AI lessons from the field
Why focus and technology go hand in hand
ING's COO Risk Leon Dusée on AI in risk management, why pilots often stall, and how a promising AI solution such as a financial health assistant could help people sooner
By Rutger Huizenga
AI is now on almost every agenda. Not as a gadget, but as a strategic question: where do you apply it, how do you scale it, and how do you prevent experiments from getting stuck in pilot phase? In the interview series "AI Lessons from the Field," we speak with leaders who have to make those choices in practice. Today: Leon Dusée, Global Head Chief Operating Officer Risk at ING. “AI can add value in thousands of places, but if you start everywhere, you often end up nowhere.”
When speaking with Leon Dusée about the future of banking, it quickly becomes clear that he is not only focused on the latest technological gadgets. Although, as COO Risk at ING, he is responsible for the security and stability of a global digital infrastructure, his focus lies on a surprisingly human aspect: making deliberate choices.
“AI can add value in thousands of places,” he says matter-of-factly. “But that is exactly the danger. If you start everywhere, you often end up nowhere.”
It is a lesson he learned in practice at the bank, where he started eight years ago as a management trainee and rose to global leadership of the operations connected to risk management. Technological choices become reality faster when the entire organization has the right focus.
“Everyone sees opportunities in AI, and those opportunities are real. But in a large organization, the temptation is strong to experiment in many places at once. Before long, you get a proliferation of pilots, proofs of concept, and isolated ideas that never truly land in the core process. The art is to say: this is where we will invest, and this is where we will not. That sounds simple, but it is not. Because it also means saying no to good ideas, to initiatives that may seem smarter or more efficient in themselves.”
Critical thinking
In the operations domain of Risk, for which he is responsible, Dusée clearly sees the difference AI can make. Much of the work involves assessing huge volumes of information: plans, validations, controls, and documentation. A large part of that is still manual and time-consuming. According to Dusée, AI can add a great deal of value here, for example by extracting, organizing, and analyzing information more quickly. But for him, the biggest change is not about speed. It is about what AI asks of people.
“AI forces employees to work more critically,” says Dusée. “You have to learn not only how to use it, but also how to question it. Where are the limits? When do you trust the output, and when do you not? What thresholds do you apply? That requires expertise, not blind trust.”
He adds: “The pace of change is high. Two years ago, we were talking about chatbots; now we are already talking about agentic AI and conversational AI. That is why you should not only train employees on today’s tools, but above all teach them how to keep adapting, spot opportunities, and help steer AI adoption.”
That is why ING, he says, is deliberately investing in training, with programs for both employees and leaders. Not because everyone needs to become a technical specialist, but because AI does not work if it remains a toy for a small group of experts.
“Often, someone spots an opportunity within their own scope, for example when creating a report. That is a good start, but for leaders the real work only begins after that: can you apply such a solution more broadly? Not for one report, but for all reports, or perhaps even for the entire process? That requires choices. You cannot do everything at once. Sometimes you have to say very explicitly: these are the six use cases we will focus on this year, and we will leave those other twenty aside for now.”
Continuing to challenge and question yourself is also important when it comes to changing processes through AI, he emphasizes.
“The first reflex is often: how can I add AI while keeping everything else the same? But the more interesting question is: should this process actually be fundamentally redesigned?”
At ING, he sees this, for example, in the manual processes used to collect data and make it available for models. The question is not only which step can be made faster, but whether the entire chain should be redesigned.
“Only then does AI become a strategic question,” he says. “Not because executives are operating the technology themselves, but because they determine what gets priority, how it is embedded and who is responsible for it.”
That, according to Dusée, is where things often go wrong in practice. Not with the technology, but with the organization.
“Many pilots fail because organizations think about day two too late. Teams build something in a safe environment and it works. But only then does the real question arise: where does this land in the organization? In which process? Which team becomes the owner? How do you maintain it? How does it connect to existing systems? If you only start asking those questions after the pilot, you get friction. Other teams have not been involved, the processes do not align, or there is simply no capacity to roll it out. Then something remains stuck in a kind of innovation limbo.”
To prevent this, ING has set up a clear structure for AI: a central analytics team that oversees the portfolio, while concrete ideas can still come from the shop floor.
“That central direction is essential to prevent teams from reinventing the wheel in their own small corner. The aim is to work in an ‘agnostic’ way: a solution that works for one type of reporting should be immediately scalable to all reporting within the bank.”
KickstartAI
The questions Dusée encounters within ING are by no means unique to banks, he says. That is precisely why he is active within KickstartAI, the partnership involving organizations including KLM, NS, ING and Ahold Delhaize, which aims to accelerate AI adoption in the Netherlands. The sectors may differ, but the underlying AI challenges are often much more similar.
“KLM’s market is completely different from ING’s,” he says. “But you encounter the same challenges. From governance and data to the question: how do you move from a proof of concept to a real product?” For organizations, he believes, the value is immediate: learning faster what works and what does not.
For Dusée, KickstartAI also has a broader meaning than knowledge exchange between large companies. He explicitly sees it as a way to create more societal impact. As an example, he mentions an AI tool around financial health that ING is currently exploring together with KickstartAI, with the aim of eventually investigating how customers can be helped earlier and more accessibly when financial problems are at risk of emerging.
Dusée: “How do you reach people earlier, lower barriers and offer help in a way that matches what people need? I find that at least as relevant. If we approach AI only as a machine for working more efficiently, we are selling ourselves short.”
But for Dusée, the value of KickstartAI extends beyond the partners alone. He sees the network not only as a place where large organizations exchange knowledge, but also as a way to make expertise more broadly available.
“There are so many organizations in the Netherlands, both profit and non-profit, where AI can be incredibly valuable,” he says. “In education, research, and healthcare, for example. But not everyone has the financial resources or capacity to apply AI quickly. How valuable is it, then, if we can contribute through KickstartAI and make our knowledge available?”
That is why he hopes more organizations will join. Not because every partner has to do the same thing, but because a larger network, in his view, makes more possible: more exchange, more learning capacity and ultimately more societal impact.
Dusée: “Those who work together not only increase the chances of scaling within their own organization, but also of creating real societal impact.”
Five lessons from Leon Dusée for other organizations
1. Choose sharply, and consciously leave things aside
According to Dusée, the biggest mistake is not experimenting too little, but wanting to do too much at once. Leaders must dare to say: this is where we will focus, and not there for now. Without focus, AI remains fragmented.
2. Look beyond the individual use case to the entire process
Many ideas rightly emerge bottom-up, but remain too small. The leadership question, he says, is always: can you broaden an application from one report or task to the entire process? That is where real scale lies.
3. Think about day two from day one
Pilots often stall because ownership, maintenance and embedding in existing processes are considered too late. Do not optimize for the demo, but for practice.
4. Invest as much in people as in technology
AI raises the demands placed on employees. Subject-matter expertise, adaptability and the ability to challenge output become more important, not less. Dusée also explicitly mentions learning on the job and a different perspective on education.
5. See AI as societal infrastructure too
According to Dusée, AI should not only be about productivity. Especially in sectors such as education, research and healthcare, collaboration can help make applications more accessible to organizations that do not themselves have the resources to lead the way.
