Vroukje van Oosten Slingeland (ING Group) on AI Applications in Legal

Vroukje van Oosten Slingeland (ING Group) on AI Applications in Legal
When it comes to generative artificial intelligence (GenAI), the egg of Columbus has not yet been hatched, says Vroukje van Oosten Slingeland, general counsel at ING. The future potential of AI is great, she thinks, also for lawyers: ‘However, human intervention will always be important to steer artificial intelligence in the right direction.’

ING’s ambition is to make banking frictionless. ING is experimenting with artificial intelligence (ai) in a variety of applications, such as for personalization and fraud detection, but also to help companies meet ING's climate requirements. Generative AI (‘GenAI’, ChatGPT and other applications that can generate their own texts and other output) obviously also enjoy the bank’s interest.
The bank is, of course, proceeding ‘prudently’ with all experiments, says general counsel Vroukje van Oosten Slingeland to Hans Albers, director of legal management consulting at Deloitte. ‘We approach it cautiously, to understand what GenAI can achieve and why. We have selected some use cases in which we look at it further, beyond that it cannot be applied due to the still mostly unknown risks we have to mitigate. For example, we explicitly do not work with ING data on public AI platforms; this is strictly forbidden and in line with ING policy. The enthusiasm is great, but we keep in mind in everything we do that we are under scrutiny and must not betray the trust society has in us. This is also one of the reasons we are not currently experimenting with GenAI in the legal department. With an application like ChatGPT, there simply is a risk of disclosing the bank’s sensitive information. We want to avoid that at all costs. To run an application like that from a location of our own is already going too far for us: it sends metadata to the provider, and we do not want that. At least not until we are 100 percent sure what data is exported and whether that is permitted.’

Many lawyers are enthusiastic about generative AI, as you yourself observed at Deloitte’s recent conference on the subject which you attended. Do you share that enthusiasm for generative AI?
‘I am optimistic, but I have yet to see what it will offer us. Typical attitude for a lawyer, perhaps. Another factor is that I have not yet seen a specific, existing GenAI application that I think will be of appreciable immediate benefit to us as a legal department. The egg of Columbus has yet to be hatched. Generative AI is still in its infancy, although developments are going so fast that it could look different in a year’s time. However, I do see all kinds of promising developments. The promise for the future is great.’

Can you give an example of that?
‘Generative AI may soon enable lawyers to do their work more efficiently because it allows them to sift through and analyze information so quickly. Think of searching through large databases for due diligence. There will certainly come a time when we will use generative AI for this purpose - provided it is secure and well-regulated and provided the price is proportionate to value.
I can also see AI playing a role in customer contact. A few years ago in the legal department, we experimented with a chatbot for people within the bank with legal questions. We kept getting the same questions from business. A waste of our capacity, we felt, to have to reinvent the wheel each time. Hence the chatbot. With the added advantage that the lawyers who had to enter the questions and answers into the chatbot system would gain more affinity with technology and become more ‘tech-savvy.’ An incredibly fun pilot. Yet at some point we pulled the plug on it. It was very labor-intensive, but the result was never as perfect as we wanted it to be. And, most importantly: the chatbot was not being used enough. People still found it preferable to just call a legal colleague. Maybe someday there will be such a chatbot again - if it is more of a plug-and-play solution that requires limited further work and is reasonably priced.
Also, perhaps AI can help get a better handle on relevant laws and regulations. The quantity and complexity of these are only increasing. I think every lawyer and legal department struggles with this, and within an institution like ING, which has to deal with various regulators, this is especially true. It becomes more and more of a challenge to keep developments in legislation and case law manageable.’

It would be even better if you could generate predictive insights with AI. For example, we are now trying to use AI at a financial institution to see if we can predict at an early stage which complaints could lead to claims and lawsuits. Also, a legal department can use AI to make data-driven decisions or support arguments with data: for example, to set the budget. If you can predict how many contracts you need to draw up in the coming year, you might be able to deduce from that how many people are needed to do that and how much budget should be allocated for it.
‘That is super interesting. But I am faced with two challenges. First, how do I make sure my team is ready to use this technology? And furthermore, how do I make sure that the lawyers are not restricted in their own development?’

Can you elaborate?
‘Consider a lawyer who needs to prepare a case for court, and the tool performs a search of case law within minutes - whereas before, you used to spend hours in the library to gather useful arguments. That sounds very appealing, but the lawyer still must make an informed distinction between valid and invalid arguments. He or she can only do that if he or she has learned to do so. It will be a challenge for universities, companies, and the legal profession to train people to be able to do that.
I myself learned the legal profession mostly in practice. As a contract lawyer, I learned to edit the old-fashioned way - mainly by doing it many times and often being held accountable to those who trained me. It is a long learning curve that in my day you could not learn by simply using the templates that were available then. The same goes for AI solutions now: you have to be able to evaluate them. ChatGPT is quite capable of fabricating a passable warranty agreement or another legal document. However, how do lawyers who have never drafted such a document before know what it should contain and what to look for? They cannot do that if they only know how to use such a tool; they also need to have logged flying hours as lawyers.
So we try to develop lawyers on both fronts. We try to familiarize them with technology and digitization with all kinds of courses. Think of days where complex topics for lawyers are dealt with, and questions like: what is generative AI? And: what do you need to know and understand about this as a lawyer? We also have legal tech talks, for example, where IT lawyers, often with someone from the business,  give a presentation on new technological developments. In addition, we also offer internal training courses together with law firms and people from the industry which cover complex topics for lawyers.’

What do you think the legal department will look like in ten years?
‘That is difficult to predict. Ten years is a long time, and especially when developments are as rapid as they are with generative AI, you are quickly blowing hot air if you venture into predictions. That is not helpful. But if I have to make an attempt: I think that thanks to tools like generative AI, we will be able to make huge efficiency gains in the coming years. But I do not believe that entire departments will be redundant. Human intervention will always be important to manage the activities of AI.
In addition, lawyers will have more time for strategic and intellectually challenging issues. To do more risk mitigation work, for example, to prevent future problems - something for which time is often lacking now.
In other words, thanks to the efficiency gains we can make with technology, we can do deeper work ourselves. New positions will also emerge. In our department, we see that happening already and will see more and more of that in the coming years. Sometimes these are people who have gone into the technical side from a legal background, but there are also paralegals among them who have a technical background and who look at optimizing and digitizing legal processes, activities, and applications.’

Does that not mean, then, that the importance of factual knowledge is shifting somewhat into the background for most lawyers? And that conceptual thinking, creativity, strategic and advisory skills are becoming more important?
‘If you boil it down completely, I do not think so. Every lawyer will eventually still need to possess the same basic skills as they do now. Simply put, you should still be able to write an argument with an introduction, a problem statement, list the relevant facts, the legal framework, or another framework, apply that to the facts and then come to a clear conclusion. Whether you do that in a comprehensive litigation piece, a memo to your boss, a memo to an internal stakeholder, or an advice. And sometimes you throw it in a bit of a different order, like with the executive summary you move to the top of the page. Those skills continue to be needed.
What does change is how you convey that argument. Because things move much faster than they used to, you have little time and you have to know how to get things across to the point, quickly, and accurately. Generative AI can of course be very helpful in this: to offer ideas like a brainstorming buddy, to draft a text or to rephrase an existing text. But here too, it is important that you know how to assess the text. And have sufficient factual and analytical knowledge to do so.’

Law firms and outside consultants working for ING also use AI. How do you expect their services to evolve? And what does this mean for their business model and fees?
‘I do not yet see it leading to the ‘end of the billable hour’, as is sometimes predicted. I do hope that the advance of AI will lead to externals working more efficiently and that we will see some of that in their invoices. But maybe the effect will not be as substantial, since those efficiency gains will be made mainly in work for which the highest rates are not charged anyway.
In addition, I would find it very interesting to be able to purchase services on a sort of pay-as-you-go basis and not necessarily on the traditional hourly billing basis. For example, I can imagine that if we need to do due diligence on a loan portfolio to see how much collateral there is, or to see what our risk-weighted assets are, we could have those contracts reviewed by a specialist party that has an AI application for that. And that we then pay for it per document.
Though we could eventually do this type of work ourselves, without a law firm or other party involved. If I were a lawyer, I would not be too sad about that. In fact, I think you have to have the courage to move away from that type of work. In my opinion that is not the most interesting work you can do for clients and as AI becomes more sophisticated, it also becomes less and less lucrative.’

This interview was published in Management Scope 02 2024.

This article was last changed on 06-02-2024