How AI Helps Banks Manage Risk
By Ai4August 27, 2020
During this webinar, industry experts discussed how AI helps banks manage risk. We’ve included a short transcription of the webinar, beginning at 3:30 of the webinar.
Stuart Neilson, Citi: So here's the question: Let's start with risk. Broadly, how does AI help a bank manage risk? Secondly. How do you manage the risk of using models? How about for this one we start with Ruchi.
Ruchi Gupta, HSBC: Okay, I'm the lucky one. I think firstly before that, one of the biggest advantages I see of AI providing us is, you have so much unstructured data, and then you have so much structured data. So all the technologies you have could definitely help you with the basic regression techniques and everything. Definitely they help you in passing out all the structure data and everything. Assuming a data size is small, but as your data size increases a lot, the one thing you have to look at is AI. That's one of the benefits AI can give you. Then secondly will be the unstructured data, in which AI gives you a tremendous advantage. So definitely in the field looking at structured and unstructured data, that's AI helps you in managing your risk and it's not just managing. First of all, you also have to figure it out where a risk is and to understand that. So that's the benefit which AI provides.
In terms of how in my current role it helps us and how does the bank manage the model risk. At least at HSBC, we have a whole framework of it where there's an actual governance process involved. Then you have the more which takes care and makes sure that we follow the proper process. That's definitely one way it helps us in managing the risk.
Stuart Neilson, Citi: Anyone else want to chime in on that one?
Anusha Dandapani, Barclays: I'll go for it. The AI risk, from my perspective, the business use cases have gone up exponentially in recent times. There is a tremendous potential that exists on how we can make key decisions using AI or ML-based approaches, especially when it comes to products that we are already using in our decision making process. However, the key is that we do foresee AI that may affect organizations or consumers or they can also create broader detrimental effects on society such as this is typically or a rise in the whole or in part. Sources that we use, especially the data that we use to train these AI systems. These risks also are not just from the AI systems itself, it’s how we use these AI systems and what sort of decisions that we drive based on these systems. Also this advising base and the poor governance assistance. Those are the risks that we use AI to help manage.
There is also a big gap that currently exists in AI governance. This gap needs to be addressed from our perspective. What’s taxonomy around it? What are the mitigation techniques that we can actually use AI to resolve, especially in the financial industry and that is something that would be very interesting to watch out for.
Stuart Neilson, Citi: Thank you. Javed, any further comments?
Javed Ahmed, Metis Corporate Training: Yeah, I would say that in the way that you can you can calculate your exposure using a calculator. AI is a tool that can help you to calculate things when you're managing risk. I think it's been around there, have been some developments, but fundamentally risk management is a personal and human-oriented process where you can use tools and the tools can help you, but it's very dangerous to rely on those tools as a crutch. So as a point of reference in terms of banks and stress tests, the Federal Reserve and it's severely adverse scenario, which was published in February of this year, forecast that unemployment would go up to about a peak of 10% by the third quarter of 2021.
If that was your worst case scenario, we were way off in terms of what actually happened and what we might have been planning for. So this is not something that could be picked up by even the best of models. I think in the context of determining how much capital to allocate for a particular loan relative to another loan, using the best models can help us in terms of holistic risk management. I don't think that the game has changed there in terms of the human element and the importance of managerial input in risk management over the last 10-15 years.
Learn more and watch the full video on YouTube: https://youtu.be/tDhJJ1Dd-IA.
Introduction In today’s digital era, computer vision stands as a transformative technology, driving innovations across...
By Sumedh DatarDecember 12, 2023
Cryptocurrency has slowly made its way into the mainstream, and more people have begun thinking...
By Trix MejiaFebruary 25, 2022