Part IThe Story
The computer belonged to a friend, and so did the hours. In the middle of the 1990s, in a Bucharest still working out what it was supposed to be after Ceaușescu, a mathematics student who had more or less stopped attending his mathematics classes waited for that friend to fall asleep. Then he sat down at the friend's machine. He had already spent weeks with a photocopy of Bjarne Stroustrup's book on C++, reading it without a computer at all, compiling programs in his head and debugging them on the inside of his skull, because there was nothing else to run them on. Now, at night, with the owner unconscious in the next room, he could finally see whether the code did what he had decided it would do. In the morning the machine went back to its owner. The friend slept. Daniel Dines worked.
The motive was arithmetic. Dines had heard that Romanian programmers doing outsourced work for American companies could earn as much as $300 a month. He was living on roughly $30 a month at the time, about a dollar a day, in an economy where most people were scraping together $30 to $50. One sum was ten times the other. A library book, a photocopier, and a borrowed keyboard in the dark separated them.
Three decades later the company he built does, in its most elemental form, something close to what he did at that desk. It sits down at someone else's computer. It looks at the screen. It operates the applications on that screen, as Dines once told a Harvard Business School podcast, "exactly like humans operate that," and it performs the work while the humans are busy with something else. UiPath raised $1.3 billion when it listed on the New York Stock Exchange on April 21, 2021, in one of the largest American software IPOs ever recorded. Bloomberg valued Dines's stake that morning at more than $6 billion. Forbes had already put him on its cover in September 2019 under the headline "Boss of the Bots" and called him the first bot billionaire, which was meant as a compliment and was received, by a man with a lifelong weakness for literature, as something more ambiguous.
He had wanted to write novels. That part of the story keeps coming back.
By the Numbers
The Long Night and the Fast Morning
~$30Dines's approximate monthly income in early-1990s Romania
9.5 yearsUiPath bootstrapped before its first outside round
$500KAnnual recurring revenue when the company finally raised
$1.6MSeed round led by Bek Ventures, August 2015
$35BPost-money valuation in the February 2021 Series F
$1.3BRaised in the NYSE IPO, April 21, 2021
$1.43BRevenue for the fiscal year ended January 31, 2025
A novelist without the talent
Onești is the kind of town that exists because somebody in a ministry decided it should. Dines's parents, a civil engineer and a teacher, met there after the regime relocated both of them to a new chemical factory town in the east of the country. That is the whole of the romance as the record preserves it: two professionals moved by the state like pieces on a board, a meeting, a son born in January 1972. Daniel Solomon Dines was raised largely by his grandparents. He remembers the period as the happiest of his life, which is not something every billionaire says about his childhood. Football. Kids in the street. A garden.
It was in that garden, at nine, that he decided he would be a writer. He later described the wider backdrop as feeling "stuck in time" — a communist country where progress had simply stopped, where the future was something announced on the radio rather than something you could walk toward. Under those conditions a novel is a reasonable ambition for a boy. It is one of the few machines you can build alone, with no capital, in a place where nothing moves.
In high school he read Jack London's Martin Eden, the story of a sailor from the Oakland waterfront who teaches himself grammar, poetry, and philosophy out of library books, writes his way into fame through sheer furious labor, and then finds the fame hollow. London's hero is a self-made man who discovers that the self he made was not the one he wanted. Dines finished the book and decided he would be a novelist. He actually wrote one. "Eventually I realized I didn't have the talent!" he told Sequoia in 2020, with the exclamation point preserved in the transcript, as if the discovery were still a little funny to him. But the love for reading stayed. So did London; decades later he would still drop him into conversations about entrepreneurship, alongside Nietzsche, to the occasional bafflement of venture capitalists.
There is a detail here worth holding onto. Communist Romania, as Dines puts it, did not treat writer as "a real profession." So the boy who wanted to make sentences went to the University of Bucharest in 1990 and studied mathematics and computer science, because he was very good at mathematics and indifferent to almost everything else in school. It was, he said much later, "a really good choice for me." It is hard to tell from the tone whether he meant it as gratitude or resignation. Possibly both.
Bridge, arbitrage, and the price of a programmer
He did not go to class. That is the first thing everyone who tells the story mentions, and Dines tells it himself without embarrassment: he showed up for exams and nothing else. The rigid, communist-era pedagogy that survived the revolution bored him. He found the bridge table instead.
Competitive bridge is a game of inference under incomplete information — you reconstruct four hidden hands from a sequence of bids and a few played cards — and Dines was good enough to earn money at tournaments. More consequentially, the best strategy literature was American. So at nineteen he taught himself English by reading an American bridge book with a dictionary open beside it, one word at a time, a slow mechanical translation of a foreign language into a usable one. Years later he would hire a dialect coach to sand down the pronunciation. The grammar came from the bridge book.
Nineteen was also the year his parents divorced and could no longer afford to support him. "I had to make my own living," he recalled to Bessemer's Byron Deeter. "It was a lot of hardship, but in a way, it was a colorful life." Hold onto that word, colorful. It returns.
What did a colorful life consist of in Romania in the early 1990s? Currency arbitrage, in his own phrase — the informal business of exploiting the gap between official and actual exchange rates in an economy hemorrhaging value to inflation. A recruitment company at twenty-one. A job-listing business. All of it improvised, all of it small, all of it survival. And it was through the job listings, apparently, that he learned the figure that rearranged his life: programmers in Bucharest doing outsourced coding for American firms were making up to $300 a month. A friend of his had just taken exactly such a job.
So: the library, the photocopy, Stroustrup, the friend's computer after midnight. He describes the ability he built in those months, reasoning about code before he could execute it, as the foundation of everything that came after. Not writing syntax. Thinking about it.
His first programming job paid $200 a month, a tenfold jump over anything he had earned before, and it came with a quirk that would turn out to be the happiest condition of his working life. The company did not have enough computers to go around during the day. So Dines started at 8 p.m. He worked through the night, in an empty office, alone with a machine that for those hours was entirely his. He has described that period to Harry Stebbings as the most formative of his early career and one of the most joyful — a state of uninterrupted flow, no meetings, no politics, just a problem and a screen.
He graduated in 1997. Half the students he had studied with eventually left the country.
— Daniel Dines, on 20VC with Harry StebbingsProgrammers are, in a way, lucky and happy creatures because they build something, they are creators. They are in a way artists; they are creating something and get to have the feeling of being in flow.
The cold room in Seattle
Microsoft came calling around the turn of the millennium. The precise year varies depending on whose account you believe — the company's biographies and the Wikipedia entry say 2000, Bloomberg says he left Romania in 2001, and one profile has him hired at twenty-nine — and Dines himself has used both framings. What is consistent is the destination, Redmond and Seattle, the division, databases, and the duration, roughly five years, ending in 2005. His official biographies credit him with designing and developing the SQL Server Agent, a component whose name now reads like an unusually patient piece of foreshadowing.
He could read English. He could not really speak it. He understood perhaps half of what was said in meetings and struggled to answer questions put to him directly. It took him several months, one profile reports, to work out that a "folder" was something more than an icon on a Windows desktop — that the word had a life outside the operating system, in filing cabinets and offices, in the analog world the software was modeled on. A man who had learned English from a bridge manual was now learning it from Microsoft, and the lexicon kept collapsing back into the product.
The engineering education was superb. He arrived at Microsoft during what he regards as its peak, surrounded by some of the most talented software engineers in the world, and he watched how a world-class company actually shipped enterprise software. He would use all of it. But the verdict he gives on the period, delivered at a SaaStr session years later, is blunt in a way his corporate biographies are not: Microsoft was "good for my engineering expertise but terrible for my life."
"I left a very colorful Eastern Europe Bucharest to go into the cold Seattle," he said.
There it is again. Colorful. The hardship of the 1990s had color; the comfort of the Pacific Northwest did not. He told Stebbings that he felt like a prisoner there, serving out a sentence while he waited for enough money to buy his freedom. Unlike most of the Eastern European engineers around him, he had no intention of staying in America forever. His life in Romania had been good. The fear that kept him in Seattle for an extra two years was practical rather than sentimental: that he would go home and find no interesting company to work for. Romania was not, as one account puts it, a place that attracted smart people. His own classmates were proof of that.
The solution he arrived at was to stop looking for an interesting Romanian company and to make one. He left Microsoft in 2005 with about $150,000 in savings. He calls it, with his customary lack of varnish, "fuck you money." His friends thought he was making a mistake. In retrospect he agrees with them, at least partly. "It was very foolish," he told Harvard's Bill Kerr, "not knowing anything about how to build a start-up."
Nine and a half years in the hole
The company was called DeskOver, and for most of its first decade it was a small group of engineers in a Bucharest apartment trying to find out what they were for. Dines co-founded it with Marius Tîrcă, a fellow Romanian engineer who appears in nearly every corporate history of UiPath and in very few of its anecdotes — the quiet co-founder, the one the cameras did not follow. Company records also list Adrian Dorache, Andra Ciorici, and Lavinia Cojocaru among the founders. What is known about the apartment is that it was small, that money was perpetually short, and that payroll was a monthly crisis.
The business began as outsourcing — Dines had lined up a few contracts before leaving Seattle — and then, because outsourcing was not what he wanted from life, it began to sprout products. The first real one was a consumer tool for Windows. You clicked on any word visible anywhere on your screen, in any application, and a small window popped up with a definition or a translation.
Consider that for a moment. A man who learned English with a dictionary beside a bridge book, who spent months discovering what "folder" meant, built as his first product a dictionary that floated over every screen. To make it work, his engineers had to solve a genuinely hard problem: extracting the text behind the image of a word in an application that had no intention of sharing it. They built, in effect, an engine that could look at a screen and understand what was on it.
"It was an engineering gig, very interesting stuff," Dines said later. "But obviously it was a flop from a marketing perspective."
Other consumer products followed. "They didn't work," he told Bessemer. What did work, modestly, was pulling the screen-reading engine out of the dictionary and selling it to other software companies as a component — an OEM business, a software development kit that helped engineers automate their own work. Running in parallel, the outsourcing kept the lights on. Dines had noticed, watching his own developers, how much of their day went to repetitive, boring tasks, and he had begun tinkering with software to automate them. It was not the priority. Survival was the priority.
By 2010 and 2011 the company was generating somewhere between $100,000 and $200,000 a year. It was, in his phrase, "a living." It was also the low point of his life.
— Daniel Dines, to Bessemer Venture PartnersIt made a living for us—the company was generating $100K-200K per year in 2010-2011. But that was really not what I wanted from life. I felt that I ruined my career.
He had stopped being a software engineer without becoming anything else. "My skills had become less relevant," he told strategy+business in 2019, "and I had learned a lot of things that nobody would have hired me for. A little bit of marketing, a bit of legal, a bit of finance, a bit of everything." He was, he said, "really pessimistic about the future." Then the plain sentence, the one that every bootstrapped founder will recognize in their body before their mind: "It is really difficult to run a 10-person company and not have any money to pay yourself at the end of the month, or not see anything in the future."
"I slowly went down into a big hole," he said. Nothing worked. "It's very difficult to sell from Romania."
In 2011 DeskOver lost its biggest outsourcing client. Exhausted, Dines did the thing that looks like courage in retrospect and probably felt like resignation at the time: he scrapped outsourcing entirely and committed the company to software. On top of the screen-reading engine they would build a business product, something a person with technical aptitude but no formal programming background could use to automate work on their own desktop. It would eventually be called UiPath Studio. The decision did not save the company. It merely made the company's survival depend on a single bet.
The numbers from this period are almost comic in their patience. Eight years, by Dines's own count, to go from zero to $500,000 in annual recurring revenue. Nine and a half years before the first outside money. A decade, he admitted to Stebbings, before he understood what specific problem the company was solving and for whom.
An astral moment
The answer arrived in his inbox.
In 2013, an employee of a large Indian business-process outsourcing company — one of the firms that employed thousands of people to keep the back offices of Western banks and insurers running — sent a cold email. He had found UiPath on his own, as an alternative to Blue Prism, the company some credit with coining the phrase "robotic process automation." His employer was drowning in manual, repetitive work across applications that offered no clean way in: no programming interfaces, no integrations, just screens and the people who clicked through them. UiPath's engine could read the screens.
"Look," Dines said of that moment, with the candor that runs through all his interviews, "in 2013, I didn't know what BPO was."
He found out. The single client generated $100,000 in revenue — more than half of what the whole company had been making in a year — but the money was not the revelation. The revelation was the size of the room it opened onto. Large enterprises everywhere were flooded with manual work that APIs could not reach. Dines recalls demonstrating his software to people steeped in the Blue Prism way of doing things and watching it automate the same processes faster and more reliably. Somewhere in this period, too, came what Jason Lemkin's SaaStr account calls "the $10,000 pivot": a small American software vendor building banking software came looking for one narrow capability, and the deal showed the team that the thing they had built for developers had customers they had never imagined.
Dines has a phrase for these episodes. He calls them "astral moments." "There are astral moments, and you need to seize them," he told Stebbings, speaking of those early clients in India. "We understood the environment and requirements. Most importantly: we understood this was a huge market; the companies were flooded with manual work and inefficiencies."
The word astral is doing interesting work. It is not the vocabulary of a growth hacker. It suggests alignment, conjunction, the stars briefly arranged in your favor — a thing that happens to you rather than a thing you engineer. Dines is consistent on this. "The desire to create a new category is arrogant and will cloud your judgment," he said at a portfolio event years later. Categories emerge when timing, technology, and market conditions align. You just need to be ready when they do. The decade in the hole was, in this telling, a decade of getting ready without knowing for what.
In October 2013, UiPath met its first true enterprise customer. That same year the company released the first UiPath Desktop Automation product line. In 2015 DeskOver formally became UiPath — the company renamed itself after the product, the clearest possible signal that everything else was finished. The consumer dictionary, the OEM kits, the outsourcing: gone. What remained was the engine that could look at a screen and understand it, now pointed at the most tedious work in the global economy.
Thank you for showing me the light
Something had to change in Dines himself before the company could grow, and he is unusually willing to say what it was.
"I was not born with humility at all," he said at UiPath's FORWARD VI conference in October 2023. "I was pretty arrogant and unhappy."
The arrogance was the engineer's arrogance — the belief that the people who build the tool understand it better than the people who use it. It did not survive contact with customers. When Dines started visiting the companies running UiPath's software, watching their teams and listening to their complaints, he found something that rearranged his sense of the hierarchy. "I actually realized they were smarter than us in using our own technology," he told theCUBE's Dave Vellante in 2019, "because they used it in the real world." He went back to Bucharest and gave his engineering team an instruction he still quotes.
— Daniel Dines, to theCUBE at UiPath Forward, October 2019From this day, I don't want to ever hear we don't fix bugs or features. When the customers say we need to do this, you say thank you for showing me the light.
"The defining moment," he told an audience at UiPath Live in February 2020, "was when we turned from a technology company into a kind of customer-first company." Humility, in his formulation, is not a personality trait to be admired but a mechanism with a measurable output: "Humility gives you the opportunity to course-correct fast. I think this is the most important business lesson that I ever had in my life." By 2023 he had refined it further — humility not as virtue but "more like a framework," a way of structuring decisions, approaching people, listening. He watched competitors lose deals because they were arrogant with customers, and he watched those customers turn around and say, in his paraphrase, let's work with the other guys; they're more flexible, open, and they listen to us.
The clearest demonstration of this happened in Tokyo.
Koichi Hasegawa was between jobs around 2016 and had become convinced that robotic process automation could raise productivity across the entire Japanese economy — a large thought for a man without an employer. He and Dines met over dinner and agreed that Hasegawa would work for UiPath part-time. About a year later Hasegawa reported that Sumitomo Mitsui Banking Corporation, one of Japan's megabanks, was rolling out a large automation program and was considering UiPath among others. UiPath was the long shot, a young company from Bucharest against bigger and more entrenched incumbents.
Dines got on the next flight to Tokyo. He promised SMBC a substantial investment in Japan and six full-time employees dedicated entirely to the bank's effort. SMBC chose UiPath — and, to Dines's surprise, replaced both of its existing automation providers with it. When Kei Yamamoto, SMBC's president of value creation, joined Dines on stage at FORWARD VI in 2023, he said the technology had been impressive, particularly the Orchestrator software that managed fleets of robots at scale. But that was not what sealed it. It was the intensity of the customer focus. By 2019 the bank was projecting $500 million in savings from its automation initiatives.
Six people. That was the promise. Not a feature roadmap, not a discount, but six human beings assigned to one customer's success by a company that at the time could barely afford them. It is a strange thing for a robot company to have won on. Or perhaps not strange at all.
There is a coda that Dines tells against himself. Even at $175 million in annual recurring revenue, deep in the hypergrowth years, a trusted customer called him to say the company had lost touch, that it was only chasing revenue now. He listened. He tells the story as a warning about how fast the arrogance comes back.
Genghis Khan in midtown
The first outside money came in August 2015: $1.6 million, led by Bek Ventures, with Credo Ventures and Seedcamp alongside. Bek's founding partner, Dan Lupu, a Romanian who had backed the company when there was very little to back, invested in common stock with a provision that kept him pari passu with future rounds — an unglamorous structural choice that would generate billions in returns. Bloomberg, writing about Bek's next fund in November 2024, summarized the outcome in its headline: a twenty-fold return. Dines still gives Lupu a shout-out by name in interviews, the way some people mention the teacher who first told them they could write.
What the seed round did to Dines psychologically is, by his account, at least as important as what it did to the bank balance. He calls it a switch. For a decade he had operated from a posture of scarcity — the posture of a man who had lived on a dollar a day and understood in his body what the worst case looked like. The seed round flipped it. Someone outside the apartment trusted them. The fear of running out evaporated, and the company could, in his phrase, go big: spend aggressively, hire aggressively, expand without waiting for permission.
He had been told, repeatedly, that the way to build a great software company was to perfect the American go-to-market first and then replicate it abroad. He ignored the advice. There had never been a home market to perfect anything in. "Starting a company from a small place with no market has a hidden advantage," he said. "It forces you to think globally from day one." In 2015, after the seed, UiPath opened offices in London, New York, Bangalore, Paris, Singapore, Washington, and Tokyo. Its revenue ended up spread, as Dines likes to point out, roughly across three continents — Europe, North America, and Asia — a distribution almost no enterprise software company achieves. "We were forced to think global," he told Harvard's Kerr.
The internal name for what came next, as Sequoia later recounted it, was the "Genghis Khan strategy": aggressive expansion into multiple markets at once, scaling operations to consolidate share in a category Dines sensed would be winner-take-most. The numbers that followed are some of the most remarkable in the history of business software.
Annual recurring revenue grew more than tenfold between 2016 and 2017, ending the year at $43.5 million. By July 2018 it had crossed $100 million. By mid-November 2018, $150 million. Headcount went from 590 in 2017 to more than 1,200 across 22 offices by the end of 2018, with 1,500 enterprise and government customers and a developer community of 200,000. In 2017 the headquarters moved to New York. Accel led a $30 million Series A in April 2017, one of the largest in Europe that year. On March 6, 2018, UiPath raised $153 million at a $1.1 billion valuation from Accel, Alphabet's CapitalG, and Kleiner Perkins, becoming the first unicorn founded in Romania. Six months later, on September 18, 2018, CapitalG and Sequoia led another $225 million at $3 billion.
The Sequoia relationship began with a rejection. The firm had initially passed, citing a lack of chemistry. Dines was honest with them about it — and the honesty, he says, provoked something like a competitive spirit on the other side of the table. Sequoia came in for the next round. Carl Eschenbach became the partner on the account in 2018. On Sequoia's founder page, beside his name, the firm placed a quotation from Dines that sounds less like a growth story than a confession: "When I started this company, my main goal was to create a place where I and the people that I work with will be happy. Changing the world came a bit later."
On April 30, 2019, Coatue led a $568 million Series D at a $7 billion valuation, with CapitalG, Sequoia, Accel, and a crowd of crossover funds alongside. Gartner announced that UiPath had moved from fifth to first in RPA market share between 2017 and 2018. In September 2019 the Forbes cover arrived, and with it the title of richest Romanian, displacing the former tennis player Ion Țiriac, who had held the position for years with a fortune of about $2 billion. That spring, after the fire at Notre-Dame, Dines gave €1 million toward the cathedral's reconstruction. In November 2019 Deloitte ranked UiPath first on its Technology Fast 500. The company now counted 3,000 employees, more than 5,000 customers, half a million developers, and more than $300 million in annual revenue.
And then the Genghis Khan strategy ran into arithmetic.
UiPath had gone from roughly $45 million to $175 million in a year — almost four times — and Dines has said that year won the market. The plan for the following year was to more than triple again, to $600 million. The company spent accordingly. Dines recalls the moment he saw the projected burn: around $300 million, against a planned $150 million. He describes the anxiety it produced as severe. The full attempt, as he summarized it at a SaaStr event, burned roughly $400 million and required layoffs. The 2019 reduction in force was, he says now, a necessary correction — one that left a leaner organization that happened to be ready when COVID arrived a few months later.
"The fine line between bold and crazy," he calls it. Four times in a year was bold. More than three times again was crazy. He does not claim to have known in advance where the line was.
When the pandemic came, in March 2020, the forecasts around him were apocalyptic; people were saying a 50 percent miss would be the new normal. UiPath modeled what survival looked like at a 10, 50, and 60 percent miss — and postponed every decision that involved letting people go. "It's one thing to have to right-size a company when the economy is booming and you know people can find other jobs," he told Sequoia. "This was different." Within a month it was clear the business was holding up. Many customers were in health care. In one case, UiPath robots saved nurses about two hours a day; by the time Sequoia profiled him again in December 2021 the figure had become three hours, which may reflect a different hospital or a story improving with each telling. Either way, he said, that mattered more to him than anything the company had accomplished financially.
In July 2020 the company raised $225 million at $10.2 billion. In December it filed confidentially for its IPO. On February 1, 2021, it raised $750 million at a post-money valuation of $35 billion. On Wednesday, April 21, 2021, it listed on the New York Stock Exchange under the ticker PATH, raised $1.3 billion, and closed its first day up 23 percent.
He was asked, around then, what the money meant. "I was lucky at some point," he told Sequoia's Paul Boutin, "to realize the true value of money—what money can buy, what money cannot buy." He also admits the frugal mindset never really left. The target for "fuck you money" kept moving — from a million to five million and onward — and he describes the habit as a kind of prison of its own. The first prison was Seattle. The second was the one he carried out.
The founder who left and came back
Public companies demand a different animal, or so the boards and bankers believe, and Dines half believed it too. He has spoken openly about a long struggle with imposter syndrome — a feeling that led him, for years, to hire for credentials over chemistry, to defer to résumés, to trust the experienced outsider over his own read of a person. He also struggles, by his own admission, with negative feedback and with firing people; his empathy runs high and he tends to delay. His thoughts move faster than his speech, he says, so he had to learn to communicate in frameworks and principles and fill in the blanks afterward. There is a "demon" in him, a kind of stage fright, that he has spent years trying to convert from ego into energy.
In 2022, UiPath brought in Rob Enslin, a veteran enterprise-sales executive who had come from running Google Cloud's go-to-market organization, as co-CEO. The logic was legible: Dines the product founder, Enslin the operator who knew how to sell to the largest companies on earth. In January 2024 the arrangement tilted further. Dines stepped down to become chief innovation officer, focused on the problem that generative AI had suddenly made existential — how a company built on deterministic software robots stays relevant when large language models can, in principle, read a screen and decide what to click. Enslin became sole CEO.
Less than six months later, Enslin resigned. Dines took the job back.
The middle of this story is the part that usually gets glossed over, as Nilay Patel observed when he had Dines on Decoder in April 2025, and Dines did not gloss it. What he tells, in various interviews, is a story about direction. During tectonic shifts, the Sequoia retelling goes, founders have a unique ability to align a company for the future; the company needed one cohesive direction, not two. He came back as what he called, in conversation with Eric Ries, a second stint — with plans for it.
The plans were drastic. UiPath tore out its older workflow engine and replaced it with one designed from the ground up for what Dines calls agentic orchestration — the coordination of AI agents, software robots, and human beings in a single process. There was internal resistance; modernization won. The company began calling the strategy its Act II. In July 2024 it launched Autopilot, integrating generative AI into its platform, and announced a collaboration with Microsoft — his old employer — to bring agentic AI into Teams. In October 2024 came a preview of Agent Builder. In March 2025 it acquired Peak, a Manchester company building agents for inventory and pricing. In April 2025 it launched the UiPath Platform for Agentic Automation, with an orchestration layer called Maestro. Later it bought WorkFusion, which automated compliance work in banking. In October 2025 TIME named the platform one of the best inventions of the year.
The skeptics' question, which Harry Stebbings put to him in December 2024 in a title that could not be printed in full, was whether agents meant RPA was finished. Dines's answer was a distinction he has repeated ever since. Rule-based work across multiple business systems, with structured inputs and stable conditions, is what robots do well and what language models do badly; models are not built to follow strict algorithms. Unstructured work — the parts of a process where enterprise knowledge resists being written down as rules — is where models shine. Enterprises need both, and they need something neutral above both. He started calling UiPath "the Switzerland of agentic automation," a vendor-neutral layer that orchestrates agents from SAP or Salesforce or open-source frameworks like LangChain alongside UiPath's own robots. Most business processes cut across many systems, he argues. An agent that only works inside one vendor's walls will not be enough. "Purity kills products," he said elsewhere, remembering the long, pointless early debate over whether screen-based automation was "real" automation compared to APIs. Customers do not care about technical purity.
He also ran the experiment on himself. UiPath's "Customer Zero" program re-implemented the company's own SAP S/4HANA system with automation at the center, working with SAP and Deloitte, and achieved what it called a 93 percent "clean core" against an industry standard of about 80 percent. Dines wants 95 or 96. The more striking figure is the finance department. As UiPath's revenue grew more than sevenfold, the size of the finance team stayed the same. More than 85 percent of critical finance workflows, from billing to reconciliation, now run through autonomous automation, while the people handle exceptions and complex cases. The work, Dines said, had become "more challenging, more fun, more interesting."
For the fiscal year ended January 31, 2025, UiPath reported revenue of $1.43 billion and a net loss of $74 million, with 3,868 employees. The stock had long since fallen from its debut heights; Eric Ries, introducing him in March 2025, described a $6.65 billion company. Dines's own net worth, by Forbes's count in mid-2025, was about $1.5 billion. He keeps planning horizons short now — six months to a year — and says forecasts beyond two or three years are not worth much. He has also said, flatly, that he will not found another company.
The work that remains
He starts most days with a book. Sometimes, Diane Brady of Fortune reported, the book is followed by a nap, and only then does he turn on the computer. Before the nap, or after it, there is Hacker News, which he reads every morning, and then the reading on marketing, legal, and sales that he does because a founder should be, in his word, versatile. "A CEO is someone who knows a lot of things about a lot of things," he told Bessemer, "but doesn't know anything really well." He is, he has said, a lazy person — "I work really hard at the company, but only at things I like. Being a lazy person, I had to build better pattern recognition in life." It is the kind of sentence that only someone who has worked through the night by choice can say without irony.
Around 2023 he began writing a book. Not alone.
He worked with two models, Claude and ChatGPT, and he did not use them as ghostwriters. He used them as opponents. "It was really like a Socratic dialogue," he told Brady. Each session was a series of hypotheses, challenges, instructions: here is the theory, read this article, now construct the opposing view. He set the two systems against each other in what he called adversarial prompts. The manuscript swelled.
— Daniel Dines, to Fortune's Diane Brady, September 2026I used two models and did adversarial prompts: This is the theory. Read this article. Create opposing views … The book grew at one point to be almost 100,000 words. AI is extremely verbose, so the real work was cutting it down.
The real work was cutting it down. The machines produced; the human decided what stayed. Three years of mornings and evenings, while he was also running a public company through the largest platform shift of its existence. "When I started, I didn't have such a clear idea of where I wanted to land," he said. "Our company strategy, the book, they evolved together." And then the sentence that the boy in the garden in Onești would have recognized: "I had many evenings and mornings where I couldn't wait to wake up and write. I was in a complete state of flow."
Flow. The word he had used for the night shifts in the empty office in the 1990s.
The Work That Remains was published in late August 2026 — 168 pages, around 40,000 words, its subtitle Human Judgment, AI, and the Architecture of the Next Enterprise. Its thesis fits on a line: AI proposes, humans decide, automation executes. Agents plan and prepare. People own anything that carries consequence. Deterministic systems carry out exactly what was approved. Dines's vocabulary for the missing layer is the map and the rails — the map a governed description of how a business actually works, the rails the machinery that executes its stable parts without deviation. He predicts that by the end of 2028, any large regulated enterprise running a high-consequence workflow autonomously will be doing it inside an engineered environment, not with a general agent dropped into legacy systems. Then he does something vendor books almost never do. "If broad enterprise deployment proves otherwise," he writes, "this framework weakens."
"I run an automation company," he writes. "That is my declared bias." Don't accept the argument on his authority, he says. Don't dismiss it for the same reason.
The argument he most wants readers to leave with is about learning. AI reasons, plans, drafts, explains; it genuinely embodies characteristics of human intelligence. But it does not learn on the job. A new hire arrives knowing nothing and becomes valuable by absorbing everything nobody ever wrote down. "The model arrives brilliant," he wrote on LinkedIn the day the book came out, "— and shows up exactly as new every morning."
He tested this in the most personal way available to him. He asked models to write fiction in a given style. The results came back bland. "An average by definition doesn't have a taste," he said on UiPath's podcast, The Path Forward. A model is an average of everything it has read; taste comes from lived experience, not memory. You can memorize every book ever written about skiing, he said, and it will not make you a skier. You have to fall on the slope. "Our memory is not our identity."
He traces his interest in the subject to a lawyer friend, who told him that her fear was not that her job would vanish but that her identity would become irrelevant. Dines wants executives to keep two ledgers — one for the visible outcomes of a job, the signed contract, and one for the hidden ones: the juniors mentored, the client relationship held together, the years of unwritten knowledge. Cut blindly against the first ledger, he warns, and you destroy value recorded only in the second. He has told executives that eliminating junior hiring starves the enterprise of its future leaders. Fewer passengers and more drivers, he writes, in a workforce where talent density stops being a philosophy and becomes arithmetic.
"Everybody feels some sort of anxiety, me included," he said, speaking with a colleague, Andrada Morar. "We don't know how our kids' career is gonna look like." He has one child. His answer to the anxiety is a line he repeats often: in times of anxiety, action is the answer. It is a line from a man who has also said that he spent much of his late twenties, thirties, and forties in introspection rather than action, and who admits to loneliness.
"I have ideas," he told Brady, "but I don't have the time and the talent to put them on paper in a structure, in a manner that can be consumed by the public at large."
The talent. Forty years after Martin Eden, the same verdict, delivered in almost the same words. Except that this time the book exists. It is downloadable for free from the company's website. On the cover, beneath the title, under his own name, in smaller type, it reads: with Claude and ChatGPT.
Part IIThe Playbook
What follows is an attempt to extract the operating logic from Dines's story — not the inspirational residue, which is abundant, but the decisions and habits that actually moved the company. Many of them cut against standard Silicon Valley advice. Several of them are lessons Dines learned by doing the opposite first. Each is grounded in specific evidence from the narrative above.
The principles
- 1.Let a small market force a global company.
- 2.Treat the long failure as formation, not delay.
- 3.Wait for the astral moment, then burn the boats.
- 4.Run humility as an operating system.
- 5.Win the anchor customer with people, not features.
- 6.Hire for chemistry; experience is rentable.
- 7.Find the line between bold and crazy — then respect it.
- 8.Defer the irreversible decisions when the fog is thickest.
- 9.Refuse technical purity.
- 10.Return when the company needs a single direction.
- 11.Be your own first customer.
- 12.Declare your bias and date your prediction.
Principle 1
Let a small market force a global company
The standard advice Dines received was sequential: dominate America, then export. He had no America to dominate. Romania offered no meaningful enterprise software market, and the absence turned out to be a gift. Within months of the 2015 seed round, UiPath had offices on three continents — London, New York, Bangalore, Paris, Singapore, Washington, Tokyo — and its revenue ended up distributed roughly across Europe, North America, and Asia, something Dines believes almost no other enterprise software company has achieved.
The structural benefit was not just diversification. A company born without a home market never develops a home-market accent. It cannot mistake one country's buying habits for universal truths, and it has to build a culture that works for engineers in Bucharest, sales teams in Tokyo, and customers in Chennai at the same time. Dines credits the global footprint with building a diverse culture and recruiting the best people available anywhere, not just in one metro area.
G
Sequential vs. Simultaneous Expansion
The conventional go-to-market path versus the one UiPath was forced into.
| Conventional wisdom | What Dines did |
|---|---|
| Perfect the U.S. playbook first. | Opened seven international offices in the year of the seed round. |
| Treat a small home market as a handicap. | Treated it as a forcing function: "It forces you to think globally from day one." |
| Win the biggest market, then expand. | Flew to Tokyo to win SMBC while still a long shot at home and abroad. |
| Concentrate revenue where the buyers are. | Built revenue spread across three continents. |
Tactic: If your home market is small, assume from the first hire that your first major customer will be on another continent, and design pricing, hiring, and support for that.
Principle 2
Treat the long failure as formation, not delay
Eight years to $500,000 in ARR. Nine and a half years bootstrapped. A consumer dictionary that flopped, consumer products that "didn't work," an outsourcing business that lost its biggest client, a founder who believed he had "ruined" his career. By any conventional metric, DeskOver was a failure for most of its existence.
Dines reframes that decade, with hindsight, as the education a CEO needs. He had picked up "a little bit of marketing, a bit of legal, a bit of finance, a bit of everything" — the skills nobody would have hired him for as an engineer, and precisely the skills a CEO requires. "My winding path was essential for my formation as CEO," he said. The technology mattered too: the screen-reading engine built for a failed dictionary became the foundation of the company. He goes further, warning that abundant capital now produces "unidimensional" founders who skip these formation years and lack the negotiation skills and resilience that only scarcity teaches.
The point is not that failure is noble. It is that the assets accumulated during a long failure — a hard technical capability, a generalist's skills, a high tolerance for pain — are often the exact assets the eventual success requires, and they cannot be bought on the way up.
Tactic: During a slow stretch, inventory what you are building that would be hard to acquire later — a core technology, a cross-functional skill set, a cost discipline — and protect it rather than discarding it with the failed product.
Principle 3
Wait for the astral moment, then burn the boats
Dines does not believe you can build a category on purpose. "The desire to create a new category is arrogant and will cloud your judgment," he said. What you can do is be ready. When a cold email arrived in 2013 from an employee of an Indian BPO who had found UiPath as an alternative to Blue Prism, Dines — who admits he didn't know what BPO meant — recognized that the moment had arrived. "There are astral moments, and you need to seize them."
Seizing it meant total commitment. The company renamed itself after the product. The outsourcing, already cut in 2011, stayed dead. The consumer ambitions went. The Sequoia retelling calls it a hard pivot, and the hardness is the lesson: half-pivots, where a company keeps the old business alive as insurance, rarely capture a new category before someone else does.
The two halves of this principle are in tension, and the tension is the skill. Patience during the years when no signal exists; ruthlessness the instant a real one appears.
Tactic: When a single customer reveals a market far larger than your current business, rename, restructure, or kill the old line within months — not years — so the organization cannot hedge.
Principle 4
Run humility as an operating system
"I was not born with humility at all. I was pretty arrogant and unhappy." Dines learned humility because he watched it win deals. Customers using UiPath in the real world, he realized, understood the technology better than the engineers who built it. His instruction — "When the customers say we need to do this, you say thank you for showing me the light" — turned a personal epiphany into an engineering policy.
The interesting move is that Dines treats humility not as a virtue but as an instrument. "Humility gives you the opportunity to course-correct fast," he said, calling it the most important business lesson of his life. He later described it as "more like a framework" — a direction for decisions, for approaching people, for listening. Its competitive function is speed: an organization that assumes it might be wrong discovers its errors sooner. He also watched arrogance destroy competitors, who lost customers that went looking for the vendor who would listen.
Humility decays with success, which is why the anecdote about the trusted customer at $175 million ARR — "you've lost touch with us, you're only chasing revenue" — matters. It requires maintenance.
Tactic: Write a literal rule for how your engineering team responds to customer feature requests — default to "yes, show us" — and audit compliance quarterly as revenue grows.
Principle 5
Win the anchor customer with people, not features
SMBC was considering UiPath against bigger, more established vendors. Dines flew to Tokyo and offered a substantial investment in Japan plus six full-time employees dedicated solely to the bank. SMBC replaced both of its incumbent providers. Its executive later said the technology was impressive but that the customer focus is what sealed the decision. SMBC went on to project $500 million in savings.
The asymmetry is the lesson. To a startup, six full-time employees is an enormous commitment; to a megabank, it is a signal of how much the startup will bend to its needs. Incumbent vendors cannot easily make that signal, because their attention is already spread across hundreds of large accounts. Koichi Hasegawa's role matters too: a part-time believer hired over dinner opened the door. Early international expansion often runs through a single person whose conviction exceeds their title.
Tactic: For the one customer that would validate your company in a new region, commit named, dedicated people in writing — headcount is the one thing incumbents cannot match.
Principle 6
Hire for chemistry; experience is rentable
Dines's most repeated hiring lesson runs against his own instincts. Imposter syndrome, he says, led him for years to prioritize experience — to defer to impressive résumés because he doubted his own judgment. It did not work. "You need chemistry," he told Stebbings. "Don't trade chemistry for experience." A team of domain experts who don't gel, he said at a SaaStr event, is "a recipe for disaster," and personalities do not change.
His method evolved accordingly. He used to ask candidates specific questions designed to measure intelligence. Now he has no script; he watches what kind of conversation happens naturally. For functions he doesn't understand, he asks board members to introduce him to great people in those areas — even people he can't hire — so he can calibrate what good looks like. The same logic ran in reverse with Sequoia: the firm first passed for lack of chemistry, and Dines's honesty about the rejection is what changed the relationship.
He also notes a trap: senior hires need a "honeymoon" of freedom, but once they integrate deeply, removing them becomes much harder. Chemistry assessed late costs more than chemistry assessed early.
Tactic: Before hiring for an unfamiliar function, hold three calibration conversations with excellent practitioners you cannot hire, then interview candidates without a script and weight the conversation's ease as heavily as the credentials.
Principle 7
Find the line between bold and crazy — then respect it
UiPath's growth from roughly $45 million to $175 million in a year won the market. The attempt to more than triple again, to $600 million, burned roughly $400 million and forced layoffs; at one point Dines saw projected burn of about $300 million against a planned $150 million, which he describes as producing severe anxiety. He calls the distinction between those two years "the black magic of hypergrowth."
The deeper lesson is psychological. The 2015 seed round flipped a switch in Dines from scarcity to abundance — a necessary switch, since without it the company would never have expanded fast enough. But the same switch, left unchecked, produced the overreach of 2019. A founder's risk posture is a dial, not a personality, and it has to be turned back as deliberately as it was turned up.
T
The Acceleration and the Correction
UiPath's growth arc, from the apartment to the trading floor.
2005
DeskOver founded in a Bucharest apartment after Dines leaves Microsoft with about $150,000.
2011
Biggest outsourcing client lost; outsourcing scrapped to focus on software.
2013
Cold email from an Indian BPO employee; first enterprise customer met in October.
2015
Renamed UiPath; $1.6M seed led by Bek Ventures; seven international offices open.
2017
ARR reaches $43.5M; $30M Series A led by Accel; HQ moves to New York.
2018
$153M at $1.1B in March, $225M at $3B in September; ARR passes $150M by November.
2019
$568M at $7B; aggressive plan burns heavily; reduction in force.
2021
$750M at $35B in February; IPO raises $1.3B on April 21.
Tactic: Set an explicit burn ceiling tied to a growth multiple you have actually achieved before; when the plan requires a multiple you have never hit, cut the plan, not the ceiling.
Principle 8
Defer the irreversible decisions when the fog is thickest
In March 2020, forecasters around Dines were predicting that a 50 percent miss on targets would be normal. UiPath immediately modeled survival at 10, 50, and 60 percent misses — and then postponed every decision that involved letting people go. "It's one thing to have to right-size a company when the economy is booming and you know people can find other jobs," Dines said. "This was different." Within about a month, the data showed the business was holding up. The eventual cuts landed on in-person marketing events, not people.
The principle is not "never cut." Dines had cut in 2019. It is that planning and acting are separate decisions, and that under maximum uncertainty the cost of waiting a few weeks for real data is usually lower than the cost of an irreversible error made on bad forecasts. The planning gave the company readiness; the delay gave it information. "Don't panic," was his summary.
Tactic: In a sudden downturn, build the full cut plan within a week, then set a fixed review date four to six weeks out and refuse to execute any layoff before that date unless cash requires it.
Principle 9
Refuse technical purity
For years the automation world argued over whether UiPath's screen-based approach was "real" automation compared to API integrations. Dines now calls the debate a distraction: "Customers don't care about technical purity." The tool should work for the job at hand.
He applied the same pragmatism when large language models arrived. Instead of declaring RPA obsolete or dismissing agents as hype, he drew a functional boundary: robots for rule-based, structured, multi-system work; agents for the unstructured parts where knowledge resists codification; humans for consequential decisions. He then positioned UiPath as "the Switzerland of agentic automation," orchestrating agents built by anyone — including competitors' embedded agents and open-source frameworks like LangChain. He even chose a model from Alibaba for one task because it was best at reading semi-structured documents. Neutrality, in this telling, is not a marketing pose but a refusal to force customers into one ideology.
Tactic: When a new technology threatens your core product, map exactly which tasks each approach does best and build the layer that coordinates them, rather than defending the old approach or abandoning it.
Principle 10
Return when the company needs a single direction
Dines handed over power in stages: a co-CEO structure with Rob Enslin in 2022, then a move to chief innovation officer in January 2024 so Enslin could run the company alone. Less than six months later Enslin resigned and Dines returned. The Sequoia retelling frames the return around cohesion — during a tectonic shift, a company needs one direction, and founders are unusually able to align it.
What Dines did on returning is the substance: he replaced the company's older workflow engine with one built for agentic orchestration despite internal resistance, and relaunched the strategy as Act II. A returning founder's advantage is not nostalgia; it is the authority to tear out foundations that a hired executive might hesitate to touch.
L
Four Leadership Configurations
How Dines's role shifted as the company's needs changed.
| Period | Structure | Phase |
|---|---|---|
| 2005–2022 | Founder-CEO through bootstrapping, hypergrowth, and IPO | Build |
| 2022–Jan 2024 | Co-CEO with Rob Enslin | Scale |
| Jan 2024–mid 2024 | Chief innovation officer; Enslin sole CEO | Risk |
| 2024–present | Sole CEO again; agentic re-architecture | Reinvent |
Tactic: If you step back from the CEO role during a platform shift, define in advance the specific conditions — such as a split in strategic direction — under which you will step back in.
Principle 11
Be your own first customer
UiPath's Customer Zero program re-implemented its own SAP S/4HANA system with automation at the center, achieving a 93 percent clean core against an industry norm of about 80 percent. The finance team stayed the same size while revenue grew more than sevenfold, with more than 85 percent of critical finance workflows running autonomously and staff focused on exceptions.
This does three things at once. It produces the most credible possible case study. It forces the product team to feel the product's failures firsthand. And it lets Dines speak with specificity about what automation does to human work — "more challenging, more fun, more interesting" — rather than in abstractions. A vendor that has not automated itself is selling a theory.
Tactic: Pick one internal function, run it entirely on your own product for a full year, and publish the headcount and workflow numbers whether or not they flatter you.
Principle 12
Declare your bias and date your prediction
In The Work That Remains, Dines predicts that by the end of 2028, large regulated enterprises running high-consequence workflows autonomously will do so inside engineered environments rather than with general agents dropped into legacy systems. Then he names the evidence that would weaken his own case: "If broad enterprise deployment proves otherwise, this framework weakens." And he states his conflict plainly: "I run an automation company. That is my declared bias."
This is rare in vendor thought leadership, and it is strategically shrewd. A dated, falsifiable prediction is more memorable than a vague vision, and a declared bias disarms the reader's skepticism faster than any claim to objectivity. It also disciplines the author. Dines wrote the book through adversarial prompting — instructing two models to construct opposing views of his own thesis — which is the same practice applied to a manuscript. Ideas that survive structured opposition are worth publishing; the rest get cut.
Tactic: When you publish a strategic view, attach a specific date and a specific observable outcome that would prove you wrong, and state any commercial interest you have in being right.
Free playbook
Get The Business Model Playbook
58 business models, one visual page each: how the money flows, the metrics that matter, and who runs it. Free when you join the Faster Than Normal email.
Free. No spam. Unsubscribe anytime.
Part IIIQuotes and Maxims
In their words
— Daniel Dines, at UiPath FORWARD VI, October 2023I was not born with humility at all. I was pretty arrogant and unhappy.
— Daniel Dines, quoted by 3 Seas EuropeBill Gates used to talk at Microsoft about a computer in every home, but I want a robot for every person.
— Daniel Dines, Sequoia Capital founder pageWhen I started this company, my main goal was to create a place where I and the people that I work with will be happy. Changing the world came a bit later.
— Daniel Dines, on 20VC with Harry StebbingsYou're dead when you stop learning. If you didn't change considerably in the last 6 months, it means you did something wrong. You have to push yourself.
— Daniel Dines, on UiPath's podcast The Path ForwardAn average by definition doesn't have a taste.
Maxims
- Small markets make big companies. A home market too small to sustain you is a reason to go global on day one, not a reason to wait.
- The hole is a classroom. The skills a failing founder accumulates — a little marketing, a little legal, a little finance — are the exact curriculum of a CEO.
- Seize the astral moment completely. Categories cannot be built on purpose, but when one appears, rename the company and burn the old business.
- Thank the customer for the light. Customers using your product in the real world understand it better than you do; build your engineering policy around that fact.
- Commit people, not promises. Six dedicated employees beat a feature roadmap when you are the long shot against incumbents.
- Don't trade chemistry for experience. Credentials can be rented; the way a team works together cannot be fixed later.
- Bold is four times; crazy is betting you can do it twice. Know the growth multiple you have actually achieved, and cap your burn to it.
- Plan fast, cut slow. Under extreme uncertainty, model every scenario immediately but wait for real data before making irreversible decisions about people.
- Purity kills products. Customers want the job done, not the architecture you admire.
- Keep two ledgers. Every job has visible outcomes and hidden ones; automate against the first without measuring the second, and you destroy value you never knew you had.
