Full text version of the conversation between Yevhen Pentsak, Mykhailo Patsan, and Yevhen Sarantsov about artificial intelligence in business, management decisions, critical thinking, and the human role in working with AI.

Yevhen Pentsak:

— Friends, greetings to all. Today we are at the Kyiv-Mohyla Business School. My name is Yevhen Pentsak, I am a lecturer in economic and financial disciplines. And today we will talk about artificial intelligence, where we are now, what are the prospects for using artificial intelligence. And we will talk with two Ukrainian experts in the field of artificial intelligence and lecturers at the Kyiv-Mohyla Business School – Mykhailo Patsan and Yevhen Sarantsov.

Mykhailo Patsan:

— Greetings to all. I am an investor, entrepreneur, founder of the WEB3 university Learn to Earn Global and the IT company AISync Global. We systematically implement artificial intelligence in business and, in principle, all innovative technologies that allow businesses to develop, be more effective, and compete in this world of uncertainty.

Yevhen Sarantsov:

— I am a guest lecturer at the Kyiv-Mohyla Business School. I oversee the program 'Artificial Intelligence for Managers'. We also have our own operational corporate accelerator at the business school, where we invite corporate teams to go from idea to a real prototype. In my professional life, I work as a business architect for 'UP Group'. This is a group of companies that includes 'SkyUp', 'Join UP!', and 'UP Universe'. And I am also an entrepreneur, experimenting with AI, creating the startup 'Collective Intelligence', in love with this idea. Glad to be on this podcast.

Yevhen Pentsak:

— It seems to me that there is no one in the world now who has not heard about artificial intelligence, perhaps about its threats, perhaps about its useful application for increasing personal effectiveness or the effectiveness of the business in which he or she works. Let's start a bit from afar.

And what metaphor for artificial intelligence would you formulate so that we are in one unified cloud environment? If you ask artificial intelligence itself now what it thinks of itself, it gives quite a lot of such metaphors. It's a kind of exoskeleton for our mind. It's a cognitive steam engine and so on, and so on. But I'm interested, you work with artificial intelligence every day at your fingertips. You grow it, you nurture it. You see what happens with it every day. What are your inner feelings, what does it remind you of?

Mykhailo Patsan:

— If we talk about it as a metaphor, for me, artificial intelligence is a catalyst for growth. It's an accelerator, you could even call it that with one word. That is, it accelerates all actions. And it doesn't matter if it's in a positive or negative direction. Depending on where you direct it, that's the direction it will accelerate. It is often compared to a system that can bring order to chaos, but that's not the case. And it's far from it. If in one word, it's acceleration. A catalyst for our growth, which allows me to be more effective personally, for my business to try ideas faster, iterate faster, and launch new products to market faster. And, accordingly, to be more, again, effective. And, accordingly, it allows for faster scaling.

Yevhen Pentsak:

— But faster doesn't mean more effective. You just said that artificial intelligence, chaos, can turn into even greater chaos. And it doesn't necessarily lead to growth. I think this will be an interesting topic that we will discuss today. We will talk about key business metrics, what we need to pay attention to, and what it influences. Thank you for your metaphorical presentation of artificial intelligence. Yevhen, what do you think?

Yevhen Sarantsov:

— On our program, there is a metaphor that we essentially use, which is augmented intelligence. Because we build our arguments that it is a synthesis of human and artificial intelligence, and we are actually looking for solutions around this. We see that it does not exist separately, although we all try to build autonomous systems, we believe in this, but without humans, they somehow all collapse. Therefore, I believe in augmented intelligence or in a supermind, where there is a synthesis of human and artificial intelligence. I will probably stop at that. Augmented intelligence is, one might say, an allegory that I personally believe in when talking about artificial intelligence.

Yevhen Pentsak:

— You touched upon such an interesting question: can artificial intelligence exist without humans? Then it turns out that this is a great threat to humanity, that it can replace the way new knowledge is generated, that artificial intelligence will do it instead of humans. You say that humans should be together with artificial intelligence. But does artificial intelligence pose a threat to humanity? Or is it still an accelerator, a catalyst, an exoskeleton, a cognitive accelerator? What is it?

Yevhen Sarantsov:

— I think it poses a threat. We cannot avoid talking about this either. To a certain extent, it paralyzes our society because it divides us into two cohorts. That is, those who will be subordinate to artificial intelligence and those who will create it. And in the last few years, working with corporate teams, we feel this very strongly. That is, first, a CEO, a stakeholder, or an owner comes to us, brings their team, and says: "teach them," so that they teach others. Then, a year later, we usually hold such final demo days with our teams, and we hear such stories. We are no longer waiting for people to learn how to use artificial intelligence; we are building it under the hood. And people simply receive tasks and perform them.

That is, we understand that in companies, no one will wait until a person wants to understand, activate this accelerator, enable augmented intelligence, and they always risk getting it under the hood. For example, it will check how you spend fuel, how you visit clients, how you communicate with them, and all this will happen under the hood, not with you. So, there is a certain threat, yes.

Yevhen Pentsak:

— If we continue your thought, it turns out that a self-selection mechanism will operate. Some people will use artificial intelligence and give it instructions, while others, a significant number of them, some say 95%, will follow instructions generated by artificial intelligence.

Yevhen Sarantsov:

— Exactly. Because experimenting both in our own company with artificial intelligence and in the Accelerator, we see a consistent scenario. When we look for how to make people work with AI, and no matter how many scenarios we try, everyone likes one: let it just set tasks for my team, and they just perform them. That is, this optimization, as the role of the executor and the role of the one who can set tasks. That is, the orchestrator, strategist, researcher, the one who seeks assumptions, makes judgments. A certain group of people accepts this as a solution, and people simply receive an orchestrated task of what they need to do. And we are actually afraid of this future. We say that this can be a real problem when 5% of people are architects and 95% are executors, for whom it is simply a complete black box.

Yevhen Pentsak:

— Thank you. Let's move on to Mykhailo. What do you think, is it a threat, or is artificial intelligence an assistant?

Mykhailo Patsan:

— You know, I want to state one thesis: we are already living in an era, and I think we will fully reach it in a year, because intelligence is a commodity. And, accordingly, it can be bought, and access to it can be obtained in some way. And there's no longer a need to educate a person for a long time for them to become a certain narrow specialist, because this experience can be bought. And accordingly, just as we buy raw materials now, we will buy intelligence to perform a certain action more effectively and faster. And, accordingly, based on this, since intelligence is like a commodity, it turns out we can go in two directions.

Side number one - we can buy it, use it, set tasks, let it set tasks for us, or let it analyze for us, and thus perform the work that we, as humans, can do best. I cannot be an expert in everything, but I can be a narrow expert and be the best. Similarly, we are now moving towards a situation where a person must be a narrow expert, but with a team of artificial intelligence agents, essentially with bought intelligence, one could call it that, it's not entirely the correct expression, but with the fact that, essentially, a person plus a team of artificial intelligence agents - this is, accordingly, a superpower, this is a superhuman who will essentially be the person of the future. Therefore, intelligence is a commodity, a person is, accordingly, a consumer of this commodity. And is this a threat? Well, a calculator appeared, everyone was afraid that people would stop calculating. Excel appeared, and then there was panic.

Therefore, it depends on the person, on how they want to develop, in what they want to develop, and what they want to do. If we, for example, look at the younger generation now, they often say, zoomers don't want to work. But why? Why should they work? First, they want to work, but they don't want to do some narrow task, that is, hypothetically, move a cup from point A to point B. Well, they don't want to do that, but they want to solve problems. And since they want to solve problems, but still don't have that much experience. But the motivation is there. And, it turns out, artificial intelligence allows the younger generation, with the right direction, to be more effective, faster, and, accordingly, to do what an experienced person does much faster, in fact. And in Ukraine, this is important now.

Yevhen Pentsak:

— But if intelligence is a commodity, then intelligence can be better, it can be average, it can be worse. This means that this intelligence will have different values. So, it turns out that it will lead to even further polarization of humanity in the sphere of wealth, in the sphere of income. Whoever can afford a superpower will also get the best economic result. And it will lead to such a great social threat. Young people or children who study from different families, with different access to educational services, may not have the opportunities that other people have. And, accordingly, this polarization will intensify already at the beginning of their life trajectory.

Mykhailo Patsan:

— I think so, but here the task of universities will change, schools will change, because access to knowledge is already global for us, access to teachers will already be more, well, it's decided that way, access to knowledge will be wider, and everything will depend on the teacher, the mentor, who will work with the person. And, accordingly, for now, I think that, yes, this polarization will exist. Currently, the discussion is generally about the fact that we effectively pay, let's say, 20 dollars for a million tokens generated by artificial intelligence, and we pay for everything, regardless of whether this token wrote 'hello' to me, or whether this token performed mathematical modeling of some large, complex business model. And, accordingly, we still pay the same cost. Most likely, this will happen, it's already being discussed.

I'm already communicating with developers, and, in fact, representatives of OpenAI and Anthropic are currently saying in interviews that we will be approaching a point where the cost of a token will change depending on the value it provides. And thus, there will be different gradations. It's the same as not everyone having access to a machine that can make a pipe. But everyone can learn how to use it and make one. And, accordingly, the same story will now happen with artificial intelligence. Just in a different sphere. It is now simply disrupting white-collar workers, disrupting the industry that required intelligence, that had to think. And, accordingly, it just accelerates it a little differently. So, whether we want to or not, we are heading into this era. We are hurtling in a race car that just needs to be steered quickly so as not to crash into a pole.

Yevhen Pentsak:

— It needs to be controlled by humans. And what type of human thinking is best suited for cooperation with artificial intelligence?

Yevhen Sarantsov:

— A small remark on the previous point. Just the day before yesterday, I had a meeting with students who asked me about artificial intelligence. Understanding the full seriousness of this technology, they said: 'What should we do? Because our lecturers forbid us to synthesize work as a supermind, as augmented intelligence.' What advice can I give to children? Most likely, in 3-4 years, I personally won't hire you if you don't show me your skills, your superpower, how you can work with this, this accelerator, and so on. But we, as humans, it seems to me, will resist for a long time, and then many dramatic events simply await us. Now regarding skills. Look, after several years of working with the accelerator, what have we seen? The teams that come to us, we call them business architects in the era of AI. That is, we involve our engineers in the accelerator who help create prototypes.

But we are super lacking precisely business architects in the era of AI, who can break down a business process, look at it holistically, understanding AI, reassemble it, and offer new value. Because we have a widespread idea of a human plus assistants. We see that no miracles happen, people are not rushing to get assistants, they will wait until they are fired, another person who will have assistants. We believe in agents as a new type of automation, automation plus AI. There, too, we are not moving quickly to a certain extent. Why? Because there are no business architects in the era of AI who can look at things holistically and reassemble them holistically. And it doesn't even matter with what specialization, because specialization becomes the second part of this role.

I can be a business architect in the AI era specializing in finance, I can be a business architect in the AI era specializing in analytics, for example, sales or marketing, but the first role becomes such, and the second role then becomes the specialization itself. And that's why we advise everyone to look at these things holistically, because there are simply no easy answers today.

Yevhen Pentsak:

— So, we come to the idea that we must be architects of our own thinking. That is, the new era, which we will call the post-artificial intelligence era, is not such a narrow specialization, but a breadth of thought at the intersection of different sciences, at the intersection of different competencies. Synthesis. And what about critical thinking? Mykhailo often says when teaching at our business school: 'Something hallucinated in artificial intelligence today.' I can confirm this, a few days ago, I used to praise it very much, we will now talk about what pivotal periods artificial intelligence has gone through, and where it is heading, but I saw that it is growing not by months, but by days, and what things it can do now.

But sometimes it has a bad day: you ask it a very simple question, and it confidently answers incorrectly. You tell it: 'Wrong,' you point: 'Look, there's this solution,' and it says: 'No, everything is correct. It means you formulated the task incorrectly for me' — and so on. This is a hallucination. And we will talk about critical thinking. If you trust it too much, it starts to lead you. Your thoughts stop working, your brain gradually atrophies, and you become completely dependent on this artificial intelligence. So, what are your thoughts on this? How to communicate with it, how to perceive it? And the role of critical thinking in working with artificial intelligence.

Mykhailo Patsan:

— During this time, over three years, we have worked with more than 80 companies, where we either conducted training or systematically implemented artificial intelligence. We have developed our own methodology, and we call it AI-thinking. Everyone thinks that AI stands for artificial intelligence, but we went the way of Apply, and called it AI Accelerated Improvement. That is, accelerated improvement.

Why do I constantly talk about speed? We need to think quickly, adapt quickly to changes, and not trust artificial intelligence. This is precisely about critical thinking, here lies the whole secret of why not to trust.

Look, artificial intelligence itself is an intellectual system for data analysis. You give it a certain set of data, it analyzes it for you and provides you with an answer. But what kind of data will it be? And what kind of information did you provide? Depending on this, it will give you a specific answer. And where do hallucinations arise? Where there is uncertainty. For example, you told it: 'Tell me what news came out today in Ukraine on the topic of artificial intelligence.' You gave it a request once, it will give you an answer. And it will be logical. If you give the same request a second time, in a different chat or another artificial intelligence model, it will also give you an answer, but it will be different. And why? Because we have uncertainty.

Since artificial intelligence, it's not even entirely correct to say artificial intelligence, it's rather a platform on the basis of which we use artificial intelligence, it will use an internet search tool to find some data that it deems necessary, and will provide us with an answer. And depending on what kind of data it collects for us, the answer we receive will depend on that. Moreover, it adapts to us. As they say: 'Every domestic pet resembles its owner,' so does every artificial intelligence resemble its owner. And if you ask a question in ChatGPT, and I ask a question, we will get two fundamentally different answers, because it adapts to that.

And, accordingly, depending on how many points of uncertainty there are, how much data we have on which we trained it, and what the quality of this data that we provided it for work is, the accuracy will depend on that. And most often, artificial intelligence is not used or people are afraid to trust it, why? Because this probability of a correct answer is significantly lower than what I expect. And it needs to be increased. And what increases this probability? The right task, the right tool, the right prompt, and the right data on which we receive the answer. And then, accordingly, we already use it. You see, it transitioned from being an instrument to just randomly consult, to being an instrument where we understand where we got this information, and we understand that we receive the expected answer.

Now I understand that it gives me information only on the data I wanted. And what is the phenomenon of, for example, NotebookLM? Whoever doesn't use it, I highly recommend it. Because we uploaded any information to it, images, videos, YouTube, and just wrote: 'Give me a short excerpt or extract key ideas and adapt them for our business' — and provided it with the information. And it will only take this data, and only based on it will it give you an answer. And you understand that this is great, it's accurate, it's what I expected. And therefore, in such a case, it really is a great tool. If it's a format of 'do an analysis for me on some random data,' in that case, we don't have critical thinking, you just always have to distrust it, it always tells lies and you always have to look at it through the prism of distrust.

Well, in principle, just as we live in the world.

Yevhen Pentsak:

— Let's move to a topic familiar to us, because there are also experts in this field, in the field of investing. They say now that 95% of all investment portfolios are built on artificial intelligence. But so far, the results are not impressive. You load them with absolutely precise data, quotation data with intervals of one day, one hour, one minute. But can this artificial intelligence, based on that data, build a highly profitable portfolio with a given risk level and so on? Are we expecting too much from it?

Yevhen Sarantsov:

— It seems to me that we started perceiving it incorrectly from the very beginning. My opinion is that when we decided there is me, there is he, she, it, whoever you like, it was a mistake. Because there is me and me, in fact. If we say that I am the key influencing factor on the model, then it is me through it. And we always explain this pattern in our programs. If I accept this as me and me, then I allow for the possibility that I can make mistakes. Because I simply might not distinguish one from the other. Or, for example, I didn't embed any thinking into this artificial intelligence at all. I didn't write prompts, I didn't write methodology. I didn't reduce uncertainty. Further in our programs, through this idea of Collective Intelligence, we show retrospective approaches, when the model, in essence, acts based on interaction with people. That is, from a group of people. And we begin to perceive artificial intelligence as statistical models.

Look, what are we used to? We are used to automation, where an algorithm is executed absolutely clearly, we are used to statistical models, where one equals one. They offered us a technology where nothing is one hundred percent. Never. Nothing is one hundred percent. And second, they offered us a technology that works based on context, and generally, it cannot calculate because it uses tools that calculate. For some, it's a revelation today that it cannot calculate; it calculates roughly the same way we do. The only difference between us and it is that it has a large contextual window that can look holistically at a large volume of data. We, humans, cannot always do that. Abstractly, we still can, but precisely, we cannot. Therefore, I believe that these old paradigms, if I may say so, either using it in a statistical format or automating it as automation, are simply a mistake.

I want to say that even if those same investors—and we know examples of investment banks that work based on news, the same cryptocurrencies—they look at news, at investor sentiment, at markets, at behavior, where it has an impact. Then why not use artificial intelligence that communicates with experts, looks at their forecasts, knows how to take context into account, understands some rating for itself. And from this context, it makes bets and performs better. No, instead, you see, we use it as a statistical model, although, in reality, I think the simplest model that simply launches a calculation tool in MATLAB and interprets the results for you would be enough. Then, initially, the model was just incorrectly built. That is, we are using a completely different approach, not the right approach, to talk about this.

A few words about hallucinations. We, as humans, said: 'This is a hallucination.' Although, in reality, it is not a hallucination; it's an incorrectly, probabilistically chosen interpretation of uncertainty. Because at this moment, when we maximize the field of uncertainty, as Mykhailo said, and I want to support him here, we can get an uncertain result. But why did we call it a hallucination? We even somewhat offended this artificial intelligence: 'You are bad, you are hallucinating.' But it's not hallucinating. When we start to deal with these hallucinations, we reach a point where the human has hallucinated. Because they are already used to setting tasks like this: 'Do something there, and generally, guess what I want.'

Yevhen Pentsak:

— Well, if we briefly summarize this part of our conversation, we are saying that the data must be verified and cleaned. You cannot throw anything at it. Because if we throw anything at it, the answers will also be anything with elements of hallucination. Therefore, there is no need to believe in miracles, but to treat artificial intelligence as a deeper level of ordinary business analytics.

Yevhen Sarantsov:

— We are not used to context. Let's imagine that it's the same data, only described as an analytical note. Like an analytical note. Then it can look at and read, understand, read a dozen such notes and find patterns there. That is, we often confuse certain things. And when we start using it correctly, then powerful magic happens.

Yevhen Pentsak:

— Mykhailo, a couple more words from you. Can artificial intelligence be used to build investment portfolios? And to what extent can those calculations be trusted and expected to yield a high Sharpe ratio, or Sortino, or some other, or do they generate alpha?

Mykhailo Patsan:

— It's actually very interesting here, because, for example, I've been in trading and investing in general since 2007. I've seen all these crises, how everything fell, soared, lost money myself, and so on. And here you just need to understand one thing. For some reason, we all idealize humans. But can a human do this? Can a human earn consistently, stably in the stock market? Can a trader trade stably? Because, for example, the average trader loses their deposit on a strategy where they trade themselves within 6-9 months. 9 months is considered professional. They are really good, they held out. And even some newcomers, who are new there, come. So, 6-9 months, no money. Accordingly, that's why they started building robots. I remember 2007. That's exactly when I started studying this. And I started working in IT in 2003. Then in 2005, I started studying robots, and in 2007, I entered the stock market.

And, accordingly, I started looking at the robots that American companies use there. That is, how they trade, how they analyze the order book, and so on. But this is all such a stochastic variable that you will never guess. And here everything depends on the strategy we want.

What is trading? Trading is actually finding inefficiencies in the market. That is, a person, they think ahead, they have certain emotions, they have certain visions of why they would buy or sell a particular company's stock. In 2021, everyone told me about Palantir: 'This is a loss-making company, it will never take off.' Those same people in 2025 said: 'This is an ideal company, good that you advised us to buy it.' This is not investment advice. And, accordingly, these are emotions. And these emotions will always be in the market. And how can artificial intelligence help trading, for example?

It can help read the context, understand it. Something that no artificial intelligence model, specifically machine learning, could do before. It read historical data, it predicted some prices, showed them, and tried to find some patterns. But in fact, all we need is one indicator to look at the people who are participants in this market and understand their decision-making logic. And in reality, in the investment market, the investment and trading market, the difference between trading and investments, as I, in fact, interpret it, is that trading is the search for short-term profit, but with high leverage. And investments are the creation of long-term capital. That is, it's more about timing, purely timing. And, accordingly, it is generative artificial intelligence, in this direction, that allows us to at least slightly understand the psychology, perhaps, the thinking of these market participants.

And, as I already said, they think very primitively, in fact. Because there is, roughly speaking, one indicator, now I've just started adding another one, which is return on investment. I invested X amount of money in this company today, how much free cash flow income will it bring me over the next 3, 5, 10 years. And depending on my timing, I now see that Micron, for example, seems like a cool company, manufactures products that are potentially needed in the era of artificial intelligence. But it cost, a year ago, even a year and a half ago, it was very inefficiently valued. Because storage, RAM, hard drives, it's all about data that needs to be stored somewhere. And without these devices, it won't work. Yes, everyone rushed to video cards, everyone forgot that storage was also needed. And, accordingly, we understand that efficiency exists. And the company is inefficiently valued.

Could we have calculated this? Yes, we could have. We could have calculated it even without artificial intelligence, actually. But AI allows us to accelerate.

Yevhen Pentsak:

— We talked a bit in general. Artificial intelligence, what it is, what it's used for, what its threats are, how to use it. And we talked about metaphors for understanding artificial intelligence. But the word 'to manage', 'manager', comes from the Old Italian 'to guide a horse with a bridle'. Can we, thus concluding this metaphorical understanding of what artificial intelligence is, consider it our horse? And we have saddled this horse and have certain levers of its control. I think that's a good metaphor.

Yevhen Sarantsov:

— Exactly, that's how you should look at it. Because our problems begin when we find problems precisely in this technology. For example, with hallucinations. Although we just gave a large zone of uncertainty and did not understand how everything actually works. Or, as with the investment case Mykhailo mentioned, that we directed it to the past, but why shouldn't we direct it to what it does best? It's a linguistic model. That is, it's a model that best understands how we think, reason, and predict. So why shouldn't it work with this data that you, Yevhen, mentioned. This is about two streams. It's what the market says and what experts predict. Essentially, why are we directing it to the past at all, if in the past, statistically, the same Python as a tool performs all this task, and artificial intelligence will simply set the task. Therefore, it should be seen as a horse. We need to mount this horse and ride. Because there is a great danger in sitting somewhere in the wagon of this horse.

Yevhen Pentsak:

— Or if you let go of the reins, it will snort, and you will fly out of that saddle. It can take you completely somewhere you don't want to go. Therefore, I also think that the mastery of managing this horse, metaphorically, reflects the mastery of managing artificial intelligence. Absolutely. If different riders with different experiences and qualifications mount the same horse, it will take them completely somewhere they don't need to go.

Yevhen Sarantsov:

— Moreover, it seems to me that this technology absolutely clearly understands who is communicating with it. We know that it accumulates memory, it synthesizes answers through this memory about the person. It acts based on who it communicates with. Therefore, if we cannot find anything ourselves, we cannot methodologically distinguish one from the other, then we will always be, as they say, slaves of ChatGPT, forgive me, as they say now, and we will always listen to stories where it guessed well, where it didn't guess so well. But it's like flipping a coin, will it work or not, will it guess or not. Reality will always be approximately like that.

Yevhen Pentsak:

— Friends, from you, as artificial intelligence experts, can you point out any pivotal moments over the last two years, how artificial intelligence has developed, and where it is going?

Yevhen Sarantsov:

— If a year ago we all bet on people needing to learn to work with artificial intelligence, Mykhailo spoke a lot about this, a human plus agents, this is another person, or a company needs to create more agents, we got Claude Code, as one of the latest phenomena, who generally came to the market and said, listen, let's put it this way, only chat is available to you, everything else, well, I'll do the rest myself, I don't need your prompts, your ideas, well, I just need to understand what context we are in, I will form memory for myself, I need API keys for certain things so that I can work, this was the first signal. But what is this signal about? It's a signal that we generally stop waiting for people to understand something for themselves and start moving in this direction. That is, stakeholders began to define more selfish strategies for themselves.

Then Hermes appeared, which is our direction, and how is it currently being received by leaders? They say, listen, if this is a model that can be next to my employee, write skills for them, create and so on, can I actually use it as a spy? Believe it or not, cynically, that's exactly the idea. He says, I want such a spy next to each of my employees, to make their digital copy, but in two months I want to ask it what impact this person actually creates in their workplace. The third signal, when I ask a manager, we had 500 top managers on the program, I say, look, you have large companies, how long do you think a person in your company can work at the level of an answering machine? A well-configured avatar, a well-configured model, it will even write like a human, answer emails, participate in meetings, Teams.

I think even Mykhailo knows how to do this. And managers say, well, probably a year. That is, I see such a shift that strategies regarding artificial intelligence will simply become more cynical. And when people, you know, who work in companies today, ask me, Yevhen, when should we start? Should we start? Should we, you know, let people learn? I ask them, tell me, friends, are you sure that anyone needs this at all right now, well, in this large industry? Will you learn this or not? Haven't you thought about it, look, the strategies of the Big Four, such as Google, Microsoft, Amazon, Grok, they always compete 3-5 years from now. Their main client is a person who is now doing homework with models. They, well, they've written us off. They say, why would we spend time on these people who need to be explained how Nokia works?

In 3-5 years, a generation will come that grew up with artificial intelligence; they will have far fewer questions about hallucinations, whether it works or not, how to work with it, what to do, how to calculate, what methodologies to use. Well, that is, we need to understand this moment, that we are truly alone here, our generation. So that's my answer. Everything is becoming much more cynical, and now the ideas that, forgive me, even Claude writes its own skills to solve your question. I was told that Mykhailo supports the idea that the topic of learning to write prompts will simply disappear over time. Or the topic, look, we made agents, you know, on these low-code platforms, n8n, and so on. The last accelerator, Yevhen Mykhailo, zero low-code, all agents wrote their own code themselves. Everything was done with vibe-coding, Claude wrote itself, wrote an agent for itself, opened an MD-file, corrected something there, did it, that's it.

We have a great prototype. We don't even need people who know how to work with n8n and those kinds of things, to connect, to assemble, and so on. Therefore, I believe that no one will wait, and this is the main change. And so, to summarize, I believe that stakeholders, people, managers who make decisions, they will make decisions not in favor of people. They will make decisions in favor of this one will spy, this one will make your digital copy, this one will set tasks for you, this one will calculate fuel for you, this one will do it like that. And believe me, even if it sometimes makes mistakes, I think they will quickly look at how many times their people make mistakes and say okay. We calculated, minus 30-40% OPEX, let's go play.

Yevhen Pentsak:

— I often ask you, what happened there, in which company? Which company laid off another batch of employees? They are all being laid off. Is this a threat that the unemployment rate will rise? Or will all those people still transform into some other role?

Mykhailo Patsan:

— Precisely answering this question and continuing what Yevhen said about spies and everything else. For example, when we enter a company, the first thing we do is sit down with a person and observe how they work. Or we interview them, like taking a photo of their workday, or in some cases, we even sit with them offline, and the person comes in and sits, watching, 'I press this button, this button, he does this.' And in some cases, we just record the screen, and then feed it to AI agents. In any case, when we gather everything, then we feed it all to AI agents. We have our internal ones built, which, in fact, act as an analyst-consultant, and as a result, we get in two weeks what others needed much more time for, even we needed much more time, back in the distant 2023.

That is, in fact, we are now moving to the point where artificial intelligence becomes a separate 'I' and an employee; we call it that, meaning it's your employee. Therefore, you need to build a separate department where these employees will work. You will need to spend money on API tokens, on subscriptions, and additional funds on computing power; no one will escape that.

Yevhen Pentsak:

— So, they need to be optimized, right? Because this resource can also be quite expensive.

Mykhailo Patsan:

— Exactly, and everyone thinks that now I will lay off 10,000 people, or, say, half the team, hire artificial intelligence, and it will work for me. But then it turns out that, for example, one of our developers spends from 2,000 dollars a month on tokens, and it almost... To make a prototype costs 300 dollars, but it's fast, and we can do many of these iterations, I mean, and thus product delivery happens not in a year, but in a couple of months. And, in general, our record was four products, fully functional, normal, full-fledged products, integrated, implemented in a large business, a large financial business, which is integrated directly by regions, by departments, and they started using it. That is, four months, four products. The point is not about laying off people, the point is about restructuring processes. And, accordingly, what do we see now? Mass layoffs are happening in the world now.

Mass layoffs, moreover, of the solvent audience, if you can call it that. That is, those who created the main economic effect for, well, a share of GDP. And what do we see now? This will not be long-term. This will be, well, a maximum of one year. Yes, we get this gap, meaning people will be laid off. And many will be laid off. But why will they be laid off? Restructuring of processes. That is, processes are gradually being rebuilt to integrate, automate, and embed AI agents, this AI department will be created, and new people who will be hired must understand the principles of AI, want to work with it, want to develop it, and, accordingly, work with new processes. Because now when we enter a company, at most, I see 15% of people who are interested in the development of artificial intelligence in the company. And even that, in fact, is a lot, usually 8-10%.

And, accordingly, this indicator is gradually growing; currently, according to statistics in Ukraine, an average of 25% use AI. Some companies more, some less, but they use it, and it's also a big question what they use it for. Therefore, there is great resistance, and there always will be. And it's not even resistance in the format of 'I don't want to use it,' but rather 'I don't have time to use it.' That is, make me a horse, let whoever wants to control it, ride it, and put me in the wagon, I'll ride there, and, accordingly, suggest what I need to do. So, most likely, we will transition to a situation where companies, whether they want to or not, must create three additional roles. The first is a strategist, an AI strategist. The second is a psychologist, not for AI, but for humans. And the third is a technical specialist who will help implement it.

Well, accordingly, it will then scale depending on the scale. Whether we want to or not, in all companies where such an infrastructure works, everything grows. Where it doesn't exist, it doesn't grow.

Yevhen Pentsak:

— Artificial intelligence is not a must. It's needed by some, while others can do without it. Because artificial intelligence doesn't come for free. Specialists, orchestrators, strategists, the cost of artificial intelligence are needed. And we measure how much it's needed, how much it accelerates business efficiency, how much it can influence capitalization.

Mykhailo Patsan:

— Exactly. Because we had a somewhat inverted business model. Because how was it before? There's a SaaS platform. That is, several smart people gathered, created a platform, let's say, CRM, which, like, helps businesses digitize their processes. They spent a lot of time and money on it, and then maintenance was almost nothing. 85% margin. That's great. That is, today we invest 100 thousand dollars, tomorrow we take a million. Brilliant. All SaaS platforms worked on this model until the advent of ChatGPT. Accordingly, after generative artificial intelligence appeared, it flipped for us. We can now create a ready product using AI tools like low-code and Codex. Just by speaking one prompt to artificial AI, it will make us a ready product - a product that we can already test, implement. Products aren't born ready-made, great.

They also go through prototyping, MVP, testing, and so on. That is, we can make it for three kopecks (a tiny amount of money). But to use it further, how many tokens it will consume, and what effect it will have, that's another matter. That needs to be calculated, because it could be that you make some product for 20 dollars, and then pay 2 million dollars for its use.

Yevhen Pentsak:

— So this is already an AI economy that needs to be calculated. Another metaphor comes to my mind: in business, they often say it's better to go badly on a good road than to go well on a bad road. So, which road to take, that's chosen by a human. And how to go, these are different tools, different accelerators, as we say, and one of them is artificial intelligence. Do you share this opinion?

Yevhen Sarantsov:

— I share the opinion that it's better to go, as you said initially, badly.

Yevhen Pentsak:

— But even better, Yevhen, is to go on a good road and go well. Right? And here, to go well, which road to take, our intelligence chooses, and with what speed to go, with what efficiency, that is already our ability to use artificial intelligence.

Yevhen Sarantsov:

— Look, if you choose to have this option in your hands, like AI-driven, AI-orchestrator, meaning we're setting a very high bar, we're not talking about people, with AI we're talking about a high bar. And not having this option in your hands, it's better to have it. And as you taught in your time, it's only a matter of the cost of expectation and the value of this option that you have in your hands. So there will come a time when models become even stronger, and those companies that had this opportunity in their hands, what does that mean? Even if they played with AI-first, even if it was super unserious, even if they were just interested in it, even if they gave up the idea of choosing correctly whether it's right or wrong, whether it works or doesn't work, whether it hallucinates or doesn't hallucinate. A moment simply comes when they have a 3-5 year gap.

I had such a situation just the day before yesterday; I was at a business club and talked to a person who is a graduate of our accelerator. After that, they invested over 100 thousand dollars in developing a platform for lawyers. Now they are redeveloping it. But what's interesting is that his path is already not easy. He still needs to rework it, and probably invest twice as much. But when I ask his colleagues in the business club, they say, we believe that he currently has a 3-5 year market gap. Ahead? Ahead. They say, there's your game. It seems not everything succeeded, it's very ambitious, it seems money has already been spent, not everything worked out, everything needs refinement. But when you look at it, they say, this person is all set, he is 3-5 years ahead of us, if it works out for him tomorrow, it's another 7-8 years, he just gets this gap, a monopoly, and will work with it.

And this difference between having an option in hand and not having one, I would advise not to ask yourself such difficult questions today. It's simply a matter of whether we are doing at least something today, whether someone is pressing a button, whether someone is creating a skill, playing with something, and that, as a rule, is always the key question.

Yevhen Pentsak:

— So, if we continue metaphorically again, we often refer to it this way: if there's a rider and a horse, you need to have a bridle. And then, whatever its qualification, it will come with time.

Yevhen Sarantsov:

— And some, you know, such a fenced-off area, where, in principle, we could play with it in the sandbox, feed it, groom it. And we liked it, let it be a hobby, but any company will encounter unexpected solutions. Because even when you come to these companies, you always find that, you know, the owners and CEOs of these companies are personally very surprised. They always say, 'We are testing the latest version of GPT. Look what Claude is doing now. And Gemini did something amazing for me.' You start looking at their company, and as they say, 'the horse hasn't even rolled over yet.' Not to mention the bridle, someone sat down, someone lost, and so on and so forth. And this, by the way, is very strange to me. If there's any technology worth playing with, it's definitely AI today. Because it's much more expensive now not to engage in what you should be engaging in. Even if it costs you 3-5 thousand dollars a month.

If you're small, let it cost you 200-300. I'm sure Mykhailo and I once calculated it together in class. I say, 'Can you make such a game for a thousand dollars?' He says, 'How?' I say, 'You'll spend 500 dollars, for example, on Claude with tokens, and 500 dollars on one and a half to two meetings with Mykhailo.' Mykhailo then said, 'one meeting, but I agree.' So here's a thousand-dollar AI game with a person who simply knows everything about AI, keeps up to date, and so on.

Yevhen Pentsak:

— Yevhen, you've already started giving such instructive advice to our listeners. And Mykhailo, what would you add here... We will conclude our meeting, starting from the abstraction of what artificial intelligence is, how to perceive it, to how to live with it, how to use it, how to turn it into an effective tool for business growth and its capitalization. What instructive words would you say today?

Mykhailo Patsan:

— We talked about changes there. I changed because of artificial intelligence, I changed the company and so on. But if at the end, a simple roadmap. How to check what level you and your company are at with the use of artificial intelligence? Just look at a very simple thing. First, how often do you use artificial intelligence and what do you use it for? Accordingly, you will see that you probably use it quite often if you are a business owner or manager. Accordingly, what role will it play for you? That is, the first level is, I would even call it the zero level, these are chats. That is, you just asked it a question, and it gave you an answer. Well, cool. Then over time, it should grow into such an executor for you. That is, not even an executor, but a more advanced assistant.

That is, you gave it a task, it read your mail, it read your calendar, it looked at who you are meeting with, it looked at some of your data, and gave you a specific answer so that you are ready. This is an assistant. Then it should be this, an executor. That is, you already tell it to move this meeting from today to tomorrow. And it already contacts, calls, writes, arranges, checks the time, shifts, and accordingly performs this work. And in this way, it can be called your personal operating system and AI system; I call it that, an AI system. First, you build your own AI system. After that, you will see that this whole system scales across your company, and in this way, you will gradually infect, as one owner said, infect the entire team with the idea of artificial intelligence. Create an AI community within your business.

Accordingly, gradually move it and build a system that will live independently, that will develop within you, and in this way, artificial intelligence will no longer be a question of whether it needs to be implemented or not. It will live there, you cannot do without it. It cannot do without you, and you will no longer be able to do without it, because it will gradually develop, and your team will develop with it, and, accordingly, it will grow with the team, because you will have a system. Therefore, this is the most important thing.

Yevhen Sarantsov:

— Alongside this evolutionary path, returning to our logic of real options and portfolio logic, you need to have at least two audacious ideas. So if there's time to try something super audacious, please try it in the next 2-3 years. This is a window of opportunity. I don't know what it will be. Look, you know, I had this experience: an association of banks, so many bankers were sitting there, and I asked them a question. We're talking about AI. I said, 'Do you agree that the next neobank is a chat?' They said, 'We agree.' I said, 'You know that Mastercard, Visa, are working on AgenticPay, right?' They said, 'We know.' Tell me, who has a backlog for all our applications for more than 2-3 years? Everyone. And the last question, who is experimenting with the next bank as a chat? Zero. So, I say, friends, can you at least do it for a laugh? You'll be whispering in a while, 'we want to.' So for 2-3 years, please try something audacious and focus on evolution. Mykhailo definitely supports this.

Yevhen Pentsak:

— So, it's time to realize such revolutionary ideas, landmark ideas. Absolutely, yes.

Yevhen Sarantsov:

— Today, you cannot not try this. Even if it doesn't work out. Alongside the system.

Yevhen Pentsak:

— Such a portfolio logic. Good. Friends, colleagues, thank you for such an interesting conversation. At least. I think it was interesting to our listeners; I heard a lot of new things for myself, I better understood where we stand with artificial intelligence and how to implement it for companies that have not yet embarked on this path.

Yevhen Sarantsov:

— Thank you very much, Yevhen. Thank you very much, Mykhailo. That was a great conversation.

Mykhailo Patsan:

— Thank you. Indeed, I caught a lot of insights, especially about these cases with strategies, with decisions, and so on. And in general, I think that in 2027-2028, we will have artificial intelligence that will warn us about risks and make decisions for us. It was a chat, an assistant, an agent, and an executor, and then it will make decisions for us.

Yevhen Pentsak:

— This is exactly what can take on the functions of a digital twin; a digital twin is such a recommendation.

Yevhen Sarantsov:

— But for those who played, for those who didn't play, GPT will hallucinate.

Mykhailo Patsan:

— Hallucination and the best human. Yes, thank you.

Yevhen Pentsak:

— That's all, thank you for the conversation.

AI in Business: Can Artificial Intelligence Exist Without Humans?

Yevhen Pentsak