Everything that is hidden will be found out, and every secret will be known. When market manipulations, regulatory violations or outright fraud come to light the fines for banks can go into the hundreds of millions of dollars. In order to protect themselves from new and increasingly clever methods, banks are using newer and newer technologies. We are already familiar with audits, transaction monitoring systems, instant messaging and email monitoring systems (written language).
Recently, two more methods have been added, the first is Behavioral Analytics: monitoring and analyzing the behavior of traders to fish out anomalies, with one of the anomalies indicating that “something is wrong” may be … the speed of entering trades.
The other is real-time analysis of traders’ voice calls. Still a large part of inter-bank communication is done by telephone. An AI system based on Machine Learning is hosted in the cloud and listens to conversations about trades.
The possibilities are much greater, I don’t think it will be applied anytime soon (privacy issues), but for example, there are already systems for analyzing facial expressions to pick out likely emotions. There are also systems for analyzing not only the content, but also the timbre of the voice, the algorithm is able to pick out emotions that accompany speech and, for example, determine with high probability whether someone is lying (solutions are known and have been used for at least a dozen years, the pioneer was the Israeli service).
All in all, a tighter and tighter system for tracking and catching anomalies is emerging, a side effect of the progressive application of AI. AI systems are becoming more and more “intelligent” and are able to catch more and more complex behaviors. And not just the behaviors themselves – better and better user profiles are being created based on deeper insights into suspicious behavior.
A side effect of voice order analysis is also a slight delay in recording the actual trading volume in the markets, until now this was impossible as voice transactions were summarized only at the end of the day. There is one company that aggregates this type of data, its representatives say they ultimately want to minimize the delay in volume presentation to a few minutes.
More at the magazine, just subscribe!