AI in oncology: saving lives with data
KWF funds hundreds of research projects every year, and the number of projects in which AI is used is growing rapidly. Some of the studies include:
"One in two people will get cancer at some point. Every year, 130,000 Dutch people hear that diagnosis. Left or right, everyone comes into contact with it." The urgency is great, Ourahou emphasizes. "That's why any innovation that can contribute is groundbreaking."
The biggest revolution of our time
Two years ago, with the emergence of generative AI, Miloud Ourahou presented this topic to the management team: "We don't know exactly what yet, but this is technology we need to test. From there it spread like an oil slick."
Ourahou calls AI the greatest technological revolution of our time. Yet, according to an MIT study, as many as 95% of AI pilots fail. "Not because the technology doesn't work, but because organizations don't yet know HOW to apply AI properly."
That insight led to an important conclusion: AI should never be an end in itself. "If it's just a stand-alone tool, it's going to fail. It has to be embedded in existing work processes."
Lessons from an experiment with a chatbot
KWF also experienced this itself. One of the first projects was a chatbot to provide campaigners with answers faster and to relieve the service team. "The result: no chatbot on the website," laughs Ourahou.
The chatbot did not immediately have the desired results. First, because the basic data were insufficient. "A chatbot based on website texts doesn't work. Those texts are made for quick reading, not for full mapping of internal processes. AI is only as good in that as the data you put into it." In addition, dynamic information proved too complicated. Information about collections, for example, varies by zip code and changes frequently, and that is difficult for a model to keep track of.
The lesson was clear: AI rarely fails on technology, but often on context and knowledge. "You can fly in an outside expert to create something, but ultimately your own people need to understand how AI works. Training and education are therefore crucial."
From machine learning to organization learning
The failed chatbot became a turning point. No longer testing individual tools, but learning organization-wide how to deploy AI.
KWF therefore developed tooling that allowed employees to think for themselves: where can AI play a role in my work, and where not? It yielded more than 200 use cases, some of which could be implemented immediately. "People in the organization often know very well what they need. As an AI expert, this gives you a much more realistic picture and better use cases."
From idea to impact: concrete successes
The new approach is paying off. Some successful implementations include:
Learning from how AI learns
AI systems learn from human data. But according to Ourahou, it can also be the other way around: "What can we learn from how machines learn?"
The essence of AI is endless experimentation: testing, failing and trying again. "Not everything works out, and that's not a bad thing," he says. In a demo, for example, an AI solution often works perfectly, but as soon as it has to take other processes into account in practice, it turns out how much extra knowledge is needed.
Therefore, don't just choose the sexiest or futuristic applications, but look at what is close to people. What is of real use to them and what they can work with immediately. Precisely the solid, practical ideas ultimately make the biggest difference.
The future: AI agents and a second chance for the chatbot
What Ourahou believes is going to have a real impact in the near future are AI agents. "Within a year, I see huge gains there. But the real power is not in the technology itself, but in the fact that the organization learns how to use AI."
And the chatbot? With the lessons learned in mind, it gets a second chance. "No medical questions of course, but for practical and supportive issues it can certainly be valuable."
From experimentation to impact
Ourahou succinctly summarizes his vision: "AI is all about experimentation, trial and error. That's not failure, it's learning. And make sure your organization knows how to deal with the technology."
At KWF, that leads not only to more efficient processes, but ultimately to the most important goal: more resources for groundbreaking cancer research.
About the speaker: Miloud Ourahou
Miloud Ourahou has been involved in AI for eight years. He studied Big Data Analytics in Hong Kong and took the post-graduate course AI for Leaders at the University of Texas, with a focus on implementing AI in organizations.