In business-critical fields, Artificial Intelligence can if applied naively or carelessly do harm, Professor of Finance Norman Schuerhoff says in an interview with finews.asia. The Revolut scandal is a good example.

Professor Schuerhoff, how – in our daily life – are Machine Learning and Artificial Intelligence (ML/AI) already operational?

My online shopping sites have started to tell me what its algorithms believe I like, need, and want to buy. Their algorithms also raise the price just before I am ready to buy – I wonder why. An app on my phone tells me which road to take to get from A to B and how long it will take me. Soon, my fridge will refill itself.

What are the fields of practice in the financial industry so far?

ML/AI can and already gets applied in many areas. ML/AI can be used to do the same cheaper, do more of the same, do the same better, and do new things. There are five types of use cases in banking and finance.

  • Intelligent automation helps to improve mid- and back-office efficiency. Applications include early warning systems, compliance, regulatory reporting, and trade processing.
  • The second pillar I call business intelligence and cognitive enhancements. This includes ML/AI applications to better understand client needs and improve client interactions.
  • The third pillar is knowledge discovery. It includes investment research, corporate statements, and legal discovery.
  • New growth are ML/AI applications for idea generation, smart alpha, and risk containment. Quantamental investment and risk management, trading recommendations and client advisory fall under this pillar.
  • Finally, new ML/AI enabled and based business models are still the sound of the future.

What are the ML/AI game changers in today's financial industry?

There exist external and internal game changers. The external ones are regulation and margin pressure in a quickly evolving post-financial crisis environment. In the current low interest rate environment, banks face tremendous margin pressure because their classical business models are not as profitable as before. New entrants, especially from the ML/AI domain, put additional competitive pressures.

«Technology in banking is still specialized, legacy, siloed, and regulated»

In the medium-term, the financial sector will better understand and explore the value from the drove of data it is sitting on. Banks are essentially data companies with banking licenses. Banks need to better internalize the idea that their value chain is built on data. ML/AI can then become the enabler of an ecosystem that serves client needs with simple, secure and enriching interaction.

Banks are not yet, or at least not enough technology companies. Technology in banking is still specialized, legacy, siloed, and regulated. The bank of the future will look different from today’s.

Are there cultural differences between American and other banks when applying ML/AI?

U.S. banks are clearly much further advanced in some areas of ML/AI applications than other banks, especially when it comes to adopting ML/AI in core competencies. Some U.S. banks have thrown hundreds of million dollars at developing ML/AI applications in various areas.

«In business-critical applications, ML/AI can if applied naively or carelessly do harm»

Yet, banks have good reasons to be cautious. Especially in business-critical applications, ML/AI can if applied naively or carelessly do harm. The Revolut scandal of 2018 is a good example.

Where do you see further room for ML/AI developments?

Deep learning and neural networks work well in Google-style applications such as face recognition. It is less clear that these techniques create long-term value in a financial market context. Financial markets are amazingly complex with millions of participants with different preferences, know-how, and sophistication. The stochastic nature of the interaction in financial markets defies the laws of physics.

In the long-run, the financial sector may develop its own tools and techniques specially geared towards understanding financial markets. Finance needs its own methods. But we are not there yet.

What are the most recent findings in your research?

In one of my recent research projects, I show how ML/AI is revolutionizing the trading of financial securities. Algorithmic and high-frequency trading have long been an integral part of liquid markets like stock exchanges.

«In one of my research projects, we explore if a machine can replace a Chief Financial Officer»

However, the trading of more illiquid and opaque financial securities, such as corporate bonds, has until recently been very different and insulated from technological innovations. With the advent of electronic trading in over-the-counter (OTC) markets, ML/AI become crucial technologies for pricing and trading.

In another of my research projects, we explore if a machine can replace a Chief Financial Officer (CFO). The role of a CFO is to advise the CEO on all financial matters and to choose the optimal capital structure and financing policies. Like a computer can learn from human drivers how to best navigate a car, a computer may learn from CFOs’ behavior how to best run the financing of a firm.

Norman Schuerhoff is a Professor of Finance at the University of Lausanne. His work has been published in the top academic journals in finance and he has won several prestigious publication awards. He is a six-time winner of the CFA Institute Research Challenge in Switzerland and was World Champion for 2018. His main research interests lie in financial intermediation, corporation finance, corporate governance, market microstructure, and asset pricing.

In ongoing research, Professor Schuerhoff and coauthors study the municipal bond market—the largest and most important capital market for state and municipal finance in the U.S. With «green» bonds issued at a premium due to strong investor demands, the municipal bond market has led the development of responsible investing.