can also talk about the Interag Europe, which is a support for digital innovations.

00:04:14: When it comes to Middle East, then the other really important point, what I would like

00:04:20: to highlight, which could be beneficial for European innovators, is to adopt a culture.

00:04:26: Because also here in Europe we have so many different countries, cultures, nationalities,

00:04:32: and what I learned over there from the SNU/NU team is basically to really shape your campaigns

00:04:39: as well as initiatives for the different cultures.

00:04:42: So it means that the different cultures have different needs, have different backgrounds,

00:04:47: and once you take care about all these things, you can also personalize all your campaigns

00:04:53: and messages.

00:04:55: And also what I really loved is the community-led marketing initiatives, so especially Middle

00:05:03: East, we can also talk about Dubai and also other bigger hubs.

00:05:09: Currently there are a lot of micro-influencers on TikTok and Instagram, and yeah, basically

00:05:16: they are selling properties.

00:05:19: So for a huge amount of money is already circulating in these communities, and all what we should

00:05:25: start doing is to somehow identify these people and also starting working together with them.

00:05:32: Because currently in Europe, when it comes to marketing, we are still really relying

00:05:36: on the old-fashioned PPC, so paper-click campaigns and maybe working with publishers.

00:05:44: But we tend to forget that nowadays the whole TikTok brought us a completely different business

00:05:50: model as well.

00:05:52: And when it comes to different cultures, it was a really interesting talk where we also

00:05:56: participated like a week ago, Erin Peters from AI Maturity Index started a conversation

00:06:04: about culture and how culture can influence AI adoptions.

00:06:12: Maybe Nadi, you remembered because you have also…

00:06:14: Yeah, she talked about what is really going on in Germany, and I also see that German

00:06:22: companies are already hesitating to implement AI, or there are a lot of initiatives, but

00:06:30: in my way it's not in a strategic level, so that we have a roadmap like, okay, this

00:06:35: is the AI potential.

00:06:37: What does this mean for our business?

00:06:39: How to integrate it and have a company-wide programme to onboard everything?

00:06:46: And that's the next thing, AI first.

00:06:48: So I hear a lot of it in the US or also in other world regions, we become AI first.

00:06:56: What does it mean?

00:06:57: So it really means integrating AI, not only train people, and have a flexibility running

00:07:05: from, or I ask something on flexibility when I have a question.

00:07:09: That's not AI first, this is an AI support.

00:07:12: AI is first, it's really having a mindset, thinking about your workflows in the business

00:07:19: and where AI can add value and then really integrating AI step by step.

00:07:25: And that's not happening in Germany.

00:07:27: And how would you define personally?

00:07:29: What is AI first approach in business?

00:07:33: AI first approach in business for me is you look at your workflow from a business aspect,

00:07:39: you see, okay, where can we add different AI approaches in our workflow?

00:07:46: It can be from AI chatbot to manual automatic workflows, to automatic AI workflows, to AI

00:07:53: agent.

00:07:54: I mean, that's the level when we talk about AI approaches, you have AI chatbots up to

00:08:00: AI agents and there are different approaches between what you can do and not every approach

00:08:06: is the best fit.

00:08:07: So if this approach is the best fit, or is the data so bad when we integrate AI, we still

00:08:13: have a bad process based on data.

00:08:15: It's a wider and broader approach for me, not only are we integrating for pilot and

00:08:21: chat GBT in, or we do a meeting transcripts.

00:08:25: And then we have a transcript, yeah, we use AI, AI, AI first.

00:08:29: That is not AI first mindset, it's really thinking about a new way of doing things and

00:08:36: a way of doing things different than before.

00:08:39: Yes.

00:08:40: I completely agree with you and just a little add, like last week, I started reading a couple

00:08:48: of blogs from Kessie Koziakop, so she's Google's first Chief Decision Scientist and she had

00:08:55: a really interesting article, it was about please cheat at your job.

00:09:00: So basically, she has a couple of questions, like adopting improvement mindset, so how

00:09:07: better, easier and faster already existing processes and mechanism, and also can be

00:09:14: scaled impossible. So basically, she's also highlighting the fact that let's start utilizing AI,

00:09:22: but first let's define certain processes and then moving forward. But also doing this whole

00:09:29: automation and all these processes with responsibility. So just taking care about the data, where it

00:09:35: comes from and all those things. And also, it shouldn't be a kind of secret that you use AI,

00:09:42: because many times in my conversations, I heard from people like, oh, I don't want to let for my

00:09:49: boss know that I'm doing this because then he's going to think, oh, that's so easy.

00:09:54: I'm cheating. I'm cheating. Yeah, exactly. And that's also a big issue because then shadow AI

00:10:01: comes into the conversation, which means basically that there are the certain approved processes,

00:10:07: but then on top, there are this secretly used AI tools, which doesn't really bring too much

00:10:14: value for long term, because if they are not integrated into the system, then how can the

00:10:19: whole machine learn from your past and then utilize all this information into the present?

00:10:24: So that's also a kind of trend, what we observe currently here in Europe.

00:10:30: Everybody now is speaking about a genetic AI. When you go to Microsoft Copilot,

00:10:36: it's always an agenda AI, but it's not an agenda AI. It's just marketing from Microsoft.

00:10:43: You have to understand we have to become AI safety. We don't have to be the tech experts

00:10:50: or the programmers, but we have to understand, okay, this is an AI workflow. This is an AI chatbot.

00:10:56: This is an AI agent. And what does this mean? What is the difference? And the difference is also

00:11:03: where, and then it comes back to the business, to my workflows, and also to my risk management,

00:11:09: where do I have to human in the loop? Where can the AI run automatically? Okay, I set it up and

00:11:16: it's running. Or where do I have to human in the loop? So not thinking about when it comes to

00:11:24: HR, for hiring, do you really like to hire your talents from authentic AI? No, there should be

00:11:32: human in loop. So you also have a risk management. Okay, what is this really risky workflow? Where can

00:11:38: AI support? Where can it run automatically? And where I have a human to check the whole thing

00:11:45: and make the final decision? So that's also an AI first mindset. I don't have to have an AI agent

00:11:53: everywhere and say, yeah, we have AI agents, now we are so great. Their first mindset is

00:11:58: thinking about the different approaches, what makes sense, can learn that and understand that.

00:12:03: I completely understand that you and Nadine, I just invite you for a short conversation about

00:12:08: to put all the things together on the puzzle. So what is chatbot? What is AI assistant? What is

00:12:15: automation? What is agentic AI? So I would say once if we put all these things together, then

00:12:23: everyone can understand us maybe a bit easier because many times I see that people are talking

00:12:28: about chat GPT as it would be the one and only thing and nothing else would exist. But I think

00:12:34: it's also important to realize that we already have already existing systems. So such as the chat

00:12:40: bots, which are already here for a long time ago. So they are basically a rule based or scripted

00:12:46: systems which designed to simulate conversations, but everything is pre-designed. So many times,

00:12:54: also in the past 10 years, there were tons of chatbots everywhere integrated. So it was basically

00:13:00: a kind of question. And then there were some predefined answers. And based on this funnel,

00:13:06: you could go down. So it was a kind of no memory or learning and it was actually as good as the

00:13:15: script it was. From your perspective, what is an AI assistant? An AI assistant is like you take

00:13:22: chat GPT co-pilots, you ask a question and you use the data from the element for a large language

00:13:29: model. So you don't give any data to it. You just use a large language model to have a problem solved.

00:13:36: Then the next thing, the next level is, okay, I give data to the system. So I don't use the

00:13:43: large language model or the data from a large language model. I enrich the data for this specific

00:13:48: task. I upload and document. It's very easy. I only use my data. And I know which data is used.

00:13:56: I'm okay. You also have hallucination then a bit. You have to care. But you know which data it's

00:14:02: used. I mean, it's automatic workflow. So then it's set up like the AI will forget the data from

00:14:09: somewhere. Exactly. The trigger is not you. It doesn't to be a human. You can also say, okay,

00:14:17: every Friday, you search for studies from the internet, sum it up, send me everything into

00:14:25: an email, I love the example also for my for my learning or when I say, okay, I have I have to

00:14:32: have the updates every Friday afternoon from the week. Yeah, the trigger is every Friday afternoon

00:14:40: at three o'clock, the eye gets triggered goes automatic in an SS feed or something like this,

00:14:45: searching for new or in your in your company, the shape and search for the documents,

00:14:51: sum it up, send you an update via an email and you updated about topics. There's no human in

00:14:57: the loop. There is no human in the loop because you know what's doing, you know which data is

00:15:02: accessed. And you know what's happening in the workflow. And the output is always the same.

00:15:07: And the next level is is a gender AI. Yes, then it comes. It's the same like an automatic AI workflow,

00:15:15: but the agenda AI can work together with other AIs and access a large language model,

00:15:21: which is the best to answer the question due to can't predict the output.

00:15:25: Exactly. And the system also learns from the last action.

00:15:31: It learning as a huge memory. And that's a problem. For example, when agenda AI hiring people,

00:15:37: you don't know the output. So yeah, you don't know why this person is rejected and why this

00:15:44: person is hired. You don't know it. Yeah, and the decision is up to the AI.

00:15:48: Yeah, definitely. And I also see as a kind of threat from my marketing experience regards to

00:15:56: agenda AI, because it depends on the certain lookback window. So how long from my past, so

00:16:04: as the whole system takes all the data, because we also had a couple of marketing campaigns.

00:16:11: And many times the issue was when we store the data. So obviously store the data means like

00:16:17: we used data from the past one year. But within one year, so many things can change

00:16:24: in people's life. So if we say that if we go for dating, so for example, there's a single person

00:16:31: then goes for a date, maybe half a year later, they are already planning their wedding. And then

00:16:38: after nine months, the little kid comes. So within one year or two years, so many things

00:16:44: can come up. So the system must learn, but not really taking all the past data into account.

00:16:50: Otherwise, the system is going to hallucinate. And I think that's one of the biggest challenge

00:16:55: what the agenda AI should somehow solve for long term. So what should be the lookback window

00:17:02: in terms of data? I think it's super important. Okay, I think that was a nice

00:17:08: part one of our podcast series. We talked about the different AI approaches and hope

00:17:13: it's clearer than before. In the next episode in part two, we talk about what these different

00:17:20: approaches means for your business, how to apply it, what are different steps and what about the data.

00:17:27: Exactly. Sofia, many thanks to attend part one. And I'm really excited to part two in the next

00:17:36: couple of weeks. Thank you for your time. Have a nice day and many greetings to Frankfurt.

00:17:41: Yeah, the pleasure is mine. Let's make things sophisticated together. Ciao. I like this. Bye.

00:17:46: That's a wrap for this episode of Designing AI Heroes. If you enjoyed today's discussion,

00:17:54: be sure to subscribe and to leave us a review. Stay connected with us at www.designingaiheroes.com

00:18:02: for more insights, resources and updates. Until next time, keep innovating and designing the future.

00:18:08: One AI powered step at a time.