This week on The Digital Download, we're diving deep into the transformative world of AI and big data with our special guest, Evangelos Simoudis. A seasoned investor, technologist, and author, Evangelos brings a wealth of experience from the heart of Silicon Valley. Having worked with startups and global corporations across various sectors, he offers a unique perspective on the practical applications and strategic implications of AI.

While many companies are eager to jump on the AI bandwagon, Evangelos argues that most are approaching it the wrong way. He'll shed light on the common pitfalls and misconceptions surrounding AI adoption and offer insights on how businesses can truly leverage its potential.

Join us as we discuss questions like:

• What are the biggest mistakes companies make when implementing AI?

• How can businesses identify the right AI solutions for their specific needs?

• What are the ethical considerations surrounding AI in business?

• How can companies ensure responsible and transparent use of AI?

• What is the future of AI and big data in the business landscape?

Evangelos emphasizes the importance of aligning AI strategies with business goals and understanding the broader implications of this technology.

We strive to make The Digital Download an interactive experience. Bring your questions. Bring your insights. Audience participation is highly encouraged!  

This week we were joined by our Special Guest -

• Evangelos Simoudis, Founder and Managing Director of Synapse Partners, author of many books, the last one being The Flagship Experience

This week's Host was -

• Bertrand Godillot, Founder and Managing Partner of Odysseus & Co, a proud DLA Ignite partner

Panelists included -

• Tim Hughes, CEO & Co-founder of DLA Ignite,

• Adam Gray, Co-founder of a DLA Ignite

Transcript of The Digital Download 2025-03-07

Bertrand Godillot [00:00:07]:

Digital download. Oops. Good afternoon, good morning, and good day wherever you may be joining us from. Welcome to a new edition of the Digital Download, the longest running weekly business talk show on LinkedIn Live. Now globally syndicated on TuneIn Radio through IBGR, the world's number one business talk, news, and strategy radio network. Today, we're diving into the transformative world of AI and big data with our special guest, Evangelos Sy Mudis, A seasoned investor, technologist, and author, Evangelos brings a wealth of experience from the heart of Silicon Valley. Having worked with startups and global corporations across various sectors, He offers a unique perspective on the practical application and strategic implications of AI. But before we bring Evangelos on, let's go around the set and introduce everyone.

Bertrand Godillot [00:01:12]:

While we are doing that, why don't you in the audience reach out to a friend, ping them, and let them join us? We strive to make the digital download an interactive experience, and audience participation is highly encouraged. Adam, do you wanna kick us off?

Adam Gray [00:01:27]:

Hi, everyone. I'm Adam Gray from DLA Ignite. I'm cofounder and Tim's business partner. And, AI, a very topical conversation to be having and something which is on everybody's radar, either because they're wowed by the possibilities that it presents or terrified by the possibilities that it presents. So, either way, it should be, an entertaining show.

Bertrand Godillot [00:01:53]:

Definitely looking forward to it. Tim.

Tim Hughes [00:01:56]:

Thank you. Welcome, everybody. My name is Tim Hughes. I'm the CEO and cofounder of DLA Knight, and I'm really looking forward to the discussion today. I think that, AI has, provides us with many opportunities as much as it provides us with many challenges. And I'm famous for the fact that I wrote the book social selling techniques to influence buyers and change makers.

Bertrand Godillot [00:02:20]:

Thank you, Tim, and myself, Bertrand Godillot. I am the founder and managing partner of Odysseus & Co, a very proud DNA Ignite partner. As I said this week, on the digital download, we'll speak with Evangelos. While many companies are eager to jump on the AI bandwagon, Evangelos argues that most are approaching it the wrong way. This sheds it'll shed the light on the common pitfalls and misconceptions surrounding AI adoption and offer insights on how businesses can truly leverage the potential. Let's bring him

Tim Hughes [00:03:03]:

on.

Bertrand Godillot [00:03:05]:

Evangelos, good morning, and welcome. And we know it's an early co an early wake up call for you today.

Evangelos Simoudis [00:03:11]:

Well, first of all, thank thank you very much for having me on your show, and it's good to see, all of you, and I look forward to this conversation.

Bertrand Godillot [00:03:21]:

Great. Evangelos, let's start by having you tell us a little bit more about, you, your background, and what led you to where you are today.

Evangelos Simoudis [00:03:30]:

Great. I'm originally from Greece. I came to The US in, '78 to study, and got involved with AI in '82 and have been in various roles involved with that field since that time. For the past ten years, I run a firm called Synapse Partners, and we do two things. We invest in early stage startups, that develop AI applications for the enterprise. I've been in venture investors since roughly February. And the second thing we do is we advise corporations and governments on the use of AI. In the process of doing both of these things, we develop some of our own IP, AI IP.

Evangelos Simoudis [00:04:19]:

Sometimes we spin it out in separate companies. Others, we we use it and license it to our corporate, customers. We have written three books in my, quote, unquote, spare time, because I've been fascinated with the topic of, how AI can play a role in mobility, both vehicles and and other forms of of ground mobility, and, continue to have the the bruises from, working in this in this area.

Bertrand Godillot [00:04:52]:

Excellent. Well, thanks for that. So, Evangelos, let me let's start with the foundational question. What are the biggest mistakes companies make when implementing AI?

Evangelos Simoudis [00:05:06]:

So, I, I've broken this down into five different categories of problems. The the first one, which I've seen over and over, is that, in most companies, and I've seen it not not in this in the current AI wave that we're going through, but, even in in the previous ones that I participated, as I said, since the eighties, we tend to look at it as a technology first as opposed to problem first. Right? Or or, what is the the issue that I'm trying to address? And we can talk about that. But so the first and foremost is that we're not we're looking technology first versus problem first. Then, other problems involve culture. A a lot of companies want to, introduce AI. They want to adopt AI. They they don't have the right employee culture.

Evangelos Simoudis [00:06:01]:

They don't have the right, overall corporate culture for it. Their technology on the technology side, you know, right now and and for the past several years, AI has been data driven. And for many corporations, they they want to apply, artificial intelligence, but their data is not in the right in the right shape. And believe it or not, actually, this is one of the the most, difficult issue to to speak to to CEOs and say that after all of these years that they have been working on their day their operations have been working on their data, their data is still not in the right in the right shape to, to be used by the current, AI approaches. There are talent issues. There is a war of talent out there, and, this is, very much both for the, companies that create the technology, but more importantly for the corporations and the governments that want to, to adopt it. And then finally, particularly in this in the generative AI way that we're going through right now, the the the cost of implementing and deploying these solutions, is is not yet, fully appreciated. I think partially because we have a number of chatbots now that we can use for free.

Evangelos Simoudis [00:07:34]:

But when you try to bring this to the enterprise and and deploy that type of technology, the the costs are significant. So, to me, those are the the the top, the top issues. And if I could kinda wrap up my this initial question with a couple of examples, kind of one good one, one not so good. So on the good side, if you look at a company like Klarkner, which is a Swedish financial services company, they started applying, generative AI because they said, we want to, to to reduce the cost of providing customer support and improve the quality of of our customer support. And they're famous for letting all of their customer support employees. I think they laid off about 700, customer support employees in the process of bringing in. But but before they took that step, they they define their problem very, very well, and they started to understand what the AI technology can bring to address this problem. On the flip side, I will say if you look at, what many automakers the approach that many automakers have taken with regards to autonomous vehicles.

Evangelos Simoudis [00:08:52]:

Again, autonomous vehicles are great consumers and users of AI, at at various levels. But, in in most cases, these, the automakers took a technology first approach without really understanding, are we ready to do it? Do our customers want it? Do our employ can our employees deliver? And as a result of that, we have, billions of dollars being wasted as a, and and still, a failed, failed effort. So so this is, to me, these two examples, are probably front and center of of the issues that I was talking about.

Adam Gray [00:09:36]:

So if you if you if we think about that, Swedish financial services company, they, they laid off, staff, a large number of the customer support staff, because they were looking for increased efficiency, and they were looking for, as so when we say increased efficiency, we mean reduced costs, I assume. And, and they wanted a higher quality of service. Now given that, I'm a consumer of your product, you're you're the financial services company, I'm a consumer of your product. I fired up because I have a question or or a concern. You know, my pension seems to have gone down this year. What's going on? Can you explain this to me, please? And, actually, part of what I want is I want reassurance. And reassurance will not be created by me engaging with a chatbot. No matter how clever the chatbot is, it'll be, given by me engaging with you as another human being.

Adam Gray [00:10:46]:

And and I think that that there's no. No. I don't think is there an existential crisis looming? So a financial services company has, has a series of algorithms and bots that work out where it should invest its money, where it should place its bets to to drive the best possible return.

Evangelos Simoudis [00:11:07]:

And the bit

Adam Gray [00:11:07]:

the bit that it's been doing Sorry.

Evangelos Simoudis [00:11:12]:

So so, this is why I said, you know, understanding the the the company's needs and then, bringing in the technology. So in this particular case, Klarna, which is the the name of of this company, is a financial service company that is involved in the area of what's called, buy now pay later. Yeah. Yeah. So, the and they have a specific type of clientele. And this is important because when when we think about understanding, you know, putting the problem first, and and understanding the capabilities of of your organization, you you need to an you need to bring together kinda like three components, if you will. You you need to bring together the organizational needs. In this case, I want to provide better customer support.

Evangelos Simoudis [00:12:03]:

I want to, and how you define that that customer support. So in in their case and based on the on what they understood of their customers, which is the second component, speed of of the of the response, is is important, and accuracy of the response is is important. Right? But because they have a specific type of customer, so let's say the second component is what are the customer expectations. K? And in their particular case, as it turns out, they understood their customer base very well. Their customer base was ready for something like that. The third component, by the way, is the organizational capabilities. And they found that they have both the technology expertise and the business process expertise to to to, both implement and impact such a solution, bring such solution to to their to their customers. And I think that a lot of times, corporations just think of the technology side.

Evangelos Simoudis [00:13:05]:

They ignore most times, actually, they ignore even the capabilities of their customers. Right? In other words, what their customer is able to absorb and what the customer is willing to live with. But but, it it it is, even though I I talk about them as a financial services company, it will not be the same as, say, going to a Goldman Sachs, which which also is applying a lot of AI, but is using a very different approach to to how that, AI technology to to which problem it is used to address and how it is used when when customers are involved.

Adam Gray [00:13:47]:

So so AI is, in this instance, it suits is well suited to delivering simple answers to simple questions for high volume, low value

Tim Hughes [00:14:00]:

transactions. The the Klarna the the Klarna case study is very interesting because what Klarna did was that the CEO basically stood up and said, I don't need any people. So they he sacked all the people, and within sixty days, he reinstated the people because of the fact that he realized that AI can't do all the jobs and that humans are invaluable. And that was the conclusion that they basically got, within sixty days of sucking all the people with an AI.

Evangelos Simoudis [00:14:29]:

Well, I will also push back though and say that, I I believe that, and this is not a AI can solve difficult problems provided that you're doing the right mapping between problem and and the capabilities of the technology. So I I think it's unfair to say that it's only used for simple problems. So if I look at, you know, the the the first AI spring, whatever you will call it, I participated in was back in the eighties. And at that time, we had rule based, expert systems, which were extremely good at for diagnosis problems. And and people, who corporations who who who did that kind of mapping, who identified the right diagnosis problem and then apply the technology to address it did extremely well. However, there were a number of other companies that were trying to shoehorn a problem to technology that was not that was not good for. Right? And and in those cases, we had failures. Later on, in the early nineties, when we had the second AI, another AI, spring, the, when neural networks started being introduced, so called statistical AI started being introduced, The process industry had a very successful solutions because these these technologies were extremely good at addressing nonlinear regression kind of problems.

Evangelos Simoudis [00:16:12]:

It wasn't other companies that try to apply or other industries that try to apply the same technology on different type of problems didn't do as well. Right? So so this is why I keep saying that we it is important to understand what problem you're trying to solve, what are the capabilities of your organization, what are the capab and then start bringing the technology, to, to address it. Every step of the every time we have an AI spring, we're making important advances. And and the advances that we're making these days with generative AI are extremely important, and and we're we're finding that we're able to to deal to address a a wider, spectrum of problems. But that does not mean that every problem will be solved with the same with the same technology.

Bertrand Godillot [00:17:08]:

You you mentioned, in in the five areas, Evangelos, you mentioned you mentioned the fact that, most of the time, data are not in the in the right shape. Obviously, we talk so we we talk initially about, you know, the tech first versus the the the problem first, and I think it's it's pretty clear what we you know, how how you should approach this from the problem perspective before you get into finding the right tech. And there is obviously the data aspect that you mentioned. I'd love to cover the talent one just after this, but what what are the your your common, I would say, discovery or basically the the the the common landscape that you're faced with with your customers on on that data front?

Evangelos Simoudis [00:18:03]:

So, I feel that we, we have to educate the customer that, even though they have spent a lot of time to get their data ready for business intelligence and and maybe even for discriminative AI, the when they try to apply generative AI, which is the core of the of the current, AI wave AI spring, the data needs to be, in a it needs to be labeled differently. It needs to be in a in a different shape. Textual data takes a lot, a lot bigger role. And, that is not always appreciated. The the point I was trying to make is that this this is not always appreciated because the, you know, CEO will say, well, but I spend, you know, so many million with these firms to to get my data ready. And, trying to to convince them that there's a difference between, putting everything in the data warehouse so that you can run a report, versus, you know, using the data to be able to to summarize your, your documents, for example. It's a very or or, we have one of our portfolio companies, is working in the, in the materials space, and they're getting a lot of, bill of material documents with with, schematics and text and, all of that needs to be properly labeled if it is to be taken, in, advantage by a generative AI, system. And, which is very different than if I wanted to create, for example, a classification model, which which, again, may be a lot easier to create than than to build something using, for example, build a system that will, that will provide, accurate customer support, to, to a range of customers.

Bertrand Godillot [00:20:21]:

Okay. So so so we talked about, you also talked about culture, and I'm quite interested in this one. You said you need to have a specific culture. How would you describe that culture?

Evangelos Simoudis [00:20:37]:

Well, I mean, look, we're we're talking right now, about systems that are becoming increasingly autonomous. And, I think we can they can range from, being assistance to to to to performing tasks fully autonomously. So let's take again an an automotive assembly line. I may have a robotic system that can act as a as a an adaptive robotic system that can act as an assistant to me. Right? You have, I'm sure you read about, Tesla's, Optimus, humanoid robots. Other companies are doing things like that. The the culture in the company needs to be able to to to be such that, they they can collaborate. They they they are willing to collaborate with with such a with such an entity.

Evangelos Simoudis [00:21:43]:

We're having we're seeing generative AI systems that are becoming extremely good at programming. And you need to be able to have a culture of people both willing to to use such systems, but more importantly, and as we saw in one of the financial services companies we worked with, they they need to be aware that these systems are not, flawless. And and to have the the culture to always question the quality of the answers that they're getting, and and being able to, either correct it. Correction could mean to improving improving the system, but also making sure that, whether it is quality of code, whether it is quality of an answer, that they do not go to their own to their own people. So so that that to me defines the culture of, that that I'm talking the AI culture that I'm talking about. I don't want to I mean, I'm sure we will see examples of, companies where, this type of technology may be sabotaged because there is employee resistance. K? I don't go there. But but I think even if we were to take the the best case scenario where employees recognize that there is there is help here, they would be able to incorporate it productively into their work.

Evangelos Simoudis [00:23:20]:

I'll tell you one of the problems that that that we're seeing and and we saw it in the particularly in the earlier phases of, this generative AI wave. Employees I mean, people started using these free chatbots at home. Okay? And they were finding that they could do, interesting tasks for their home environment using these these tools. Create the recipe, to do x. Give me a a a an itinerary for vacation in Venice. You know? And then they will go and they will go to their company, and and they think that they they could do with the same means. They could use the the same tool to do something that is, very, specific to their company. Now if, if I do a recipe and rather than putting three teaspoons of olive oil, I put four, no big deal.

Evangelos Simoudis [00:24:18]:

Right? If I try to do the same thing in a biotech company, there might be a lot of problems that are created. There are definitely advantages for for having this consumerization of AI. This is not something that we had in the previous AI sprints, for example. In fact, I I was saying to a CEO, corporate CEO, recently that the last time that I saw something like this was, I used I used to run IBM's AI division in the nineties. And, I remember, visiting, customers, and they will that time, AOL mail was just being introduced from mid nineties. And and they would say, well, you know, my employee came in and they told me I have this, email at at home that allows me to communicate with my mother or whoever, you know, very easily. Why can't I have the same kind of email at work? Right? And I think we're seeing something similar right now with AI, particularly with generative AI, where employees use it at home, and they say, well, this is really fantastic. I mean, I can ask you to write a a a message to my to my son.

Evangelos Simoudis [00:25:29]:

I mean, it it does it immediately, and it's pretty good actually. Right? Why can't I use it to to summarize my, policy document at, at work? And that's where the the problems start occurring because, you know, writing something that has the the the seal of of a corporation is very different than writing an email that has the seal of your of your household. And and and but but, again, these to me are at the heart of the of the culture issue that that I'm that I was talking about.

Tim Hughes [00:26:02]:

I think I think you're right. I mean, we we all have seen its situations. I mean, the classic one was bring, bring your own device, which is, when, you know, when I was at work, the company provided me with a mobile phone. And then Blackberries came along, and I said, I don't want a company mobile phone. I want a BlackBerry. And they said, we can't use a BlackBerry because because the email system is different. And we've had exactly the same, haven't we, which is where we've we we're all able to use AI systems. We've got access to them for free.

Tim Hughes [00:26:31]:

But we've come to work and immediately put confidential information into, a public, LLM, and, and then it's available to everybody.

Evangelos Simoudis [00:26:42]:

Right. So so I I think famous examples of this are relating to the the so called rag, retrieval of metadata generation that that is used to fine tune, these large language models. And and a lot of times, employees, especially earlier in the day, will use, will expose this kind of, proprietary documents with proprietary information in the in an effort to, fine tune, a model, and and create risks for their for their corp, for their corporation. By the way, when I was, talking earlier about costs, these are some of I mean, again, we talk about risks and and culture. There are huge costs associated with how do I fine tune a model in order to make it productive for my for my use case. Right? So now I've identified the the problem. I know that this type of technology will be appropriate for solving the the problem. Now I have to worry about, you know, how much is is this gonna cost.

Evangelos Simoudis [00:27:53]:

And, again, there there is deception. You know? When I use it at home and it's free, I don't think about, well, what what will it cost if I were to to take, you know, a million documents and try to push them through so that I can then generate whatever answers I wanna be generating to my millions of of customers out there? That's that's something that is still that is just now starting to to dawn on, on corporations.

Tim Hughes [00:28:22]:

But from a I mean, a lot of people have been using it. You know, they've they've created their own internal ones. Some of the leading edge companies have created their own LMMs to to use, and they're also using it as a way of accessing knowledge bases, aren't they? So, so if you wanna get, you know, in the past, you used to have to go as an employee, I'd have to go to some intranet, and search for something and not be able to find it. And now you're actually able to go to the, the internal LLM and just basically put the question in, and it does the searching for you. So it saves so much time.

Evangelos Simoudis [00:28:57]:

It it does. And, again, I I I will I will I will say the following. First of all, I think the how to use or what is the impact of chatbots on on the search workflows that we have created is going to be a very interesting disruption. And I'm very certain Google and and other companies that, make their living primarily from search are spending long hours trying to think of what what this is gonna be. On the flip side, the and now we have a number of these foundation models with their associated chatbots. The, we we need particularly, again, since we continue to talk about, corporate environments, employees need to be trained to not to confuse eloquence with accuracy. You know, we use three or four different, chatbots in our own advisory work, and we we have a lot of problem. And we use the the paid versions, by the way, not not the, not the free versions, which are supposedly better, more accurate.

Evangelos Simoudis [00:30:26]:

But we continue we we we constantly find problems with the results that they they return. And and if you, again, as I said, everything is presented in such an eloquent way that is very easy to say, oh, this must be right if if it's written like this, but but it's not always, the case. And, again, having the, having the ability and and having the, the desire to to, to check on the quality of the answer is is an important, is an important criterion here for for the use in the corporation.

Tim Hughes [00:31:01]:

We had a we had a guy on here talking about AI who, said that he'd, his daughter would come home with this maths problem. And she said it's really easy that you just put it into, ChatGPT and it spits out the answer. The teacher knew that it spat out the wrong answer. But and and and so we then have this discussion this discussion about what is truth Because because when ChatGPT pushes out the wrong answer, that's deemed as being the truth, whereas it it was the wrong answer.

Evangelos Simoudis [00:31:35]:

Right.

Tim Hughes [00:31:36]:

And and and and and and so what was happening is was that there was this discussion amongst all the the children about, you know, dad dad dads were basically being applied to see what the right answer was. But they were saying because Chat GPT, this is the right answer, and, of course, it's not.

Evangelos Simoudis [00:31:50]:

It's not. Right. No. I mean, interesting

Adam Gray [00:31:53]:

a really interesting physical example or or visual example of that is apparently if you, if you ask an AI agent to create you an image of a clock or a watch saying, you know, just before, just after midnight, in fact, any time, it always comes back with a picture of a watch at ten past ten because that's the most attractive look. So almost every picture on the Internet is of a watch at ten past ten because it's the most visually balanced look. So no matter what time you ask an AI agent to produce, it always comes back with an image of ten past ten. Yeah. And and the problem with this, obviously, is that that because the nature of how a large language model works, it looks for patterns, doesn't it, in the way things happen? And everything drives you to a particular outcome, and that outcome may or may not be correct.

Evangelos Simoudis [00:32:45]:

Yeah. So, we we have made so you touched upon several different interesting topics. So let let me let me try to, untangle this a little bit. First of all, there's no question that, we've made tremendous progress through neural networks in all their incarnations, deep neural networks, convolutional, and all of that. And, in the in the current the the the the current, AI spring is, largely based on the advances that we have made through deep neural networks. But as you correctly pointed out, these techniques, no matter how complex they are, they they look for patterns. And and the world is I mean, you can do a lot by looking at by identifying patterns, but, there are many times that you don't want. In fact, it is interesting if the search problem is a very good example of that.

Evangelos Simoudis [00:33:53]:

When I'm when I'm searching for something, I'm looking for something very precisely. You know? I I want it to be correct, which rather than a pattern. Okay? So so how, for example, a company like Perplexity, which is using this, generative AI in in the search problem, how they approach their their problem compared to how OpenAI, which is has a foundation model, is approaching the problem, is is very different. Right? I mean, there are several commonalities, but the the end result, is is different because the objective is is different. So, and this applies not only to, to problems that can be addressed using text, but also, as you said, to problems that need to be addressed in using multimodal, models, images, voice, you know, whatever. So so, again, we need to be we need to be very mindful. The the point that we're making frankly to to our customers is that, while there is a a there's gonna be a need for this general purpose resources, this foundation, this frontier models as they're also called. If you want to use generative AI effectively in your corporation, in your organization, you really need to be thinking about specializations of models.

Evangelos Simoudis [00:35:26]:

So we call them XLMs because they might be large language models, but they might be small language models also. I mean, it it doesn't mean that you need, in the example that I often use, to to convey this point is, if, if all your company is doing is providing financial advice, we started this conversation today, You don't need the model that knows Roman history. You don't need a model that that understands biotech. Right? Because having this model or having a model that has all of these capabilities makes it extremely large, which means extremely expensive, expensive to maintain, you know, all of that. So, there's no question that there is a there is a need for these general purpose resources in the same way that we have general purpose search resources. But, again, it's it was happening with diagnostic system back in the eighties with, operational research systems in the and and and nonlinear regression systems in the early nineties. Here, with generative AI, we really need to be thinking about how am I going to use this technology for the problem that I that I have, and how much baggage, how how much do I need to carry forward in order to get me to the right to the right capabilities. And, and this is something that is just now being understood.

Evangelos Simoudis [00:37:01]:

So if you look at what Google is doing, with DeepMind, they have developed specialized models for biotech. They and and they're now doing research for, new drugs. They have developed you talked about math. They had developed a special model for, reasoning around math problems, and it's extremely good. But but it's not the same kind of model. It may have the same, the same technology underneath it, but it's not the same model as a foundation model. And I think this type you know, teasing out these differences will be important for the for the corporation to understand as they're trying to say, here's my problem. What's the technology that I need to use in order to to address it?

Bertrand Godillot [00:37:53]:

We have a comment from the audience. So, Adam, welcome. AI is out full with research and the decrease in the amount of time taken. However, you must invest some time in validating the output, and I think we, we discussed that. But that ties to a new question, on my side about skills and roles that you may want to consider to be successful in an AI implementation in your business.

Evangelos Simoudis [00:38:24]:

So I'll I'll go back to, to to Clara and and tied also to one, financial service customer we worked with. So the problem we worked with was how do I accelerate my software development, progress? And could I let most of my software engineers go and and and rely on these kind of systems to to generate the software that that I that I need for my corporation. And, what we showed in the process of doing that is there might be certain opportunities for, reducing your programming staff and making your the remaining staff more productive. But, you need to worry about two things. You need to worry about who is checking the quality of the output. And To do the right checking, you need more senior people. So what and we obviously more expensive. And the second thing you need to worry about is the the the people who are senior today in your organization, at some point, they're going to leave.

Evangelos Simoudis [00:39:37]:

They're going to retire. They're going to go to a so you need to have a pipeline of people from your junior staff that is becoming more senior. Right? So, to go back to to Adam, to a point that Adam made earlier, you cannot willy nilly go and say, I'm going to get rid of everybody in my in my company and replace them with this, with agents no matter what autonomous agents no matter what they are. Because we're not yet at the stage where we can trust, the the quality of of this, of this output. So you, we're trying to to to answer your question, you you really need to have the the the people with the right expertise who can check the output. These tend to be more senior people. And then but you also need to to have the to create the pipeline of talent that will become senior and, will so you will always have in your corporation that body of talent that, will, will make sure that what your what is being generated is of the of the right quality, of the right accuracy, and and for the right for the right audience for the audience that you've chosen.

Adam Gray [00:40:55]:

So so here's here's a question. AI is, doubling in its capability every year, let's say. Probably a bit faster than that, but it's doubling every every year. So what you say about me needing to validate the output of what AI provides me with, is true today, but it won't be true tomorrow.

Evangelos Simoudis [00:41:15]:

Yep.

Adam Gray [00:41:16]:

And and this this begs, a larger question about the role that AI has in society, I think. So, when we had the industrial revolution, people that used to work the land and the mines and whatever else and do physical work, they had to to leave that because one machine could do the road the job of a hundred hundred men, and those people went to work in cerebral intellectual pursuits because that's the bit the machines can't do. So does the advent of AI and and the the exponential growth of AI capabilities, does that mean that we as humans have to go back to doing things that machines can't do? So creating furniture and whatever. Or does it mean that that we all get to live on a desert island where AI goes to work in our place and we don't have to do anything? Because because this change is gonna happen within our lifetime, isn't it?

Bertrand Godillot [00:42:18]:

I love option two, by the way.

Tim Hughes [00:42:22]:

Yeah. So do I.

Evangelos Simoudis [00:42:25]:

In 2019, I gave a talk to to the European Union, to a group in the European Union, about this this point. What do we do as and and that's by the way, in 2019, we generative AI, was cannot burst out in the scene. But, I was, I was worried because, you know, as you said, whether it is generative AI or discriminative AI, you know, the advances are being made, day. And I said, okay. Well, there there will come a time that, entire parts of the of the population will not will not find, will not be needed for for the services they provide today. And, actually, I walked out of that day long session, following my talk, completely depressed, because I, I had come to the conclusion that nobody had given any thought to that, to that potential. And, today, I I will I will make the same statement, frankly. I mean, we're talking about humanoid robots.

Evangelos Simoudis [00:43:49]:

We're talking about, systems that can program automatically. I mean, a variety of different kind of agents. Couple of months ago, I I had a session with, Indian, public company CEOs, who, you know, provide, whose companies provide a lot of engineering resources to corporations around the world. And their question was, you know, what are we gonna do if if our programmers are no longer needed? Right? Because we have, and we're talking about millions of people. And this is not like, 500 people in some corporation. This may be 20,000,000, 30,000,000 people. We haven't thought about this problem, frankly. I'm I'm painting with a very broad brush here.

Evangelos Simoudis [00:44:41]:

I'm I'm sure there are people who are thinking about it intellectually, but I have not seen the I I have not seen the the action or even the preparation for action that, should be whether it is at the government level, at a corporate level, we haven't seen that. And I don't think we have any good answers yet.

Bertrand Godillot [00:45:08]:

I have I have a I have a question for the investor. Because you said earlier that, you know, the CEOs and the the the exec committees you talk to don't really appreciate yet, the amount of investment that they need to put into this. So my question is about, are we into because this is really disruptive. And I think the example you take on on coding is is a very good one because I'm I'm a I'm a almost a daily user of, of this type of, of of of agents, and I'm I'm blown away by what by what they are able to do. So so the point is there is a big disruption. So are we into the area of return on investment, or are we into the area of it's an enabler. We need to go there with regardless of the cost. What is the, the feeling of your customers around this?

Evangelos Simoudis [00:46:14]:

Oh, I I thought you were going somewhere different, actually, but I I'll I'll answer your your question first. So in the case of generative AI, I think the the two major if we can classify the the the two class of problems we're looking at is is productivity improvement and cost reduction. Mhmm. We have not yet seen we I'm talking about, our company. We have not yet seen use cases where it is used as as a new revenue stream, right, for for I mean, to add to the revenue streams of of the corporation. Now there there are, there there are new companies that are being created. Waymo is probably a very good example where you're you're using, this type of technology to to create a new company with a new revenue stream, with a new business model that you you didn't have you you didn't have before. But, if if you look at, you know, corporate we talked about Goldman Sachs earlier.

Evangelos Simoudis [00:47:22]:

We talked about Klarna. We talked about, you know, Unilever. You know, those companies that are applying generative AI, they're they're doing it for productivity improvement and, cost and cost reduction. When when, Elon Musk talks about, Optimus, his human eye robots, coming into the into the assembly line or BMW working with Figur to do something similar, it will be, strictly for cost reduction and productivity improvement. So right now, there is a massive experimentation going on, on on what around use cases that where this technology can be applied. And as I said, a lot of times, the technology is applied before even the use cases is identified. I think we're we're coming though to the point, and I consider 2025 to be a a critical year for that where, many corporations are now starting to ask, how do I go from experimentation to deployment? And and that's where they start coming to face with what is gonna be both the the cost of that deployment, what's gonna be, what's gonna be the talent that will be required, what's gonna be the the the infrastructure that needs to be put in place. So we, we talk we work a lot with corporations that are interested in, what's broadly called ethical AI.

Evangelos Simoudis [00:49:02]:

Right? So they want to understand, as I go from experimentation to deployment, how do I avoid, biases? How do I how can I create systems that explain themselves, that are transparent? You know? And and what is the tooling, that I need to put in place? Tooling, by the way, could be technology related tooling, can be organizational tooling. Right? What kind of do do I need to have an internal do do does my risk assessment committee need to now take, an additional role to, to make sure that, this output is is ethical? So suppose, for example, you're using a generative AI system, to to screen prospective employees, And and a a prospective candidate comes back and says, why did you reject me? Can you explain to me why why you rejected me? The system if you're using a system to do that kind of filtering, you you better have a very good, explanation why that why that decision was was reached. Right? So so this is, so so all of these are investment require investments. Right? This is not going to happen, for free or automagically. I mean, and, and I thought the point that you were going to ask me, frankly, was, you know, we have now thousands we have billions being invested in startups, and and and more mature private companies. What are we going to see? What kind of return investment we're going to see there? And, my my short answer to to that is that we're gonna have a lot of roadkill. And, because right now, you know, there there is not only there is a lot of investment in the field, for companies that may not be solving a a really important problem even though they're using AI. But there's also a lot of sameness, for every, sub sector where we want to apply for for people rather apply entrepreneurs are applying AI.

Evangelos Simoudis [00:51:15]:

We have tens of of competitors. If you look at I mean, you talked earlier about programming assistance. Look at how many I mean, in our database, in our firm's database, we we may have, 35, 40 startups that are working, on developing programming. And each one of them, by the way, has taken serious money, in terms of amount. If you look at customer support, tens of of start you look at MLOps, tens of start ups. I mean, so so this, this will have to, we'll see where it goes, but I I I think the the ROI on that is also going to be an interesting one to, to follow.

Tim Hughes [00:52:02]:

I think we're also seeing with AI, aren't we, where people are coming out and saying, hey. This is the new best thing. And the next day, people have actually leapfrogged that, and and and that's actually now old tech and and and it's a new thing. And then the next day, there's a new thing. And

Evangelos Simoudis [00:52:17]:

Right.

Tim Hughes [00:52:17]:

I mean, if, you know, if you look at, I've got my name down for the Perplexity, web browser purely because, I can, because, you know, what perplexity are trying to do is jump over what Google are doing. But, you you you know, Google have got billions of of of, of, money and resource. So they're they're they're obviously not sitting there with their feet up. So it's and and the other the other thing that's interesting is the is the move from, software as a service to service as a software. And, you know, OpenAI created that new thing that does your calendar, and it's supposed to be pretty rubbish, but it's version one. Yeah. So so so so people are already looking at ways of of automating, stuff. Adam and I had a meeting with a guy, last week, and he said to us, he said and he's about 29, 30, Adam.

Tim Hughes [00:53:19]:

Yeah. And he said, I'm using AI for all of my internal emails, and I'm not using AI for my external emails. Now

Adam Gray [00:53:27]:

Interesting.

Tim Hughes [00:53:28]:

Interesting to interesting to hear what somebody that age was saying in terms of how they saw AI. Now I've I've presented this to, you know, to loads of people. What do you think? From people who said, well, what's the point of having a company if all of the internal emails are are AI generated? And then and what does he think about the people internally? But this is somebody at that age, and this is what they're thinking of doing with AI. And that's a different change in the way that we work and the different way that we have society.

Evangelos Simoudis [00:53:59]:

Actually, I'll tell you one of the things that I have found now as a venture investor go ahead, Bertrand.

Bertrand Godillot [00:54:04]:

No. No. No. That's fine.

Tim Hughes [00:54:05]:

No. He wants to we need to end the show, but you go ahead. Okay.

Evangelos Simoudis [00:54:09]:

One of the most interesting as a venture investor is that, we're seeing companies that are, startups that, promise to achieve a goal with fewer people than we would have thought in previous generations of startups. And all this is due to the use of, particularly programming, assistance, but also other type of AI tools. We'll see whether that succeeds, but but there is that promise which we find which we find interesting.

Bertrand Godillot [00:54:41]:

Evangelos, this has been great. Unfortunately, too short. But,

Evangelos Simoudis [00:54:45]:

Thank you.

Bertrand Godillot [00:54:45]:

Where can people learn more? Where can we get in touch with you?

Evangelos Simoudis [00:54:50]:

So they can I'm pretty active on LinkedIn. They can reach out to me through that, through our company's website, Synapse Partners, Google c o. And, yeah, I think LinkedIn is probably the the easiest way to, to connect with me. Yeah.

Bertrand Godillot [00:55:10]:

Okay. Great. We now have a newsletter where all of these details will be, will be, given, by the way. Don't miss an episode. Get the highlight get the show highlights, Beyond the Show insights and reminders of upcoming episodes. You can scan the QR code on screen or visit us at digitaldownload.live/newsletter. On behalf of the panelists and to our guests, Evangelos, and to our audience, thank you all and see you next time. Bye bye.

Bertrand Godillot [00:55:45]:

Thank you very much.

#AI #BigData #Innovation #SocialSelling #DigitalSelling #SocialEnablement #LinkedInLive #Podcast