AI Technology in Finance Industry

Artificial Intelligence (AI) is driving business disruptions in almost every corner of the finance industry. Banks have been employing AI to restructure their anti-money laundering and fraud detection endeavor for several years. Insurance companies have been utilizing AI to enhance products, strengthen claims process, predict and preclude fraud, as well as boost customer satisfaction. Investment firms, in turn, have implemented AI, and especially machine learning (ML), to optimize portfolios, execute trades, and provide personalized service based on individual client needs. In PwC 2019 AI survey of US Executives, executives from the finance industry said they expect AI endeavors to result in an average improvement of 50% on revenue and profits, 48% on customer experiences, and 42% on innovative new products.

AI encompasses a large collection of technologies that are considerably different in capability. It can be an autonomous intelligence which augments human intelligence substantially . Some of the relatively simple AI tools have been adopted widely. For example, intelligent process automation (IPA) can be found in many banks. IPA handles routine tasks and repetitive processes that require little problem-solving and judgement, such as data-entry or basic calculations. Thus, employees will have more time for responsibilities that are more complex and rewarding.

Machine Learning (ML), a more advanced AI technology, is now a common tool for many hedge funds. Especially in hedge funds with quantitative strategies, ML is regarded as the core of their future competitiveness. “Quant” funds rely on algorithmic or systematic programs to identity new themes and trading signals. The models are running on a continuous basis, requiring a large amount of reprogramming by human quantitative analysts. This is where the AI models bring in improvement of efficiency. These self-sufficient ML models, do not require human programming, are able to adapt to evolving market conditions by themselves. This feature has a consequential time benefit since human reprogramming does not match the speed of automatic adaption accomplished by these ML models.

According to an article by Boris Friedman, published on Preqin Insights, Artificial Intelligence funds are outperforming the overall hedge fund benchmark. During the past year, performance tracks of 152 AI hedge funds scored an average three-year cumulative return of +23.87%, while the Preqin All-strategies Hedge Fund Benchmark has an equivalent record at +26.96%. Though a 3% difference in return may not validate an absolute advantage, taking risk factors into account presents a brighter future. Risk-adjusted parameters of AI funds, three-year volatility and the Sharpe ratio data, is more favorable than the Preqin All-Strategies Hedge Fund Benchmark – Al funds have a 3.20% volatility and 2.96 Sharpe ratio while all hedge funds have a 3.87% volatility and a 1.40 Sharpe Ratio.

A specific application of machine learning will be a market adaptation system which focuses on market participant sentiment for a particular investment target during a specific period of time. In this case, the system will search and collect a wide range of information from the internet, then measure and transform public confidence into a proprietary indicator, on which traders would base their decisions.

While an ML model is impressive enough, AI has more potential for excavation, which may realize a full autonomy system. Artificial Neural Networks (ANN), inspired from biological nervous systems and brain structure, are computational modeling tools that have recently emerged and found huge markets in many disciplines for modeling and solving complex problems in industry. It could be regarded as an information processing system which has enormous learning and generalization capability. Different from machine learning in general, as mentioned above, ANN is commonly referred to as deep learning, which is a subset of machine learning, and has networks capable of unsupervised learning from unstructured or unlabeled data. This means that every time an AI model gains access to a sufficiently large data pool, it can constructively form connections regardless of how irrelevant or poorly-organized the database might seem. When applied to the finance industry, ANN may discover linkages and predictors that are impossible for human brains to identify. Such a feature is extremely pertinent and powerful in an unstable market.

Nevertheless, before the finance industry pushes AI technology to the spotlight, it must conquer numerous challenges, such as security, privacy, bias, and regulatory issues. The greatest challenge may be gaining customer trust. In the same PwC 2019 AI Survey, ensuring the trustworthiness of AI systems is the top concern of 40% of the financial industry executives. 64% of them intend to create AI models that are more transparent, explainable, and provable; 60% of them plan to enhance AI security with validation, monitoring and verification; 54% of them expect to strengthen relevant governance. Another challenge ahead is the shortage of workers who have the tech skills and expertise needed. Almost a third of the financial executives in the PwC AI survey are worried that they may not meet the demand for AI skills over the next five years. Around 60% of them said that they are planning to implement continual learning programs which incorporates AI-knowledge to upskill their employees.

Works Cited

1. https://www.preqin.com/insights/blogs/the-rise-of-the-machines-ai-funds-are-outperforming-the-hedge-fund-benchmark/26411

2. https://www.pwc.com/us/en/services/consulting/library/artificial-intelligence-predictions-2019.html

3. https://www.barrons.com/articles/how-artificial-intelligence-is-already-disrupting-financial-services-51558008001

4. Hsieh C-T (1993) Some potential applications of artificial neural systems in financial management. J Syst Manage 44(4):12–16

5. https://www.investopedia.com/terms/d/deep-learning.asp

6. Bahrammirzaee, Arash. “A Comparative Survey of Artificial Intelligence Applications in Finance: Artificial Neural Networks, Expert System and Hybrid Intelligent Systems.” Neural Computing and Applications, vol. 19, no. 8, 2010, pp. 1165–1195., doi:10.1007/s00521-010-0362-z.