1 month ago
4
ARTICLE AD BOX
As financial institutions race to modernise, many still struggle to turn AI hype into tangible results. Marnix van Stiphout, ING’s chief operations officer, says investment in staff education and sustainable scale-up is the key to turning this tech into a powerful asset.
Since stepping into the role in 2021, van Stiphout has led a major transformation of ING’s global delivery hubs – including a 33% increase in headcount in the Philippines in 2024 alone – to support smarter, faster AI deployment across the business. The bank’s distributed analytics team now spans Amsterdam, Manila, Istanbul, Bucharest, Madrid and beyond, rolling out AI to enhance everything from transaction monitoring to marketing.
While ING has been leveraging machine learning for years, the company recently launched generative AI chatbots for retail markets. The bank is also developing agentic AI to streamline mortgages and bring non-digital customers into the fold. In this interview, edited for length and clarity, Stiphout discusses the still untapped potential of emerging AI tools – and how companies can work to stay ahead of the curve.
How to succeed with AI as a financial institution? Have a smart training strategy for staff, says ING’s Marnix van Stiphout. (Photo: ING)Can you describe ING’s approach to AI adoption?
Marnix van Stiphout: When it comes to AI, we’ve been using machine learning for years, particularly in retail but also across other areas of the bank. There are all kinds of decisions you need to make in banking — how to handle collections, or fraud – and there’s a lot of data we can use to analyse what the risk and response of a customer may be. In these instances, you can use machine learning models to make your decision and, ultimately, make your systems more powerful and more robust.
Over the past 18 months, we’ve expanded into generative AI. One key example is our chatbot rolled out to most of our retail markets, which we developed in partnership with Google. We have five focus areas: know your customer (KYC), hyper-personalised marketing, wholesale banking lending, chatbots in contact centres, and IT software engineering. Beyond that, we’ve started piloting agentic AI solutions across mortgages, and we’re exploring voice-based agentic AI in contact centres to complement our chatbot systems.
Aside from speed, what are the benefits of using AI in this industry?
You could certainly argue there’s a democratisation of digital services with the adoption of these AI tools. Even as a digital-first bank, we’re aware that not all our customers are fully digital; for instance, older clients or people who, for various reasons, prefer not to use apps. That’s where agentic AI can be a game-changer.
If we can deploy voice-enabled agentic AI in our contact centres, we can offer a digital experience to those who wouldn’t normally engage that way. Voice agents could proactively reach out, help with daily financial check-ins, or suggest actions and relevant products. Whatever it is, these non-digital clients will suddenly gain access to very modern digital services.
In your view, what’s the biggest disconnect between the promise and reality of AI in banking, and what’s holding it back?
Machine learning has been around for years, whereas generative or agentic AI is really only at the beginning of its journey. So we’re still seeing where these tools are going to go.
Something that has proven problematic is that some companies have chosen to be as broad as they can be in the application and rollout of these tools, and have really encouraged experimentation across the organisation. Having no focus leads to only limited impact. Where ING differs here is that we focused on a couple of really big areas where we believed we could generate the most impact, and we made sure we focused our people on those teaching areas.
I think one of the main reasons that some of these things get more difficult than people anticipate is that you need to do a lot of training internally. You need to mobilise people around these topics, and that takes time. So, the combination of new technology and new training, if you do it across 50 different things, it’s very difficult to realise. If you do it across five things, it’s easier. It’s not really about AI – it’s about creating impact. So you need to be very careful about what you do with AI and how you choose to adopt it.
When it comes to retraining staff for AI, what are the challenges?
It starts with training the right technical teams, data scientists, AI specialists, and engineers who can build and deploy models effectively. But more challenging, I would argue, is that to deploy agentic AI across something like mortgages, you need your existing teams to be able to work on the solutions. That means you need to retrain people across the bank, not just those in the tech roles. For example, your business teams, your product managers, need to be trained gradually on the topic. You need all of these people to have the skills and capabilities to develop these solutions, and that’s a pretty big task.
AI adoption can’t be a niche initiative. If you want to apply it meaningfully at product level, everyone involved in the product lifecycle needs to understand how to work with these tools.
Take KYC, and customer due diligence within that. Both processes are data-heavy, manual processes, especially for business clients. An agentic AI solution can help gather missing data or even reach out to customers directly in a structured, digital way. But that only works if the process is built correctly and the right teams are on board and trained.
We see major potential across mortgages, KYC, and other product lines. But implementing these solutions at scale takes time, thoughtful design, and an organisation-wide understanding of how to work with AI.
When handling customer data, there are naturally questions over security. How do you incorporate that into your strategy?
Maybe one of the most positive sides of banking when it comes to AI is that we are heavily regulated across companies, so data security and data quality have always been big topics. That’s not to say there’s nothing more to be done, but I would argue we’re well-positioned as an industry to take on this task, more so than other sectors.
In terms of what we are doing at ING, we have a very comprehensive risk framework with plenty of scenarios that we go through for any of our AI solutions in terms of data security. We make sure there are boundaries to what we do and that we stay within the guardrails identified.
Has it been difficult culturally to get people to embrace AI?
At the moment, no, but it’s a different landscape today than it was yesterday. In 2025, there’s a lot more understanding of AI, not only within banking but also more generally, than we saw even last year. Whether it’s in machine learning or other areas, people have started to use some of these innovations themselves, and they see the opportunity.
That said, we are mindful of the personal impact. There will be less manual labour and more automated work, and that will create a different work environment. People are already embracing it, but we do have to be mindful of the social consequences.
To prepare for this, ING is focusing on reskilling and re-educating people. For one thing, we have launched a whole new curriculum for people to get better at data fluency. So, we’re trying to bring people as best we can from where they are today to a new type of job.
What are going to be your next areas of focus for AI at ING?
We’re continuing to scale in several core areas: KYC, lending, personalised marketing, and engineering. These remain key focus areas with a lot of runway.
Agentic AI is now being applied to mortgages in the Netherlands and will expand to other countries. Beyond that, we’re looking at investment products, lending, and improving straight-through processing across the board. We’re also using GenAI and LLMs internally to accelerate product development. For example, applying LLMs to policy documentation can help our teams get to market faster by automating compliance checks and internal approvals.
In short, AI is becoming embedded in how we design, launch, and operate our products, and we’re really only at the beginning of that journey.
