Artificial Intelligence and Machine Learning in the financial industry, that was the title of my virtual presentation for the Hispanics in Tech ERG at Capital One and I have to tell you, it was a good one. It’s not just because I love the topic, it’s also because the interest of the audience in this subject makes it super engaging. So, what is Artificial Intelligence anyways?
It sounds simple, artificial + intelligence = awesome. The reality is a little more nuanced as these systems are incredibly complex and can revolutionize the financial industry, and already are in many ways, but they also come with a slew of challenges and limitations.
In my presentation, I covered some ways in which the financial industry has been an early adopter of this technology in many silos, such as fraud management. According to IBM research, fraud costs the financial industry approximately $80 billion annually, which is not chump change. Using AI, the false positives for bad financial transactions and fraud overall have gone from a whopping 80% approximately to less than 1%. You may have noticed that when you travel now, your card gets canceled a lot less and you have a lot less trouble. At least that’s happened to me, a frequent traveler, that used to be a serial angry customer when my credit cards were canceled anytime I left Miami, and trust me, it happened often.
In this presentation, I also covered a little bit of the origins of Artificial Intelligence in the mid 20th century and how new technologies have propelled its use in the last 10 years. I also covered some use cases for regulatory compliance, a huge market that’s not as popular as Fintech, actually called Regtech, that’s expected to reach a market value of over 6 billion this year alone. We also discussed some front-office applications such as Customer Service chatbots, Robo-advisors, and other practical applications down the stack for risk assessment and credit scoring among others.
Other topics included the challenges around Artificial intelligence and the gargantuan task of feeding the machine with enough examples to train the systems to get the expected results. Data requirements for training deep learning systems are substantially larger than traditional analytic setups with some cases needing millions of labeled examples to properly train the AI, not a small feat.
We also covered some of the current challenges for financial institutions to implement AI including compliance and the very nature of financial data that might be siloed in different systems, departments, divisions and not readily accessible (on purpose). These being some of the technical hurdles that don’t take into account the political ones that exist in financial organizations.
Last but not least, we covered some of the ethical challenges that have arisen with some case studies that highlight the good intentions but incorrect applications of this technology and how much care and human quality control is needed when they are deployed. This is not only to avoid unintended consequences but also to stave off disaster when things go wrong.
Overall, it was a fun presentation, if I might say so myself, it’s always a challenge to talk about such an interesting topic with such a specialized and smart group of people. This was one virtual presentation for the books!