How AI Is Taking Over Investment Banking
Did you know in 2000, there were 600 traders at the U.S. cash equities trading desk at Goldman Sach’s NY headquarters employed 600 traders, buying and selling stock for the investment bank’s large clients.(Source: R. Martin Chavez, Deputy Chief Financial Officer of Goldman Sachs ) Today there are just two equity traders left. The reason for this drop? The rise of technology. So much so that the 600 traders were replaced by 200 computer engineers. Goldman Sachs isn't the only firm to benefit from implementing technology, recently JP Morgan hired a “Global Head of Machine Learning” from Microsoft, who specialises in Natural Language Processing.
The reason AI and automation are spreading so rapidly throughout the investment banking sector is it can be applied numerous ways and every way it is applied it has the potential for massive cost savings.
One obvious way in which Investment Banks can use AI is predictive analytics.Complex neural networks can spot trends and patterns in data that even experienced traders can struggle to see. A large criticism of predictions of future prices revolves around the “random walk hypothesis” that revolves around the idea that asset prices move randomly and it is impossible to consistently outperform market averages (More on that in “A Random Walk Down Wall Street”, by Burton Gordon Malkiel).However even if arguments like this one are true, the fact that speculators may believe there is a pattern and act on it may self fulfil the prophecy, theories of patterns in prices like “Elliott wave theory” have become apparent.There is also the study that a monkey choosing randomly outperforms experienced investors but hey may of just got lucky we'll never know.
One of the first large applications of AI in predictive analytics was in 2017 where the Dutch bank ING launched a new AI-based platform called “Katana” aimed at bond traders.The platformed served as any other trading platform showing price and relative information however it also comes with an AI predictive tool that uses data and learns from it too make predictions.
Another recent player in the AI predictive analysis scene is the platform called “The Kore” by a small London based company known as Kortical. The platform works by letting users input data and its corresponding parameters.The software then proposes the optimal ML model to use on the data for the most accurate predictions.
Its success are listed on its website, with the platform being involved in everything from healthcare to retail however in terms of banking two of its successes include:
Predicting 83% of previously undetected bad debt for a bank with over 20 million clients using existing credit score information as well as showing additional insights about each customer
An investment bank needed to calculate the predicted ROI on around 5 million trades a night which was being done through 80,000 computers. The bank was using human analysts to help classify the trades in ways that might help make this process faster but was constrained by the sheer number of trades, Korticals platform was then used to create a predictive ML model that could predict how long it might take to calculate the ROI for trades at the firm using past data. It worked by finding patterns in the data such as what factors caused slower times to predict ROI, these patterns were then used to improve time taken to predict ROIs by 30%.
We can expect to continue to see more and more investment bank and hedge fund type firms to start to shift their focus onto how they can apply AI to maximise their profits. The rise in algorithmic trading in the recent decade, due to a lot of firms capitalizing on the High frequency trading (HFT) strategy in which stocks are bought and then sold in fractions of a second. This strategy is a form of arbitrage in which the HFT algorithm spots a price discrepancy and proceeds to quickly profit of it, means that it has never been easier to implement AI systems into existing trading algorithms as they already in use.
Another application for AI in investment banks is to aid in the automation of data collection.AI is perfect for this task as over time they can learn very accurately what type of data is best for whatever it is needed for.
An example of this is the American firm Sigmoidal, which offers its services in improving firms systems by implementing AI to its clients.It does this by guiding as well as providing software to firms that are interested in applying AI.
The company reckons the system can be implemented to help discover investment opportunities. In summary the use case says the firm would help the automation of web scraping of relevant information.It would then use AI NLP software to filter out irrelevant data and then using a technique known as “named entity extraction” which identifies important information to the investor in the text, based on given criteria.Using this information it groups it in terms of relevance and importance and sees any correlations and returns the information in the form of a report using Natural Language Generation (NLG), a artificial intelligence based technology that can generate human-sounding reports and summaries. This is but one possible application of AI in data collection.
Overall we can expect to see firms of all sizes continue to innovate and find new applications for AI in investment banking, which may lead us into a new age of investment banking primarily dominated by machines. However for now we can expect to see AI being integrated into more and more platforms to aid current investors with ever improving predictions.