The Beginning of AI: From Unknown Study to Hidden Opportunity
Over the past three decades, Artificial Intelligence (AI) has developed at a rapid pace and is now more integrated into our daily lives than we could ever have imagined. Jan-Willem Gefken, director of the Palga Foundation and professional in IT management for data-driven organizations, already saw this coming in 1987 when he began an unknown new study in
Artificial Intelligence at the University of Groningen. “The professor, Bert Mulder, started a new multidisciplinary program on AI (…) if he hadn’t told me, I wouldn’t even have known it existed.” AI remained a mystery for a long time; “I simply told people I studied computer science. That was immediately understood.”.
The 1990s: Replicating vs. Supporting humans
In the 1990s, AI enthusiasts were divided. On the one hand, the ‘hard’ camp believed that computers could replicate humans, while the ‘soft’ camp believed that humans could not be replaced or replicated but could be used to create useful tools. This is comparable to today’s terms: generative and predictive AI.
This perspective fits a pattern Gefken mentioned during the interview: “Throughout history, people have described themselves through metaphors rooted in the technology of their time, such as steam engines (emotions as steam) or clocks (wound-up mechanisms). With the rise of computers, the metaphor reversed”: computers were suddenly described in human terms. This partly explains why early AI researchers focused so strongly on modelling human reasoning and behaviour in code.
“Throughout history, people have described themselves through metaphors rooted in the technology of their time, such as steam engines (emotions as steam) or clocks (wound-up mechanisms). With the rise of computers, the metaphor reversed”
Gefken also explains that language posed major challenges for AI, especially when sentences were not properly formulated. In such cases, AI failed to capture context and true meaning. “I take the train to the city, then the computer assumes there is a city inside the train.”
The 2000s and 2010s: Data as Superpower
After graduation, Gefken wanted to apply his new knowledge. He first worked at Shell on pattern recognition, then at the Dutch Ministry of Defence on human–machine interaction, and later joined Ordina, a consultancy and ICT services provider.
A true breakthrough came at the Dutch Railways (NS). He led the migration to an early version of online cloud technology, bringing data from multiple offline locations across the country together on one central platform. This ultimately became the foundation behind the “customer journey” philosophy, where customers could access all NS services via the OV-chipcard, a project that earned him the title of IT Manager of the Year.
Data also became a tool for security by using pattern recognition to analyse and predict behaviour. An example from his first years at NS: train toilets were frequently set on fire, a costly and dangerous problem. The algorithm developed by Gefken and his team predicted the next target based on timing and railway routes of previous incidents. “We set the algorithm up in six days (…) we caught the arsonists on the train, while they were in the process of starting a new fire.”
By the 2010s, AI had shifted from back-end infrastructure to consumer-oriented applications. For Gefken, this became visible with the NS Travel Planner app. “The best part is that you simply enter your journey (…) and if there is a disruption (…) you immediately get a notification with an alternative route. Then I think: ‘Yes, this is real AI.’ It thinks with me, it thinks things through.”
At the same time, workplace tools such as transcription software began replacing repetitive tasks. For consultants, this marked an early shift in client expectations. If passengers could expect real-time alternatives, companies soon expected the same intelligence in logistics, supply chains, and support. Still, AI was far from widely implemented and the potential risks were hardly recognised.
The 2020s: Generative AI, an oracle?
Today, as director of Palga, Gefken manages a national pathology database of 88 million records. AI supports both research and daily decision-making. Yet he emphasises that AI should not be treated as an oracle. “If you don’t know how to ask questions, or how to ask critical questions (…) then the algorithms will not come up with them themselves.” What matters most is not how advanced the tool is, but whether the human using it knows how to frame the right questions.
What matters most is not how advanced the tool is, but whether the human using it knows how to frame the right questions.
Reputation is Everything: The Risks of AI
Not all experiences with data applications are positive. At RDC, a provider of data services for the mobility sector, where Gefken served as managing director, sensitive information was mishandled within the data environment. Media attention threatened to escalate into a national scandal. “What you need to do then (…) is simply write off six months as a loss. It is very costly to communicate, to act, and the company will be at a standstill for half a year.”
“When you work with data, your reputation is your most valuable asset.”
His message was clear: “When you work with data, your reputation is your most valuable asset.” It serves as a reminder that the greatest risks of AI are not technical but reputational, because in working with data, client trust is everything. Therefore, transparency, governance, information security, and bias management are just as important as the algorithm itself.
Looking Ahead: What’s Next to Win with AI?
Despite his enthusiasm, Gefken doubts whether AI will ever replace humans. “Brains are too complex; they are a chemical ‘factory.’ We don’t even understand how they work. The idea of replicating them? I don’t believe in it.” Still, ignoring AI is no longer an option. So, what should consultants take away from three decades of AI hype and hope? Three lessons stand out:
- Ground AI in business problems. Don’t sell visions of humanoid robots when the real value lies in logistics, planning, and process efficiency.
- Protect your reputation above all. Data without trust is worthless.
- Keep judgment human. AI extends thinking, but consultants remain responsible for asking the right questions.
- Three decades of AI show that, it is essential, but not sufficient. The winning consultants will be those who effectively integrate AI into processes while maintaining trust and ensuring that judgment remains firmly human.
“Brains are too complex; they are a chemical ‘factory.’ We don’t even understand how they work. The idea of replicating them? I don’t believe in it.”
This blog is part of the student competition in Management Consulting Master Program at the School of Business and Economics.
