Artificial intelligence is moving from experimentation to execution across banking, but scaling the technology effectively remains a significant challenge. For ING’s wholesale banking business, the approach has been to build the foundations first: modernising its core banking platforms, developing shared technology capabilities and focusing on AI use cases that can deliver value across its global footprint.
With the bank operating in more than 35 countries, this scalable operating model has become central to how ING approaches innovation. From machine-learning-powered algorithmic trading and AI-assisted lending to its exploration of agentic AI, the bank is now looking at how emerging technologies can accelerate processes while maintaining appropriate governance and controls.
In this interview with The Digital Banker, Bouke Hoving, CIO, Wholesale Banking at ING, discusses the bank’s AI strategy, the evolution from machine learning and generative AI towards agentic AI, and the importance of responsible deployment. He also explores AI’s implications for the banking workforce and how the bank is addressing the energy demands of the technology.
How has ING implemented and scaled artificial intelligence across the wholesale banking unit, and what were the key challenges you encountered during this implementation? What were some of the key lessons you learned in the process?
When we talk about AI, it’s always a question of how to scale it. Before we get into AI, it’s probably good to spend a bit of time on the overall operating model of the wholesale bank.
We operate in 35 countries and serve customers in more than 100 countries globally. Our operating model is highly centralized, with engineering platforms built and managed from three main locations around the world. We build platforms with sufficient flexibility so we can serve corporates and financial institutions across our diverse network. That’s something we’ve been working on for more than a decade.
Also, here in Asia, we’ve spent a lot of time, effort and investment modernising our core banking platforms, which were originally put in place 40 years ago. We’re now very well advanced in that process. Out of the nine markets where we had those platforms, we’ve completed the transformation in seven.
In general, we see innovation scaling fairly well across our network, and that’s something we benefit from. When it comes to AI, we’ve taken a focused approach, initially looking at use cases where we saw strong proof of concept and where we were equally convinced we could scale them across the network.
We’ve looked at use cases such as contact centres in our retail franchise. Specifically in wholesale banking, we’ve invested significant time and effort into KYC, lending, and even before generative AI, financial markets and e-trading.
Here in the region, last year we brought our algorithmic trading business live in Singapore. That capability allows clients to interact with ING digitally when hedging their currency exposure, with much faster execution powered by machine learning. It’s a good example of how innovation travels well on a scalable technology foundation.
As the focus shifts from generative AI to agentic AI, what potential do you see here, and where do you see the opportunities and risks?
We see agentic AI as the next step.
Lending is a good example. To give some colour on how that evolution has progressed over the past couple of years, we’ve seen a lot of promising AI use cases, even before generative AI through machine learning.
We were already implementing automated decision-making for lower-risk, smaller-ticket credit requests. For larger wholesale credit applications, credit analysts still need to form their own judgement.
There, we’ve been using generative AI as a co-pilot, taking away much of the heavy lifting from the credit analyst by pre-populating documents, searching for data in annual reports, and requesting information from local credit agencies. This helps accelerate the process and enables faster decisions.
As the next step, we’re also looking at agentic AI. We’re learning from our retail banking colleagues, who have already successfully brought agentic AI into production. In the Netherlands, we’re using it in the mortgage application process, and that serves as inspiration beyond our current implementations of machine learning and generative AI.
How has ING approached responsible artificial intelligence deployment? How do you ensure innovation is balanced with the appropriate levels of control and oversight?
Before any team is allowed to bring something into production, they must follow our governance processes to ensure the appropriate guardrails are in place.
What is also specific to our approach is that we’ve built a strong foundation and only focus on use cases that can scale across our footprint. That allows us to spend sufficient time rigorously testing the solutions before bringing them live and only launching those that can scale globally.
How is ING balancing AI innovation with sustainability, given concerns about the energy required to support AI models?
When it comes to sustainability, it’s very important for us to first look at whether a more traditional digital approach can solve the problem with lower energy consumption.
We use AI and generative AI only where they truly add value. For many cases, our existing straight-through digital processing already meets a significant portion of customer demand.
Secondly, we’re passionate about our global operating model, ensuring innovation travels quickly while focusing only on use cases that truly matter and can scale. That means we don’t waste effort, or energy consumption, on solutions with only marginal benefits.
Last but not least, we invest heavily in educating our staff on how to limit unnecessary energy consumption when using AI.
Concerns around workforce disruption continue to grow as AI adoption accelerates. From your perspective, what should banking leaders prioritise to ensure AI investments deliver lasting value?
If you look at how AI will affect jobs, we do see some impact, particularly on junior roles. There has been some reduction in headcount across the industry, and we’ve seen that within the wholesale bank as well.
At the same time, if you look at the medium to long term, every major technology innovation has raised similar concerns. Ultimately, we’ve always found there’s still so much work to be done that some of those negative effects are offset by new opportunities.
It’s still a little early to predict exactly how things will develop. We’re cautiously optimistic, while recognising that some junior roles may inevitably be substituted by generative AI.
As for ensuring AI investments deliver lasting value, I think every technology leader is trying to make the best use of generative AI to capture its promising potential.
It’s crucial to get the AI basics right first. Once you’ve established those foundations, revisit your existing work. You’ll find your traditional modernisation challenges can be addressed much more efficiently.
Once you’ve mastered that, you can begin reimagining business processes, where there’s phenomenal potential. However, if you approach AI incrementally while leaving your business transformation and technology modernisation untouched, you’ll struggle to develop sufficient AI maturity to make a meaningful impact.
That’s why it’s vitally important to build strong AI capabilities first, then revisit your modernisation and business transformation agenda. It’s easy to say, but still hard to do every day.
Is there anything else you’d like to add?
I think we’ve covered the overall AI strategy. Perhaps it’s good to round off by touching on our global operating model transformation and how we’ve organised the workforce supporting these initiatives.
As I mentioned briefly, we’re going through a major modernisation of our core banking landscape, which is now entering its final phase. In Asia, seven of our branches are already operating on the modernised core banking platforms.
We’re already seeing tangible benefits for clients through a much better and more consistent digital experience across our network. It’s also helping us introduce additional capabilities into the region. I mentioned electronic and commodity trading, but more broadly, innovation is travelling much faster across our network—not only AI innovations, but also products that previously weren’t available in this region.
Secondly, as we roll out these global platforms across Asia, Europe and the Americas, we’re also globalising our workforce. Previously, engineering was largely centralised in Europe. Today, we have three global development centres – in Manila, Bucharest and Amsterdam – where these platforms are built.
Regional locations such as Singapore play a crucial role in adapting those platforms for local Asian markets.
In essence, our strategy rests on three pillars: a global operating model built on shared platforms, a globalised engineering workforce across our three development centres, and AI powering all of this to help us move into a higher gear.
