Daniel Dines, named The Boss of the Bots by Forbes, founded UiPath in 2005, after returning from the United States to his native Romania. His goal, as he emphasized, was to build a company that would help save people time and stress related to boring, repetitive and time-consuming work. Today, UiPath is headquartered in New York and is one of the world's leading automation companies, with Dines' net worth exceeding $1.5 billion.
In an interview with RMF FM, Daniel Dines talks about his book "The Work That Remains", written with the help of AI, in which he advises how to overcome problems in implementing automation. He argues that artificial intelligence is only a tool that will not replace people. In his opinion, it will gradually become more and more difficult to imagine areas in which AI will not be able to increase our capabilities, but he is convinced that it will never turn against us.
Grzegorz Jasiński: Our guest today is the founder and CEO of UiPath, Daniel Dines. Good morning. You have just announced the release of the book "The Work That Remains". What is it about?
Daniel Dines, CEO of UIPath: Well, I'm the CEO of an automation company. Over the last thirty years, I have worked with many clients to help them optimize their business processes. I see many people having difficulty implementing these solutions, especially in the context of corporate processes. And I ask myself – what is happening? Because on the other hand, we see extraordinary expectations related to AI. I think we heard at the beginning of the year that we would soon have millions of Einsteins in one data center. If we took literally what it means to have Einstein in a data center, we would all be redundant. However, if we understand Einstein as a person, with a body, experience, intuition, will, cultural background, and stakes that he risks in his career, then this is how I perceive this formulation. And I think that when we talk about "millions of Einsteins", many of us think the same. So I started digging deeper into what the strengths and limitations of AI were, and how to create a method to work around those limitations when implementing it in the enterprise. This is what my book is about.
This is a book for business leaders, but I would like to ask about a topic that is more important not only for employers, but also for employees. What to do to maintain employment in the AI era?
First of all, you need to be very proficient in AI because there is no doubt that it enhances our human skills. Secondly, we also need to develop those features where AI has nothing to offer. So what are these features? AI's biggest limitation is its inability to learn on the job like humans do. When I ask you a question, you will remember it, absorb it and update your knowledge the next time. AI in an enterprise always starts "clean", each time it has to download all the company's knowledge, and such a resource simply does not exist. When entering a company, no one receives a manual saying: "This is how you should learn." The best way to keep your job is to become a critical part of the enterprise, a "memory part" of the company. This cannot be fully translated into a document or notebook that could be used by AI. Artificial intelligence cannot initiate actions. Although all companies have goals and results to achieve, their implementation requires people's initiative, even microscopic decisions. For example, I have a feeling that the customer is dissatisfied and has called, so I will offer him a discount... AI cannot support this type of initiatives. And what is very important, you have to be part of a job where you have responsibility and something to lose. AI cannot provide this. You can't make technology legally responsible for something that goes wrong in a company. It is always up to a human to ultimately approve every AI action. If the AI makes an incorrect payment, someone has to pay for it. If AI makes a decision that is inconsistent with regulations, someone has to explain itself. This responsibility is a key feature of human work.
Do you think there can be processes or incentives for companies to retain employees not only for profit, but also - let's call it - for the greater good?
I don't think companies should be charities that employ people "for the greater good." But they will do it in their own interest. In my book I write about hiring juniors. Today it may seem that many junior positions are no longer needed because AI could replace them. Even if we take a junior programmer - AI can generate code that a beginner would write. But the essence is that juniors in the company will turn into seniors in the future. Without them, in a few years the company will become empty inside. AI, as I said earlier, does not maintain the active memory of the company and does not carry its culture. So if we stop hiring people, training them and promoting them in the organization, the company will become completely empty in a few years.
Business is mainly based on selling things to people, so that they can buy them, they need money, and therefore work. So if someone says that AI will take half of the jobs, it may ultimately harm these same companies, there will be no one to buy their products.
Yes. This would be a self-fulfilling prophecy leading to collapse. If we leave people unemployed, that's obviously the biggest, most crucial problem. My book explains why this won't happen. AI has the ability to replace selected, clearly defined outcomes in the system. Of course, AI can replace visible effects, such as performing a specific task, e.g. a lawyer issuing an opinion within clear regulations. But it cannot replace everything that is invisible: personal commitment, customer relations, influence on others, the way you train others, the feeling of being part of the company. It's a lot - my memory, my existence in the company. This is an important thing to understand: our work is not just an activity, we are part of a whole called an enterprise that is changing. AI is not transformed by working in a company, there is no appropriate technology for this. Let me give you a metaphor I use in the book: AI is like someone who has read a book about a company but hasn't lived a day in it. AI does not know our mistakes, does not understand the realities, and has not been "imbued" with the culture and way of operating. This is not enough to actively co-create a company. It's like watching skiing videos on the internet, then getting on the slope and seeing that you have no idea how to go down. This is what AI is like today...
What do you think about the possibility that other companies will force prices of services down thanks to AI? There has been information that banks expect law firms to reduce prices due to the use of AI. Is this the beginning of a larger process?
But this always happens in the economy. This is the case with the commercialization of most goods. This is the natural order of things. Looking from the perspective of industrial development, this pattern repeats itself all the time. Of course, some banking costs will come down thanks to AI, allowing you to offer better service to your customers. This is ultimately good for consumers. For me, AI is practically like the Internet, which has brought more prosperity to the world. Telephones and computers are better tools that gave us prosperity. Of course, because of computers and phones, jobs have changed and the same is happening now with AI. But we shouldn't get hysterical that it will suddenly become self-aware and destroy humanity and all jobs. Ultimately, it is we, as humanity, who decide which jobs can and should disappear. We are not doing this solely for a very limited "one percent" elite or some metaphysical goal of creating a "new species" that is AI. This is total nonsense. If we think of AI as a tool, a lot becomes clear right away.
I will ask about liability and cybersecurity of the systems built by your company. How to create systems that work properly and do not make mistakes that are harmful to people?
This is exactly my point of view. AI is a tool. Someone may use them in the system, or set an opposite goal: "try to break this system." Someone may enter a bank with the intention of stealing, but this is a human intention. It is crucial to understand this difference. When it comes to cyber attacks, AI can test various scenarios and it practically costs the attacker little, and the entire responsibility falls on the company that has to protect itself against it. This is the reality. AI that breaks into banks should face the same legal consequences as a thief. Of course there are state issues here, it's more complicated. But the development of technology is part of the story...
Currently, we are dependent on electricity, the Internet and now also AI. How to avoid excessive dependence? How to build a system in which a company cannot simply be "off the grid" and paralyzed by lack of access to AI models?
I think the answer is multi-dimensional. In my book, I described a kind of "constitution of a new enterprise" using AI. One of its principles is that whenever you can define something in the form of rules, it is better to use software to do it rather than artificial intelligence. Anything that can be done with rules, it would be foolish to hand over to AI. The software can be controlled, turned on and off. AI should work in a "harness", i.e. surrounded by very deterministic software. This is one of the tools that allows you to stay in control. Secondly, I also believe in the optionality and sovereignty of AI, more and more companies will try to minimize the costs of used tokens, e.g. by redirecting tasks between different models to reduce the cost of AI consumption for each activity. I also think that many countries or institutions will host their own models on their own servers. If you're applying AI to well-defined, fully understandable tasks that aren't particularly novel, you don't always need the latest, most advanced models, sometimes simpler ones will suffice. They can be run locally in the company, and only up to, say, 10 percent. For the most innovative tasks, you need to use the "top of the line" model. To summarize: you need to surround AI models with software, run them locally, route tasks to different models for freedom. This is the direction that companies will take to maintain control over their fate.
Some say that AI is currently too expensive, sometimes even more expensive than human labor. What is your opinion? And more broadly, will high business expectations for AI automation be met?
I assume that the cost of tokens will decrease. For clearly defined domains and tasks, AI will be cheaper than human work. The current situation is, in my opinion, temporary. Enterprises will benefit from a combination of human talent, human intelligence, human characteristics and AI. I don't think human vs. AI competition involves exactly the same activities. If AI can perform a task, over time its cost will certainly be lower than that of human labor, regardless of quality or other factors. Maybe not today, but definitely in the future.
We fear that AI will take away lower- and mid-level jobs. Do you think you can also touch the highest ones? Is it possible or advisable to create agent AI for your position, for example CEO?
Today, AI helps me a lot and I spend hours a day using it. He supports me in creating strategies, checking proposals requiring my assessment, and writing documents. I even wrote a book using Claude and ChatGPT models. This helps me a lot. But replacing me as the final person responsible for the company is impossible today. Let me be clear, AI does not replace jobs, only the tasks within those jobs. It may reduce the number of people working in a given role, but it does not completely replace the field. I don't see this pattern emerging, looking at both the current state and expected development of the technology. As long as AI does not learn on the job, acquire individuality, intuition, and personality traits, there will be no mass replacement of employees by AI.
We started the conversation with the topic of your book. You wrote it with the help of AI, specifically Claude and ChatGPT. You mention it directly. What do you think about writing books this way?
I learned a lot about working with AI. This is not an easy task. It's not like I asked "write me a book about AI" and got a ready-made text. The number of my AI conversations and prompts was longer than the book itself! I used AI to tell a narrative, organize a structure, and test ideas. There were times when the AI suggested something interesting that I included in the book, but I was the director of everything. For many hours I browsed, thought, introduced new ideas, changed the order of chapters, and analyzed the coherence of the book. All in all, AI is a great tool. If I wrote the book myself, I probably wouldn't be able to do it, I don't have that much literary talent. Besides, it would probably take two years, but we finished it in four, maybe five months. So it's a great accelerator, accelerator, but it doesn't replace me. Not at work, not in my hobbies, not as a book author.
At the end of the conversation, I would like to ask for something positive, or in your opinion, more positive about AI. In what other areas of life can AI really help? Maybe where we don't think about it yet?
It's hard to imagine a place where AI won't be able to help. In my private life, AI supports me a lot in taking care of my health - it monitors the results of blood tests, suggests supplements, diet, I analyze everything with AI. This works very well. This will reduce the overall cost of health care in the future and make people healthier. I see how AI accelerates research into new drugs and therapies. This is a tool, a powerful accelerator. If AI can reduce the time it takes to discover a new drug from 5 or 10 years to 2 years, that's a huge change. Instead of developing one or two drugs in 10 years, the company will develop five of them, this adds up. In my case, similarly, before AI it would probably have been difficult for me to write one book in my entire life, now I can think about one or two books a year, also about broader topics, for a larger audience, not only business ones. It's an amazing tool. I am not a critic of AI, on the contrary, I am a fanatic of its use. In my company, I told everyone: we need to become symbiotic with AI. This is our best development, it can support initiative, will, individuality, personality, intuition, as well as the ability to reason. AI puts humanity's vast database at our fingertips and the extraordinary ability to synthesize and work with that data. Even these AI "hallucinations" are something other than a logical fallacy, humans make mistakes too, and AI is quite good at reasoning within strict rules. If we put all this together, we have already become "superhumans" compared to ourselves five years ago.
