ING’s chief operations officer (COO) has said that “everything and anything” the bank does in product fulfilment “can and probably will” be transformed by agentic AI solutions.

Speaking to FStech, Marnix van Stiphout added that all banking operations, from lending and mortgages to investments, could gain value from the technology.

Agentic AI is able to solve highly complex and multi-step problems through sophisticated reasoning and iterative planning. What sets the technology apart from previous iterations of AI or genAI is that it can make decisions autonomously.

The COO's comments come as the Dutch bank trials agentic AI with voice agents in Spain and Germany to handle routine call centre tasks using advanced speech technology.

From next year, the bank is also planning to use the technology to initiate and complete mortgage applications in the Netherlands and Germany.

"We'll take it to other countries too," van Stiphout told FStech. "That's going to give us a lot of extra productivity and time to focus on those products for both existing and new to bank customers."

He said that in mortgages, one of the agents will work on document reading.

"In different markets, or even in just single markets, there's a whole range of documents that you can expect from customers," explained the COO. "So, when you develop a document reader agent for the Dutch business, we'll then use that in the German business, and in other markets."

The company's agentic AI focus supports its plans going into 2026 to continue growing the business, with the bank adding segments such as business, private and affluent banking in Germany, Italy, Spain, and Australia.

"We use the opportunity of scaling our technology into those markets to create depth and value," continued van Stiphout.

He explained that agentic AI solutions could be used across all lending products, not just mortgages, including business banking lending, consumer lending, and wholesale banking lending.

"These agents will be doing a lot of work across the whole operational space," he added. "It will change what we ask our colleagues to do, and people will become less product-oriented and more process or agentic-oriented."

Complexity in upskilling as the bank scales agentic AI

When asked how AI training and upskilling is going as the bank advances with agentic AI, van Stiphout said that at the moment it is "manageable".

"When you do this for a couple of agentic solutions, however big they are for us, it's relatively straightforward to do," he explained. "When you need to do this across lots of different markets and products, then the reskilling becomes far more complex."

The COO added: "The straight answer today is that it's fine, but if you would ask me to look ahead six months or 12 months, then I still have some uncertainty on how we're going to do this at scale."

GenAI progress

After FStech spoke with ING's chief technology officer Daniele Tonella in June, when he described the bank's 40 genAI use cases as at different stages, with some industrialised, and others scaling or in early exploration, van Spiphout said that six months later the company has introduced behavioural modelling in anti-money laundering (AML).

“There's a lot of machine learning still being launched there," he explained. "Also in anti-fraud, for example, and e-banking models, it's very important for preventing scams. So we have been very active over the past six months in a couple of markets."

The bank has also recently been working with a new Google engagement suite to launch simultaneous voice and chat solutions in its contact centres.

"We engineer something, and it gets launched in voice and chat at the same time," he explained.

The move comes as the bank has been working on removing friction form the contact centre over the past 18 months, which has led to fewer friction-related calls.

"When we see in certain markets that a journey on a chatbot is creating unsatisfactory outcomes for clients more than you would expect, we turn it off," he added, giving an example of one way it has reduced friction. "We have human chat on the back of it or a call on the back of it."

He said that this is showing good results, adding that while there are no plans to solely rely on AI for the contact centre in future, the bank does want to have a completely Straight-Through Processing (STP) approach so that customers can decide whether they want to call in or take a self-service route.

ROI

When asked whether the bank distinguishes between AI projects that have a fast return on investment and longer-term, sustainable projects, he said that the bank has only gone for the "biggest cases" and will continue to do so.

"That's also where we've got the biggest yield, but also the biggest chance of success because the people that have built these big environments are capable of actually building them again differently," he continued.

---