This is the second episode in the Accern Series on the Mighty Capital Podcast. Once again, our host is Accern Co-Founder and CTO Anshul Pandey. Here, Anshul sits down with LPA Consulting Partner Daniela Rothley and LPA Manager of Data Science & Machine Learning George Karapateyan. Their conversation centers on the role of banks in digital transformation. Along the way, they discuss the recent cautiousness in capital markets. They also consider the role of artificial intelligence (AI) in ESG (environmental, social, and governance). Finally, they discuss what banks can do to reduce risks when scaling AI solutions.
On the recent cautiousness in capital markets
Daniela notes that in capital markets in recent years there has been much activity around innovation in artificial intelligence. Over the past year, however, these markets have become more cautious. Innovators are realizing that AI solutions are not so easy to implement. And with inflation, war in Europe, and a looming energy crisis, investors seem to be getting more conservative.
Daniela says, “A lot of clients are sitting and waiting. They are still very open for a lot of things and discussions. But they are waiting to implement something really new. So a few years in the past, I would say they were willing to take more risk. Now they are much more averse to risk, from my perspective and overall business sense.”
According to George, “In the flow of priorities that the bank has for themselves, there are so many projects that are already running. Sometimes there is fatigue of different topics that need to be managed at the same time. That innovation or AI are some things that need an additional effort. And sometimes people either stay away from it or wait for some kind of already out of the box solution available in the market that they could buy instead of creating a project from it.
“More regulatory required topics, for example, ESG, are coming more into the forefront. Or topics related to sanction checks. For example, we saw with the current geopolitical situation that there were new sanction packages coming every few months. This was already putting a lot of administrative effort on the financial institutions to adjust their processes. And this was coming more into the priorities list rather than initiating new things. I would describe it as more tactical planning and tactical implementation. Rather than long term strategic planning for the current market.”
On the role of AI in ESG
Steve Jobs said you can always connect the dots looking backwards. For instance, we can now trace how ESG (environmental, social, and governance) became such an important topic. It makes sense when looking at all the different things that are happening at a macro level. But what are some of the recent ESG developments that stand out, and does AI still play a role?
George says, “One area where we see some automation opportunities is, for example, in carrying out ESG customer due diligence questionnaires. The point of these is to understand what are some of the policies that the client or the counterparty has with regard to E, S, or G factors. These questionnaires are sometimes very tedious to respond to. When you need to do it on a much larger scale, and when your relationship with the counterparty is more of a customer and seller relationship, or financier and a client relationship, you might want to also automate and make this process as simple as possible.
“One way to do that is, for example, to extract from existing unstructured sources of data potential suggested answers to these types of questions. Where a human can have a look at and confirm. Rather than search in thousands of pages of documents to find one specific policy or criteria for measuring this or that. This is where I see a potential application of AI, in text analytics and natural language processing.”
On how banks can reduce the risks in scaling AI
While there's a lot of appreciation for AI within product management generally, that sentiment might not be shared everywhere. Indeed, it’s important not to delude ourselves that just because we understand AI and build AI models, that the rest of the world also interprets and sees it in the same way. Right now there appears to be a lot of backlash against the phrase AI. So how can banks reduce the risk of scaling AI projects to really deliver transformation?
According to George, “The technology itself has moved from its hype phase to a more mature phase. There is a lot of disillusionment among the people who have applied it, or tried to apply the technology to solve one or the other problem. They realize that it's not just, give it to AI and it will do it for you. But rather, you need to define concrete problems, and source the necessary data. You need to be able to build a good model. You also need to monitor the model, update it, and then implement it. There are a lot of challenges connected to this.
“That's why I often avoid even using the word AI in conversations. Because it's just creating the sense of, okay, there is some kind of a magic machine that is going to do everything that I want. Or the narrative is, oh, there is some kind of a machine that is going to take my job away. Which in either case is not true. That's why I often just call it an intelligent solution, or smart solution. Because this goes into the narrative that we augment the process. We improve it. Make it smarter. And we take out the dumb parts of the process, which are more just simply looking at things, searching for keywords, or doing repetitive tasks, and automating that in a more clever way.”
The reason for the backlash
George continues, “I think that one reason for this backlash is an internal conflict within financial institutions. On the one hand are the practitioners that want to apply the technology. They are from IT or quant teams that know how to code themselves, and they are very excited to try out new technology. On the business side, they just want something that functions and is very fast. And sometimes the business has an idea. But for them to have this idea tested out, it takes a very long process and a long time to get to the result of that. Oftentimes, the business just avoids doing this.
“This is where I see, for example, no code actually bridging this gap. Because business-oriented users can test the hypotheses by just simple drag and drop configurations. The barrier to that is just going to be the IT departments in financial institutions. They might see a threat to that. But I think this trend cannot be stopped because it is quite big.”