Bonus episode: Panel discussion about the state of AI in today’s world

Welcome to a special episode of AI Unscripted, recorded during our recent strategy session in Antwerp. This conversation brings together leading voices to explore the urgency, risks, and opportunities of the AI revolution—and what it means for your business, your teams, and the future of work. 

Panelists Xavier Verhaeghe, Rik Vera, Joris Van Der Gucht and Prof. Dr. Ann Nowé draw compelling parallels between the current AI wave and the early days of the internet, noting that AI is an active, rapidly evolving tool that demands swift action from organisations. Unlike previous technological shifts, AI’s accessibility allows even small teams to build impactful products quickly, reducing barriers to innovation and enabling a new generation of creators. 

The discussion highlights the importance of responsible AI adoption, focusing on transparency, explainability, and the need for interdisciplinary collaboration. Social scientists and technologists are working together to address bias and make sure AI aligns with human values. Human input remains essential, guiding AI systems and validating their outcomes, especially as large language models become more prevalent in business and daily life. 

Real-world examples showcase how AI’s reshaping industries, from gaming—where AI enhances user engagement—to professional services—where clients increasingly seek validation of AI-generated advice. The panel acknowledged the unpredictable pace of AI’s evolution, urging leaders to foster a growth mindset and embrace experimentation. 

Ultimately, the session underscores the critical role of leadership in navigating AI-driven change. Storytelling, vision, and engagement are key to overcoming resistance and making sure organisations thrive ethically and responsibly in the age of AI. 

Join us and listen to all episodes on www.pwc.be/aiunscripted

Transcript

Special Panel Setup

SPEAKER_01 0:00

Welcome to another episode of our AI unscripted series. This one is a relatively special one. Because a few months ago we hosted our periodic um strategic uh session with our entire partner, the director and manager group in Antwerp. A very interesting event, uh insightful, thought provocative here and there. And one of the topics related to AI. A session hosted by Xavier Verhagen, our AI uh lead. And um again, as I was saying, interesting. That's the reason why I thought it's a very good idea to put it also as a special uh uh episode in our AI unscripted series. So here you go. Have a look, and um yeah, hope to uh see you uh very uh soon again in one of our next episodes of AI unscripted. Thank you. Bye.

SPEAKER_02 1:02

Hello everybody. Hello, good. Yeah. I was just checking if there was anybody in the room, but yeah, you're still there.

SPEAKER_03 1:10

They don't realize it, but you're talking to a black wall. Sorry guys. With a lot of light wall.

SPEAKER_04 1:15

A lot of light.

Urgency, Layoffs, And Opportunity

SPEAKER_02 1:16

Rick, what an amazing presentation, uh, energizing presentation um for an amazing intro to the topic of AI. Uh, thanks for setting the stage for a panel conversation with a top panel. Uh I think that through the presentation or during the presentation, you've seen that there's tremendous change coming out of us, but already happening now. I'm not sure if some of you watched uh news lines today, but Abian Amru, for example, announced um laying off 25% of their workforce, a bit more than 5,000 people, um, by the 2028.

SPEAKER_03 1:52

Amazon is laying off 16,000 people for the moment.

SPEAKER_02 1:56

Yeah. Um, but there's a lot of positive elements in the transformation of AI. There's a lot of topics to be discussed on uh about. And what we see is that there is, I would say, so much potential, so many good stories, some many scary stories as well. Um, and everybody has a feeling that okay, that wave is there, we need to start. And still, when we have conversations with clients, some of the clients still say, like, oh, what should we do? Where do we start? How do we start? And there's a lot of questions coming at us. So during the panel conversation, we're gonna focus on three things. Uh, one is we're gonna talk about this business potential and the sense of urgency that you created as well. Uh, the second element is we need to bring some reality as well. What are the risks? What are the elements that we need to be thinking about? Because our clients and our teams really want to have trust advice. They want to have trust in what we say, what we bring. We should not only do the replication of the shiny stories, but make sure that with what we say is correct and that that we lead to the right path. The third element, and I think it's an important one with this audience, is we will need to rethink what the impact is on our work for strategies, what it is on us individually, but also for our teams, for recruitment strategies, and so on. So we're gonna try to touch upon all those topics. Um but maybe a first one, and if I may, Rick, uh you talked about all these great stories and the sense of urgency, the wave,

AI vs Internet: Active Tool Shift

SPEAKER_02 3:27

and so on. There has been a number of changes over the past 200 years. Um I think a very impactful for me, probably the best comparison is the internet, but there was also a lot of doom and gloom. And uh I how do you compare the AI wave now with the initial moments and then I would say the uh well, the omnipresence of the internet or maybe mobile?

SPEAKER_03 3:50

Um and I have the pleasure to have met uh Satin Venus-Lee, the inventor of the internet, a couple of times because I was running a project together with him for the Flemish government. Um and that was 1993 when they made it public. And that was an inter that was after thought it's the internet moment. We are living a new internet moment. But the big difference between the internet, and by the way, Sir Tim Banasley is not happy with what we've done with the internet. That was not his intention. Um, but it's not the internet, because the internet protocol has hardly changed, but it's what people and companies have done with the internet. And the first, the pioneers, were the ones that have conquered the world, and the others were way too late. Look at e-commerce, for instance. Um, we are living a new internet moment. But the big difference between the internet and AI is the internet is a passive tool. AI is not a passive tool. The internet gave us like 30 years, and I still still see companies out there that think about starting a digital strategy with 2025. Um, we will not have that much time with this active tool because every day that we're using AI, it gets better and better and better. Remember the first versions of Chat CPT versions, the versions that you have today, that's already a huge difference. And we haven't seen anything yet. So the difference between those that use it and those that use it to create something, and those that are still waiting, the difference between winning and losing is never going to be more devastating and is never going to happen faster than today. And I'm absolutely convinced about that.

Accessibility And The New Level Playing Field

SPEAKER_03 5:36

Joris, any any additional reflection on that?

SPEAKER_05 5:38

No, no, I uh 100% echo Rick and so it's uh what we always call the the wild west in technology.

SPEAKER_03 5:45

So the wild, wild west, the WW. Exactly.

SPEAKER_05 5:49

So it's never been so exciting to be in technology, actually, but uh let's be honest, nobody knows as well what's going to happen. So uh every week there's something new, a new release of a model, a startup popping up that in two months' time hits 100 million in ARR. Can you imagine that? Uh so everything goes so fast, so quickly, and we get so much information, an MIT study that says like it's not delivering the value. Next day you see an amazing, impactful uh startup then again. So it's the Wild West, but I agree 100% with Rick. It's uh the new platform shift. We had already some platform shifts the past few decades: the internet, mobile, the cloud. So AI will probably reinvent everything uh in a way how we look at technology.

SPEAKER_03 6:34

But it's a new level playing field for everybody. So the next, and I really mean that, the next Bill Gates, the next Jeff Bezos might be a 13-year-old somewhere in in a basement, and they don't even need the basement, they don't even need a garage. Yeah, they just uh have access to all types of AI tools and they play around with it. For us, it's all new, and for them, it's only normal.

SPEAKER_02 6:54

But but there I and I will immediately come back to Anne about this. But I mean my first AI project was in 1994. It was forecasting for ABNB. And I mean, you're part of the AI Innovation or uh Innovation Center of the VUB, or one of the oldest ones on the European continent. Uh I mean you have seen AI already. I mean, that wave has been coming already so long, you know what it can do and what it can't do, etc. How do you look at it right now? Because we're all like excited, and I am I must admit I'm kind of like one of those AI positivos. How come that suddenly now the change is happening? And how do you look at it? Because you have been following this for so long.

SPEAKER_00 7:35

Okay, but we'd say it's not the first wave of AI. The first wave has been like in the 80s, where a lot of expert systems were being built, and they were actually used in industry. So we got a lot of requests, so not me personally, but but then in the lab, in the beginning of the lab, uh, to build these expert systems, and still a lot of these techniques are used today in call centers and and uh things like this. So um we speak about winters and and uh summers of AI. Now it's like uh a heat wave. Uh it will not really cool down, to my opinion. It will cool down a bit, but not uh as we had the winters before. Um but I think a big change is also that uh a lot of um trends are now and products because we're talking about AI tools, not so much about the AI techniques, which are of course behind the tools and which remain much more stable. Um but it's it's the big tech who has a lot of power to predict what's going to be the like the next product that will be made accessible to a wider public, and that's very different from the waves we have seen before.

SPEAKER_05 8:48

And the accessibility in itself, I think uh everybody here can become a coder. So I don't know if you have ever tried lovable, like you can build a product in an hour time. So the the access to build technology, the access to AI has fundamentally shifted uh compared with the 80s, where you really needed to be the domain expert, super unique skill sets. Today we all can start with adopting and building an AI.

SPEAKER_03 9:14

Everybody, everybody can use

From Expert Systems To Today’s Heat Wave

SPEAKER_03 9:16

everybody can use a complete sales and marketing organization in no time, and at zero cost.

SPEAKER_02 9:23

Can I just make it a bit more pragmatic and tangible for the audience? Yeah, can you just explain a bit what you're doing with Rafical and also how it has influenced because your company is like seven months old? Seven months, yeah. Seven months on the market, yeah. Seven months on the market. And can you explain what it does and and and how you could move that quickly? Yeah, because it makes it maybe tangible for the audience on with a real case.

SPEAKER_05 9:47

So, some background. Uh, two decades ago, I was a consultant myself, and it's quite a funny story because at some point I had a PWC logo on my LinkedIn, but that was because of an acquisition that PwC has done, the acquired Hansen and Partners. Okay. So I was what at some point your colleague. Okay. Then I started Silverfin, which was 2013. Uh, the main question was what is the cloud? Can you imagine that? And now with the accessibility of uh Gen AI, Agentic AI, we are building the ABC, the all the building and the complex for the uh expert firms for the knowledge industry. So we really believe that we can offer expert firms an orchestration platform where they uh can orchestrate AI agents, uh different tooling, and we help these firms with, first of all, tangible cases. We start with a deep research agent, with an inbox agent, with a cross-cell agent that automatically identifies cross-cell opportunities across a client portfolio, and then gradually we'll evolve in that change journey.

SPEAKER_02 10:51

Yeah. And you mentioned in one of our previous uh discussions also about the speed of programming and the type of resource. Because I mean, if when you talk about AI, some people will may still know. I think three years ago, even in the newspaper, it was like, oh, we all need data scientists and AI and so on, the use will not move forward because actually we don't have enough data scientists. You have hundred data scientists in your company?

SPEAKER_05 11:16

We are 20 experienced uh technologists.

SPEAKER_02 11:19

Okay.

unknown 11:20

Yeah.

SPEAKER_05 11:20

So how how do you I find it super exciting to build a technology company in this time because if you can start building from the ground up, there are basically no rules anymore. I will give you some some practical examples. So um within Revical, uh all our client-facing calls are recorded in a note-taker. And I think you all have a note-taker. That's not what I want to say, but that means a product manager, an engineer can just prompt all these calls, can look for the specific snippets where a client is asking about uh or is sharing his frustration about a workflow, what he wants to improve, like on a week time. Before that, very practical if you were building uh technology companies in the in the cloud

Startup Speed And Agentic Platforms

SPEAKER_05 12:07

uh era, like that took some time before you gathered that feedback. And uh then it means that a product manager, in example, can just start vibe coding. I don't know if you know the term vibe coding. I like that term very much, to be honest, and can start designing a product where in the in the in the past you needed to have a designer, you started building some mockups, you uh started sharing them with engineers. All of a sudden, the product roadmap itself can be done by one uh experienced person, a handful of coders that use AI themselves to code, and then you can really at only seven months time, you can release a full product. I think that's a total difference and a total difference in size of investment as well that you need to do to go to market. And I find that super impressive.

SPEAKER_03 12:53

But you see the the the potential difference in speed, the speed that you can make at zero cost versus how traditional companies are coping with exactly the same problem because they still have all those functions, they still have all those people, they still have the slowness. So you speeding up at incredible speed, at no cost, for them it's very expensive and slow. So the difference between those that win and those that might lose is going to grow by the day for the moment.

SPEAKER_02 13:25

And and I I did a number of boards presentations meetings uh recently, and everybody would say, like, oh yeah, AI is important, therefore, also it was on the board agenda, everybody's excited, see the positive. And then you see at the end surfacing the questions. Um, but but uh we first need to have our data in order. Or we didn't think about AI. I mean, the number of excuses are are like piling up. Um, if you look at I would say the um I mean some of the comments are also about risk and the bias, and and I mean there's a lot of things that are like um surfacing of people that have probably read articles over the last two and a half years. But what is your view on this? Because again, you see well, you don't only look at the business transformation side, you also look at a lot of the different dimensions in your research. Can you elaborate a bit on that risk element?

SPEAKER_00 14:19

So uh bias uh or the problems of bias or how to de-bias AI models, how to make them more explainable. So uh that got much more attention the the past 10 years, I would say, uh, because of questions from businesses or organizations. Uh, but I would also phrase it like it's thanks to AI that we pay attention to bias, because uh it's not only AI that is biased, it's also humans, it's also other decision-taking uh uh models that can be biased. Um so it has driven our research. Um we come from an and if you talk about the 80s and the expert systems, they were very transparent. They were modeling expert knowledge very in a very transparent way, and you could interact with these systems, you could ask questions like why and how did you come to that conclusion? So extremely transparent, but it's because of how we move to other AI approaches, which we call the sub-symbolic or numerical approaches, the statistical approaches, where um yeah, we lose a bit of transparency, and now we do research on how to get that back, but also how to communicate to people. So uh the language models are very powerful because you can naturally uh have a natural conversation, and and this is also important to explain to an AI

Data Excuses And Getting Started

SPEAKER_00 15:50

uh what are your objectives, what are your concerns. So there's a big difference between uh the mathematical object you want to optimize, so to say, what we call a loss function, things like this, and and what does the community or or the company requires as a solution? So bridging that gap where it before we did it through interviews, now we want to make it more um uh we want to support it more automatically. Um that's where we put a lot of attention today. So even before we built the AI solution, we want to already facilitate uh a conversation. We have an EU project on this, um, but that's an interdisciplinary approach that you have to take. We have social scientists on board and so on that also uh learn us how you can have that conversation. So AI has always been very interdisciplinary, but it has uh gone away a bit away, it has become very uh computer science, mathematically driven, and now we come back to this more uh interdisciplinary approach, and that's also how we want to train our students uh in at the university as well.

SPEAKER_02 17:09

Talking about training at the university, how is it shifting, in fact, the uh the education programs? Because when you I mean the examples that you gave, it's it's kind of like this human in the loop, and you need to potentially validate, but actually it's even more than that. It's it's asking the right questions. You need to steer AI in the right way. It's not just validating the outcome, which is often confusion that that is happening. Uh, but how do we make sure? Because it's it shows actually the criticality of having the right curiosity, the the critical mindsets.

SPEAKER_03 17:42

The that's why the 5C mindset. The 5C is actually coming back at that. And and you're absolutely right. You need two things. You need data on the on one hand, and you need the what does it need to do for us. That's the reward structure on the other hand. And part of the reward structure is it's an alien intelligence, it's more and more a black box. We don't know what's happening into the black box. So we need to be very clear in the reward structure. What do we want it to do for us? And as long as we're using Chat GPT, that's not going into territory that might be dangerous. But

Bias, Explainability, And Human Values

SPEAKER_03 18:15

once we go into what is compassion, what is critical thinking, um, we need that human input because we need to tell this let's not call it artificial intelligence, it's alien intelligence. We need to instruct that alien intelligence what it is to deal with human beings. And we are humans, so we don't have to explain to another human how it is to be a human, but to an alien intelligence we need to explain what it is.

SPEAKER_02 18:44

If you then look at uh again, maybe uh going further in on on the on the topic, the reliability, um, there is also the concept of oh, there's an error when Gen AI is coming with an answer and and so on. And often the debate is yeah, but what does need to be kind of like does everything need to be 100% correct? Are humans always 100% correct? And then I hear, oh, it needs to be the human validation, but when we look at some of the Gen AI and AI systems, we're talking about a massive number of transactions, it's actually something that a human cannot control. So, what are the the current uh evolutions in reliability from a I would say more scientific point of view? And then I may come back to you on how important is the hundred percent reliability or correctness from a business term in what type of use cases.

SPEAKER_00 19:40

Yeah, with with uh with the large language models, we we don't really have good uh uh approaches uh to um let's say to validate them, because the they're so broad, uh they can basically talk about everything. So the typical way we um either prove that an AI system is doing what it should do, or how we uh, for example, evaluate a machine learning model, that cannot really be applied to these LLMs. So um, but uh as soon as you then uh focus more on on a specific area, you it it becomes again doable to to use the techniques we have. It's more of an issue if you just let it open in the wild and everybody can interact with the LLM and do whatever they want. Uh there's no guarantees uh of how the system will exactly behave. It's trained to please you, but that's it. Yeah.

SPEAKER_02 20:39

But we're sometimes uh or or the teams are sometimes writing controlling agents. So you have the agent that does what you ask it to do, and then you have a controlling agent to uh but that's that's more protocol, I would say.

SPEAKER_00 20:51

We have been talking about the internet protocol. Uh, in a lot of these multi-agent systems, we basically have a kind of protocols, and then you can again start to prove things that the system will behave as it should behave, as long as it is not too complex.

SPEAKER_02 21:08

Yeah, yeah, yeah.

SPEAKER_00 21:08

Yeah.

SPEAKER_02 21:09

Can you take it a bit from uh a business cycle on the on the correctness or 100% reliability?

SPEAKER_05 21:14

These types of debates like uh can go very high level, broad, but I think the the opportunity these days is with the the the vertical agentic solutions going deep in a specific industry. And I think today the combination Gen AI, agentic AI today, because we uh quite often mix them as well, so can bring that transparency, can bring that reasoning, can bring that visibility, how an agent has gathered information. But of course, in the face where we are today, there always should be a human in the loop. That means at maybe the advisor side, but that means as well as at the coder who's developing uh developing the specific uh vertical software, there is always a kind of agent in the loop to improve that.

Reliability, Controllers, And Human In The Loop

SPEAKER_05 22:00

Model and on the on the 100% correctness, I think if you transparently are able in a specific vertical to articulate really the reasoning of agents, where did they gather the information? And a human can then uh add another prompt and say, like, yeah, but please can you look into XYZ as well? That is, I think, what is mature today in technology. If if we really expect that AI already comes up with the the most brilliant text memo without having any context, I think that will be impossible. And then I think we'll we'll be dreaming. We'll get there, for sure. I'm a firm believer in that, but today we need to look at the technology, look at the specific verticals, and try to fix very tangible use cases together with that human in the loop.

SPEAKER_03 22:44

Yeah, yeah, yeah. I think that um and I know I'm different, but I I'm I'm I'm a creative thinker. I from time to time I miss the first versions of Chat GPT. I loved the hallucinations. Because the yeah, I really I um I loved them. You know, my next book, the title of my next book is an hallucination, was an hallucination. I was introduced on stage and I was introduced by a number of books, and one of the books that were mentioned, I did I haven't read I haven't written that book. But I thought, wow, this is a brilliant title. And I really from time to time I missed those hallucinations because I love to be challenged, I love the to to connect the dots. And now from time to time I find it utterly boring because it's yeah, it's so it's so careful to not make any mistakes that you can't even force it to make a mistake because it refuses to make a mistake. While the first versions that just went in all in and they came up with whatever answer, and from time that answer was so wow, okay, I can do something with that. And now it's so utterly boring. So it's like reading an encyclopedia. Yeah. So it depends, it depends on what you want to accomplish. But from time to time, I wish I could switch to the vast versions again.

SPEAKER_02 24:07

If you it makes me think about some people that say we want to go back to the Nokia phones. Sorry, this is for the older people here.

SPEAKER_03 24:14

Yeah, no, I don't want to go back to the Nokia phones.

SPEAKER_02 24:17

No, maybe uh um a general question because we're talking about transformations and so on, and uh we see like three levels. So, for example, the co-pilot where you have personal productivity, where people can in their personal day-to-day activities do things faster, better. Uh then you have more the process re-engineering, some like uh some of the agents that you uh develop and so on. And then you have like the real disruption area. Some, I mean, companies that are suddenly delivering a type of service that is was not existing before. Any perspectives on that? Any examples that are worth sharing with the audience? Any uh on the three levels or or on on maybe an exciting disruption story or so?

SPEAKER_05 24:57

Well, I think the the pattern you can see if you look at uh, in example, Silicon Valley, like the the war for talent is becoming like a Champions League, it's like transfers of the messes of the Ronaldos

Copilots, Process Re‑engineering, Disruption

SPEAKER_05 25:09

going to the metas of this world. I think that is for the first time ever that that goes at that size. You mean about the the salary that the salaries, like but that is where we are going, like we are going to like the the true experts, that will make a difference. So uh in terms of um disruption itself, I think if you can build from the ground up with true experts, then you can make a difference. So that is a trend you see. If you look at the what Cloud is doing with their financial services product in the the business analysts world, so the thing they have done is they have hired a bunch of very experienced business analysts. That's totally new compared with with how a software was built like a decade ago.

SPEAKER_03 25:52

Um what I see in um and I work quite a lot with established companies. Uh, yesterday I was uh lecturing at London business school and I had the top ten banks, global banks, with me in my classroom. Um, and then you you I do an exercise, and the exercise is try to develop a superpower. So I I mix them in teams, and every team is a bank, and try to come up with a superpower, and most come up with superpowers that are related to AI. But the second part of the exercise is you can also come up with kryptonite, and you give the kryptonite that makes your the others powerless to the other teams. And they come up with lists of kryptonites. Why do I run the exercise? Because at the end of the exercise, I ask them how much kryptonite is still in your own organization, and they can develop whatever superpower, and that's the the the the problem with companies for the moment is they have access to the superpower, but they have so much kryptonite in the organization that slows them down, that makes them bureaucratic, that is going to have a committee to consult a committee and to check a committee just to find out, you know, is the difference between in in Silicon Valley the building a rocket and we are discussing where it's going to land and what is the color of the fuel. Um and that's a kryptonite. And companies not only need to focus on developing that superpower, but also how the heck are they going to get the kryptonite out of the organization? Because they have a thousand and one reasons not to act. And then they're going to keep on dreaming about it and it's not going to happen. So I think that is very important as well. Um yes, we need to be careful with AI, absolutely. But if being careful means we're not going to do anything, um we can only find out what we need to do once we start experimenting. And and uh that brings me to another topic, the UAI X. Sorry. I'm talking to the UA to the uh to the Commission on Thursday, and I'm going to sh uh kick ass. So tell us uh a sneak preview. This is confidential, so this this is all I'm going to talk about the seven sleeping

Kryptonite Inside Large Organizations

SPEAKER_03 28:08

pills of the of the European Union. Okay. Yeah. Yeah, you can imagine what the seven sleeping pills are. It's uh regulation, um, it's paralyzed, it's uh um it's all that type of stuff. What what what slows them down? It's about the kryptonite. We we try to regulate something that's not are there yet. Let's regulate when something is out there and when we see what works and what doesn't work. Yeah, we need to be careful. But if we are so careful, I I really worry about this continent.

SPEAKER_02 28:47

But I'm at the risk of being an AI positivo, I I tend to agree. But uh if you look at the broader perspective, there's also different I would say reasons why regulation uh maybe not the regulation that is there now, but why regulation may be there and and you have seen evolution of the regulation discussions about it.

SPEAKER_00 29:07

What I appreciate on the AI Act is that it's talking about the applications and not the techniques. It's not that a certain exploring a certain technique is is not allowed. It's it's what do you build? What is your application? But what what I personally find very surprising is um why is law number one not AI is not above the law. So I sometimes think, why do you need to reinvent specific laws for AI? Uh if I talk to companies and and we're discussing uh a project and and then of course at later on it becomes a potential application. Um yeah, it it has to follow the rules and and and it depends on on the area. If it's it's something medical, it will need to go to medical certification. If it's uh a machine, um yeah, then then it has to follow other rules. Um so you you cannot come up with a solution that at the end will not lead to a certification. So why? Uh I mean for me it's it's giving the wrong message. So if you have four water.

SPEAKER_03 30:15

That is something out of this world. But we've it's the second time we've seen this with digitization too. We we have not, I mean, the rules that are that apply in the normal world, we forgot to apply them in the digital world. Why? Because we called it virtual. We never took it seriously. But it is a real world. Look at your kids, they grow up in this world, and for them that's a normal world. But it's not regulated.

SPEAKER_00 30:40

Yeah.

SPEAKER_03 30:40

Why do we have to apply different rules for digital? And you're absolutely right, it's just the rules of living together.

SPEAKER_00 30:48

And there's some blind spots of new things, okay.

SPEAKER_02 30:51

But let's cover it But I think that's uh an important reflection for all of us to make as well, because um using any AI tool, and we have the luxury within PvC to I mean have access to quite some AI tools to do our work, uh, but it does not kind of like accept us from thinking about it responsibly, about how

Regulation, Risk, And EU “Sleeping Pills”

SPEAKER_02 31:10

we use it and how we use the result.

SPEAKER_03 31:12

Like it like in normal life. Yeah, normal life. If you have, I mean, I get into my car, my car is a weapon, but I'm not going to use it as a weapon. Yeah, I mean, that's that's exactly the same. But there's a new EU car act being launched.

SPEAKER_02 31:26

No. Okay, yeah. Um I'm I'm just going a bit to the dark side. Um we talked about oh you've talked about in your presentation about the the planes. Yeah. But actually, the uh you talked yours about uh games and the dark sides. Uh yeah, you're not the dark side, but you talk about gaming and AI and and how the the interaction uh started being sticky with using can you explain a bit on on the full story? Uh this is about games where people think that they are playing against another human, yeah, and actually the the person that they're playing against is actually AI.

SPEAKER_05 32:05

I I think there is a global trend happening where like a lot of public uh gaming companies become private again. Last case was a uh Saudi private equity company that bought like a super big uh public gaming company, and one of their uh or part of their thesis is that they can really double down on AI investment to make players super hooked or addicted in the platform. So uh imagine if you buy a game or you start playing online and it's a multiplayer game, so you start to play human to human, you need to go through a learning curve. And nobody likes losing, so quite a lot of people stop because they're playing against somebody who is super like advanced and already playing it for years. So um they will uh totally or are uh fooling us because uh we think we are playing human to human, but actually we play against AI humans.

SPEAKER_02 32:59

So you you win a bit and you lose.

SPEAKER_05 33:00

You win a bit, you lose a bit, you win a bit, you lose a bit, but you really get hooked. So uh and the gaming industry in itself is already massive uh profitable once done right. It's it's amazing.

SPEAKER_03 33:10

If you talk to our Minister of Education, because this seems like a very good idea to educate people. Yeah, exactly.

SPEAKER_00 33:16

Yes, well, there was actually a link, I would say. So games and AI uh that goes together already for decades. Because we we like to to explore games because they're a model for many things that also happen in the real world. They're easy to simulate. Uh you can even have self-play. You don't need to play against a human. The AI can play against itself and become better. That's what they did with AlphaGo. Um, but this this uh engagement of people is actually very relevant if it comes to education, and that's a big challenge of how to motivate uh a student just enough. I mean, you you can't do that. Exactly, just enough that just enough, that's and and translating this in a formula is very difficult. Um so that brings and that's hyper-personalized and even hyper-contextualized. Yeah, so in an e-learning context, exactly the student model, that's quite okay, but then the teacher model, how much you want to challenge the the student so that the student feels uh engaged and willing to learn, but is not disappointed. Um that's that's really difficult.

SPEAKER_02 34:30

But it's uh a very interesting route as well for some of the work that we're doing on change management and on training and so on, because it should be a reinvention of how we bring in and help people to get trained at their own pace, in fact.

SPEAKER_03 34:45

It is the martian, don't go for big bang innovation, it's step by step by step and challenge by challenge by challenge. We should do that for change management, we should do that for gaming, or we do that for gaming, but we should also do that in education. It's it's the same thing.

SPEAKER_05 35:00

That's the accessibility again. So, have you ever tried to put an entire PDF of a book in ChatGPT

Dark Side: Gaming, Engagement, And Addiction

SPEAKER_05 35:06

and ask to interrogate yourself? Like talk with uh uh ChatGPT, learn from it, like the barrier to start studying with AI, it's almost a companion to study. Like, can you imagine how beautiful that is? Like back in the days, you needed to print your PDF, you needed to uh read everything.

SPEAKER_03 35:25

I'm old enough to needed to go to the library.

SPEAKER_05 35:28

Okay, sorry for that. But isn't that amazing? Isn't that really lowering the barrier? Is that isn't that super accessible that like children that want to talk while uh studying now can talk with ChatGPT and and learn in that perspective. I think that's really amazing.

SPEAKER_02 35:44

But but that's actually still I mean, you see a lot of um I would say uncertainty, fear, doubt as well about AI. Whereas I mean you say, oh wow, great, and look at the possibilities. This is a difficult balancing act within not only I would say organizations, companies, but also I would say the broader human scale because of the accessibility, it's in everybody's daily life, uh almost uh Yeah, but it's always the same when something new comes in our lives, we always have the five stages of grief.

SPEAKER_03 36:16

We deny it, we get angry about it, we start bargaining about it, then we get into a depression, and then only then are we going to try to accept it, and once we've accepted it, then we're going to see the bright side. And some people are still in the anger phase, some people are still in the denial phase, some people are still in the bargaining phase. Okay, it depends on but in the end, you have to go through the five stages, and then in the end you're going to accept it. We've seen the same with the internet, we've seen the same with so many new technologies, and this is just a new technology.

SPEAKER_02 36:48

It's just a new technology, but at a very fast pace. Um maybe a bit in in your angle of uh reinventing certain businesses, but uh there was a study that in the UK 60% of the accounting firms see their clients coming with already advice that they searched on the AI platforms uh and then asked to validate it. Uh is this something that you see happening in a in a broad way? Is it uh I think we did that study. Oh, you did a study?

SPEAKER_05 37:23

Oh, okay, okay, yes. Like, yeah, we really did, and we now use it for marketing purposes. Okay, so it's a really study that we've done. Actually, I can share it with it is it's like the Dr. Google 2.0, basically. Like the past decades, everybody went to Google, asked a question, and came with an answer. Today, I think so. We did the study with like more than 500 decision takers of like larger accounting firms in the UK uh to look whether or not they already see these type of patterns. And they they they said yes, we are confronted more and more that we need to give almost a second opinion. Like it's not Dr. Google anymore. It feels like a second opinion. Yeah, yeah, validate, and uh they uh it's it's quite interesting, I think, I find, because they they have the feeling that the client is better educated. Probably it's not, but they are they have just better materials, like just more materials, maybe, because it's a longer answer than an answer in Dr. Google. But they almost need to give a second opinion. And in that second opinion, I think don't uh held it against me. I think 50%

Education, Personalization, And Study Companions

SPEAKER_05 38:28

of that second opinion, they need to give additional context of the client case itself or legal information that's missing in in the answer itself. So they they become more and more reactive in that second opinion. So I think that is that's like that's because of that accessibility. Like everybody throws something in ChatGPT just to ask something and comes out with an answer which we all think it's 100% correct. So and then they ask for reconfirmation. Yeah.

SPEAKER_02 38:54

So it it shows in fact how businesses professional services are are evolving.

SPEAKER_03 38:59

Yeah. Um, I gave my um um my private banker a heart attack a couple of weeks ago because um I asked ChatCPT and then came with the answer of ChatCPT to my private banker, and he thought I was going to another private bank. That's it.

SPEAKER_04 39:18

Um it feels like you're educated, you know everything. Yeah.

SPEAKER_02 39:23

Maybe uh a little bit more forward thinking. Um I would say look five or ten years ahead, but with the current pace of acceleration, I follow this daily podcast that I mentioned already to a couple of people. I the speed at which new things and new capabilities are being uh mentioned, it's it's uh enormous. I know that you bought your crystal ball at Amazon, you can't predict the future. But is there is there anything that you I mean uh if you look at the singular singularity university, they're predict that's actually the Yeah, but they thought it would have been they they always said 2043.

SPEAKER_03 39:59

They're absolutely wrong about that. It's going to happen way faster than that.

SPEAKER_02 40:03

So that that is the the human intelligent and and uh and the AI or the super intelligent or the what will be at the same level or uh well surpassing.

SPEAKER_03 40:10

Yeah, but it even that is is a non-discussion because it it's it's again it's alien intelligence. Um and it's it's so difficult.

SPEAKER_02 40:18

But if you look at the next TS, do do you see something where you think that's uh okay, there will be a limiting factor, or there may be acceleration?

SPEAKER_03 40:26

I mean, some people say robotics uh or humanoids type of uh in in 2020, um, and and we all know him, it's Peter Abel. I was with Peter Abel in at Berkeley University, and Peter had always been a guy that was very careful with AI. He said, Don't over don't be over excited, don't expect too much. And then in 2020, all of a sudden said it's going to happen in between now and five years' time. In five years' time, AI is going to be as normal in your daily lives as the internet is today. With 2025, I was absolutely right. Uh but then in 2023 I saw him again, and

Second Opinions And Changing Client Behavior

SPEAKER_03 41:05

that was just after the dawn of Chat CPT. And Peter had been involved in the early stages of ChatGPT, and he said, I'm so surprised about what it can do. So even the guys that are in the center of what is happening there, and most of the guys that are something in AI have been PhD students of Peter Abel, even he is surprised time and time again. And one thing that he refuses to do is he says, I can't even predict what's going to happen next year. So don't ask me what's going to happen in the next five years. I cannot. Now, if he is in the center of that universe and he doesn't dare to predict it, don't ask me to predict it.

SPEAKER_02 41:49

Well, I guess that everybody has imagination. It's probably difficult to predict the speed.

SPEAKER_00 41:54

But maybe to come back to what you said, uh it's completely true that um even the people who supported the uh approach that was taken by these large language models, uh, it's just one methodology within AI. And the people who who are following other approaches who have other views on how language models uh should be built, everybody was indeed surprised. So nobody had was able to predict that it would be that performant, uh even if it hallucinates from time to time. Uh it it surprises everybody. Um that's that's completely correct. The the difference with robotics is of course that you need a robot. You know it it's uh it's not just software, it's it's more difficult to uh replicate the robot, and and it's uh a lot of engineering effort still in these robots. So it's not only about the AI, it's also everything around it. So as soon as AI is what we call embodied, it also depends on other technology, on on uh having um powerful batteries, good materials, uh things like this. So sometimes the AI idea is already there, but it needs to be having a body, and and that also needs to be developed.

SPEAKER_02 43:19

So my next panel conversation would still be with humans.

SPEAKER_00 43:22

Uh yes. Okay.

SPEAKER_02 43:24

I'll send my avatar.

SPEAKER_00 43:26

But that's software.

SPEAKER_05 43:28

I I think for the the software industry, the coming decade will totally reinvent software. So if you look at the the presentation the Microsoft CEO has given, I think now a few months ago, um he's preparing us for an agentic industry. So he's probably not allowed to just share his ideas, so he's basically preparing us for an agentic industry. And basically, what he's saying is that uh over time, and we don't know when, is that next year, is that five to ten years? That's the big question. Like the agents will have. Have all the intelligence. That means that all system of records that we know will

Forecasting The Next Five Years

SPEAKER_05 44:05

just become the crut databases, like just entering some information. But all the intelligence itself will be with the agents. That means like the way how we look at software, like let's run a query, let's click this button, let's update that, will be totally reinvented. I don't think we will have that much user interfaces anymore. We will have a system of agents that helps you with sharing like background of a client, a specific advice that they gathered, but we will not need to run through all the system of records. It will become basically what we always wanted: system of actions. And that's what he literally says in his presentation. We're in need of system of actions that actually become your super brain and help you with your uh your ABC. You're always I always forget it, all your boring and complex.

SPEAKER_03 44:53

For me, that's good, that's good news. If we don't have to do the boring stuff anymore, okay, and most of the problems that we need to solve on the global scale are wicked problems. Wicked problems are problems with a lot of variables that are all interfering with one another. Our human brain can't cope with that. Give a human brain four variables, and our brain melts. And in most of the wicked problems that we need to solve, we need to cope with 20, 30, 40 variables. Look at climate change. So we are going to need AI to help us in those complex issues as well.

SPEAKER_05 45:29

100%.

SPEAKER_00 45:29

Yeah. And it actually goes uh in two uh facets. It's it's AI that helps us to simulate possible worlds.

SPEAKER_03 45:39

Exactly. It can simulate a million simulations in no time.

SPEAKER_00 45:45

And it also can help us how to optimize this system.

SPEAKER_03 45:48

Exactly. Yeah.

SPEAKER_02 45:49

So for us it's uh We're not good in optimization.

SPEAKER_00 45:51

Yeah.

SPEAKER_02 45:52

Before going to a close, any other elements that you want to share with this audience, knowing that this team is kind of like a group of leaders that actually is helping the next generation as well of our company to survive and thrive in in the next decades.

SPEAKER_03 46:08

It's simple. Imagine whatever, you can build it. And I really mean this. This is this is what is science fiction, what you think is science fiction. We have this superpower for the moment, we have this magic stick. Most of what we imagine can be built. It might take six months, it might take six years, but we're going to be able to build it. I'm I'm absolutely convinced about that. So I do a pre for more imagination. Any additional thoughts on that?

SPEAKER_00 46:39

I fully agree with that, but I also find it interesting in your presentation. Uh at some point you said, okay, people should also be able to ask questions. Uh so I'm fully agreeing that people should start experimenting AI, don't be afraid of using it. But on the other hand, don't be afraid also to ask two people who know how AI works if something doesn't turn out how you expected it, because we also

Agents, Systems Of Action, And Wicked Problems

SPEAKER_00 47:06

see that people try to build something, they forget about a certain aspect, then they say, Ah, it's not working. Uh, but then yeah, if you look at it, we say, okay, but maybe you forgot this or this. So asking questions, asking for help uh for people who might have the answer, besides ChatGPT, uh, I think that's also very important.

SPEAKER_03 47:28

I fully agree.

SPEAKER_05 47:30

More than ever have a growth mindset.

SPEAKER_02 47:32

More than ever. A growth mindset, yes. Okay, so maybe a couple of points, and I just want to check whether you agree with uh those and a bit of things that I distilled out of the conversation. Um one is there's a huge potential, positive potential with AI if we use it the right way. Agree?

SPEAKER_05 47:56

Definitely agree.

SPEAKER_02 47:58

Second, the change impacts is potentially much bigger. The kryptonite elements is we should not underestimate how to get organization, but also people to get people along in that journey, because it's not something, it's not just a technical matter. It's uh a thing about moving a society, uh people, organizations. It's probably one of the more complex elements in the change. If you look at existing uh organization, I'm not talking about a startup, but the technical element is complex, but having creating the the change and and and the move that's probably uh the more difficult area. Is that I agree as well?

SPEAKER_03 48:38

Change is about people, and I I had it a couple of times in my keynote. It's about storytelling, be good in storytelling because you need to engage people. Um change is not a system. You need a system to help you. People need to believe in the change. I believe in the formula to change. There is always a natural resistance for people to change. People do not like to change. And that's not because they don't like to change, it's our nature to be not inclined to change. So you need three components to overcome the natural resistance for people and companies to change. There needs to be a compelling reason to change. But most of us wait until we're on a burning platform, and then it's obvious that's a red that's a red box, then it hurts. And then you say, now we need to change, but then you're actually already too late. Um then you need to develop a vision. What do we want to accomplish? And then you need to translate it into first steps. But I prefer to start with a vision. Start with that dream, tell people if we don't realize that dream, we're going to end on a burning platform. So, how are we going to avoid ending on a burning platform and translate it into this is step number one, this is step number two, and this is step number three that you need to take and repeat it, repeat it, repeat it. Be the best storytellers on the globe, because otherwise you're never going to get your people to

Imagination, Help Seeking, And Growth Mindset

SPEAKER_03 50:00

believe you and to be feeling engaged, and change will be something on paper, but not in reality.

SPEAKER_02 50:06

Yeah, yeah, yeah. And I like the like the visual of Patrick this morning with the burning platform with the calls, the the fire. If you just stop and and and stare, if you're on a burning platform, people will run all directions. Um, and the last element is I mean, there's a crucial role for leadership for the people here to to make sure that the whole organization is not only thriving and again uh benefiting from all the possibilities that there are, but also making sure that the organization is, I mean, the people are taking with us, but also that we apply AI in the right way. I mean, the trust in AI, how we do it with the in the ethical way, and so on, also agree that that is a crucial part of uh of the the journey going I the journey on ahead of us.

SPEAKER_00 50:54

Yeah, I would say that every company has to have its values and and know what it stands for, and that's important, of course, if you then also use AI that's that you translate that.

SPEAKER_02 51:04

Yeah, yeah, perfect. With this, I think I want to thank the panel. Thanks for the hugely interesting uh insights and for your presence. I want to thank the audience for listening to us and uh enjoy the rest of the day. Thank you.