Roman Stanek asked an American marketing adviser to invest in his company. The response came with a geography lesson. Regis McKenna, whose work included Apple, took him by the shoulders and explained the limits of his investment territory: “I would never invest in a company east of Palo Alto, let alone one in Europe!”
Stanek’s company was in Prague. By that reckoning, it had missed the acceptable neighbourhood by several thousand miles. The story would be less entertaining if the company had disappeared. It was NetBeans, and Sun Microsystems bought it in 1999. A border that sounded permanent turned out to have a reasonably short shelf life.
The introduction had come through Esther Dyson, an investor Stanek describes as a mentor and friend. He met her by chance at a European conference. The relationship opened doors into American technology circles and continued across his businesses. Before there was a succession of acquisitions and funding announcements, there was someone willing to make an introduction.
That distinction matters in Stanek’s life. He grew up in a small town in communist Czechoslovakia and studied computing theory before he had regular access to the machinery itself. He has said he did not use a computer until his early twenties. Entrepreneurship was hardly an obvious career plan in a system organised around state control. Curiosity arrived before the equipment, and considerably before the venture capital.
The students had the software
The founding story of NetBeans is particularly good at resisting the solitary genius treatment. In 1996, a group of Czech university students began a development environment called Xelfi. Their idea was to build a Java programming tool using Java itself. The students had code, ambition and a product they were trying to sell. Stanek was looking for talented programmers and a business to build.
He found the project online. In 1997, they formed a company around it. The architecture belonged to people including Jaroslav Tulach; the commercial opportunity was something Stanek could help pursue. The name NetBeans came from an early plan involving network-enabled JavaBeans components. The plan changed. The name, with the stubbornness names tend to acquire, stayed.
Tulach later recalled a useful moment at JavaOne in 1998. Stanek brought Jonathan Schwartz to the company’s stand, where Tulach demonstrated the software. Schwartz was involved in Sun’s Java-related acquisitions. The combination is revealing: someone needed to make the tool work, and someone needed to get the right visitor to stop walking.
Sun acquired NetBeans the following year. In 2000, it released the software as open source. The student project had become a commercial company and then part of a much wider developer community. Stanek’s contribution was consequential without requiring him to be cast as the author of every line. Finding people, finding a market and arranging the encounter were substantial work on their own.
- 1997NetBeansTools for Java developers
- 2000SystinetInfrastructure for web services
- 2007GoodDataAnalytics delivered in the cloud
A second company, a different kind of plumbing
Systinet followed around the turn of the millennium. Its subject was web services: helping separate software systems work together. The tools were less visible than a consumer application, but the problem was ordinary enough. Businesses had systems that needed to communicate, and those connections needed to be discoverable and manageable.
By 2006, Mercury Interactive had agreed to buy Systinet for $105 million in cash. Mercury’s announcement described software for governing service-oriented architecture, the industry’s term for organising applications as cooperating services. A company could have plenty of useful software and still need a way to keep track of how the pieces fitted together. Systinet occupied that space.
The buyer was Mercury. Hewlett-Packard acquired Mercury later in 2006, which explains why both names appear in accounts of the transaction. The $105 million was the company’s purchase price. It belongs in the business history rather than a calculation of Stanek’s personal fortune.
Looking across NetBeans and Systinet, a pattern becomes visible. Stanek built businesses around changes in how other people built software. Java needed tools. Connected applications needed coordination. These were markets whose usefulness depended partly on what other engineers would do next. Being early could create an opening; it could also mean waiting for the surrounding world to catch up.
Mercury Interactive’s announced cash purchase price.
Two years building, then the customers
Stanek founded GoodData in 2007. This time the opportunity was business intelligence delivered through the cloud. Companies were accumulating data, while access to useful analysis often remained concentrated in specialist departments. He wanted analytics to reach more people and to travel inside the products they already used.
The timing offered few comforts. The financial crisis tightened access to capital, and cloud software was still establishing itself. Stanek later described 2008 and 2009 as years spent building without customers. By 2010, there was a product ready to sell. The sequence sounds tidy in retrospect. Living through the gap between finished engineering and paying demand is a different matter. In October 2013, Business Intelligence Group named him Entrepreneur of the Year. Its announcement pointed to GoodData’s development since 2007, putting formal recognition several years after the period he remembered as building without customers.
One decision helped define the business: let other companies embed GoodData’s analytics in their own software. That made the customer’s product the place where the analysis appeared. A user could benefit from GoodData without ever making a deliberate visit to GoodData. For a founder, there is a particular kind of discipline in building something that may do its job without introducing itself.
The relationships grew beyond startup financing. In May 2020, Visa announced an investment and strategic partnership with GoodData to develop data products and customer insights. For Stanek, the ambition involved helping companies turn the information they already held into something their customers could use. A payment network offered a substantial setting in which to test that proposition.
The CEO brings a ceremonial pen
Stanek’s public persona has changed along with the businesses. In a 2022 conversation, he looked back at a video interview from roughly fifteen years earlier and found his younger self strikingly aggressive. He said he was trying to become calmer. His joke about his contemporary role involved being brought to a customer with a ceremonial pen ready to sign the contract.
There is affection and a little self-deflation in that image. It also suggests a founder learning to work through a larger organisation. In the same conversation, he emphasised thinking strategically, understanding where markets were going and developing partnerships. Those skills become more useful as a company’s future depends on a network of other companies.
His impatience with unreliable numbers remained. In 2022, he explained how repeated copies and transformations could leave business users unsure where a result had come from. GoodData’s move toward services accessed through APIs was meant to reduce that confusion. He connected the approach to the earlier ideas of service-oriented architecture. The Systinet chapter was still turning up in the GoodData conversation.
That makes his career more interesting than a list of exits. The subjects recur: how software parts cooperate, how definitions survive movement between systems, how a customer can trust the result. Technologies change their names rather often. The organisational difficulties have a talent for renewing their contracts.
“I actually think that headless is another manifestation of SOA.”Roman Stanek, 2022
An AI agent needs a definition of revenue
By January 2026, Stanek was pressing a precise question about enterprise AI: who owns the meaning of the business’s numbers? In banking, he argued, infrastructure, controls, financial definitions and customer views often sit with different teams. An AI system inherits that ambiguity unless people resolve it. Buying a fluent interface does not settle an argument about what counts as revenue.
Asked what good AI and data would look like inside a bank, he offered a pleasingly unglamorous answer: “'Good' is boring, in the best possible way.” The desired result was a meeting where people could use the numbers, an explanation that could withstand scrutiny, and intelligence available inside everyday workflows. Boring, in that setting, would be quite an accomplishment.
GoodData’s acquisition of Understand Labs in September 2025 added data storytelling and agentic analytics expertise. On April 30, 2026, the company adopted the GoodData.AI name. These moves put its longstanding analytics work inside a more explicit AI strategy. Stanek’s stated ambition was for AI to become a dependable part of ordinary business decisions.
September 2026 brought AI Observability, a product intended to connect measures of AI usage, quality and cost with the steps behind an individual response. On October 6, GoodData.AI announced its Agentic Serving Plane. The proposition was infrastructure between enterprise data and the applications or agents consuming it, with business definitions and access rules carried along. These were product announcements, with the practical test still residing in customers’ operations.
In an accompanying October essay, Stanek set out why agents change the workload. A person might inspect a dashboard. An agent can pursue a task through a chain of further questions. He argues that this makes consistent calculations, bounded permissions and predictable execution more valuable. The claim fits his commercial interest, and it also gives a clear account of the problem he has chosen to work on.
Nearly three decades after NetBeans, Stanek is again building for another kind of user. The next request may arrive from software acting on someone’s behalf. The company that answers it will need to know what the question means and what the requester is allowed to see. After all those years spent crossing borders, he is paying close attention to where the boundaries belong.