How AI is like a Ferrari (and you better know how to drive it!)

 

We are meeting here in Zurich, where you just spoke on the topic of AI and how it’s going to revolutionize and transform finance. One of the messages that you tried to convey was: “AI as a tool.” Can you elaborate on what you mean by that?

I wanted to make the difference between “AI as a tool” and “AI with agency.” “AI as a tool”—depending on the type of AI you’re using, depending on the model you use, and how you train it—can be a safe environment, where you can work with it and you can use it to your advantage to the objective that you have. And it works. We’ve seen it. It does a good job. But when the finance sector decides, “well, we’ll just let them do whatever they want to do,” that, first of all, raises a huge level of risk. When these systems start having agency—so when they become autonomous, then they have their own agency—they can decide what they want.

 

What does that mean? What kind of limitations or risks are there?

We have limitations in terms of risks of nonconformity of the outcome, which the regulator wants to know. You use AI, it has to give out the same outcome every time. So, for instance, it doesn’t discriminate against women or people of different backgrounds. But it has to be consistent. Now, with agency they can decide differently. That’s a huge risk. So I try to make the difference between “AI as a tool” and “autonomous AI with agency” simply because between these two areas there’s a lot of risk. One needs to understand it first, understand the tools, understand how they become autonomous, what it means, evaluate the risks. And only then, perhaps, they can say, “OK, so let’s do a little bit of autonomy here.”

 

It was interesting what you said during the panel about literacy, that basically everyone from top management and executives to the receptionist needs to understand it.

I have been advocating AI literacy from the chairman of the board—not even the CEO, the chairman of the board—all the way down to the receptionist. And the clients we work with, actually a few years back, they actually started investing in AI literacy. And they have seen a lot of mobility within the organization, where people would move—say from the reception desk into the data science team. That’s a big advantage for an organization, because it’s the same person. You have your people moving within the organization instead of going out and trying to bring talent in. When people at the board level, who have an oversight function, understand this technology, they understand the upside, and they can use it. They think, OK, we can use it as a tool, as a growth tool, as an expansion tool, because we understand where we want to go with this strategy.

 

What kind of banking divisions will be most affected by AI?

Everybody will be affected. There is absolutely no function in financial services or in any institution in financial services that will not be affected. But this is a very important point I’d like to make here: When they look at adopting AI, the understanding is sometimes like, “oh, we’re going to do a little bit here, see how it works, and then maybe we’ll test it, we’ll put it into practice, in production, we’ll go live.” Excellent! Actually, my recommendation has always been to try to define your strategy. Where do you want to go? How many clients do you want to reach? Assets under management, depending on your business model, jurisdictions, where do you wanna go? Outside? Stay in? Move to other parts of the country? Move to other countries? So that is the business strategy, and the technology has to follow it and has to support that strategy. Where financial services institutions typically lose out—because they invest a lot of resources in building and investing in AI, testing them, maintaining them—it’s when they look in a siloed approach. “OK, we’re just going to deal with the finance function or just with HR or just back office or just settlements.” Actually when you do that, it becomes very expensive over time, because you have to patch all the different projects across the organization, and sometimes they don’t talk to each other, and then you have an extra headache.

 

So you really need to think of where you want to invest and where the opportunities are?

You need to have a good old-fashioned business strategy. Once you’ve identified that,
then you go, “OK, so what tools do we have?”

And then AI is an opportunity?

AI is absolutely an opportunity. It gets you from A to B really fast, but you need to know where you’re going. And that actually concentrates minds, concentrates attention, and they go, “we have a destination.” And sometimes you can deploy AI and you can deploy very simple models. It’s like the equivalent of a little tricycle: You get to know how it works. But sometimes you can just jump straight into a Ferrari and go to the destination. Generative AI typically is associated with a Ferrari because it gets you so quickly to where you want. But you need to know how to drive a fast car.

 

When you know in which direction you go, you also need the framework, a regulatory framework that supports it. What can you say about that?

Well, coming back to the analogy with Ferrari. We drive fast cars on roads that have demarcations. We know which side of the road we drive. So we stay safe. We need those regulations, and I’m a big proponent of the work that the European Union has done in the space of AI regulations. And I think in the UK we need to learn a lot from what they’ve done. And I always keep saying in the UK post-Brexit that we should be more like Switzerland.

Why?

Because I think we have so much in common in terms of ideas, types of constituents, everything we have at our disposal as countries. We need to be tactical and strategic, just like the Swiss—and as clear and predictable. I think we have a lot to learn from you
guys.