In January 2020, the Cambridge Centre for Alternative Finance (CCAF) released a study on the impact of artificial intelligence on the financial industry. One of the most comprehensive global surveys in the field, it included 151 respondents from 33 countries/regions, including existing financial institutions and fintech companies. The study yielded the following findings.
At least 77% of respondents believe that AI will be of high importance to their organizations in the coming years. Almost 64% of respondents intend to make money through customer acquisition, customer service, risk management, process automation and new products through AI. AI is now widely used in risk management, with an implementation rate of 56 percent among companies.
Traditionally, HFT firms and hedge funds have been the main AI implementers in finance, but more recently, FinTech firms, insurers, banks and regulators are catching up.
Some of the uses of AI in this industry include robo-advisors, backtesting, model validation, portfolio composition and optimization, stress testing, algorithmic trading and regulatory compliance. Let’s learn more about AI applications in finance.
Artificial intelligence and machine learning algorithms are gradually revolutionizing financial risk management. AI-driven solutions can provide insights into
Determining the amount of loans to customers.
Generate warning alerts to traders about position risk.
Enhance compliance and limit model risk.
To understand why respondents in the CCAF study cited risk management as a primary focus for implementing AI, consider the case of Baidu.
The most famous search engine in China is Baidu. (because Google is banned there). In 2016, Baidu sought the help of ZestFinance-ZestFinance is a U.S.-based fintech company focused on AI products. Baidu’s goal was to offer small loan deals to retail customers who purchased products from its platform.
However, the lending situation in China is in stark contrast to the Western market – the former is quite risky as more than 80% of people do not have any credit status or credit rating. Therefore, there are no existing methods for determining the reliability of borrowers.
ZestFinance solved this problem by analyzing Baidu’s large customer data set, especially search and purchase history. In this way, they use AI to assist Baidu in deciding whether to lend to a customer. By 2017, a survey found that , Baidu microloans had grown by 150% without any significant credit losses.
Since ZestFinance processes financial data through proprietary technology, the full details of their AI solution are unknown. However, it is known that their process combines two machine learning algorithms: decision trees and clustering.
For example, if a customer’s search history indicates extensive visits to gambling sites, they are grouped into clusters associated with higher risk. On the other hand, if borrowers are responsible for online spending, they are grouped into lower-risk borrowers. With automation, Baidu’s finance staff can easily review these applications and approve loans based on people’s risk.
Investment firms have long used computers for trading. A large number of hedge funds rely on data scientists to build statistical models. However, the approach has significant limitations – it uses only historical data, which is mostly static and dependent on human intervention. As a result, these calculations are difficult as the market changes in any way.
Fortunately, modern AI models have made great strides through algorithmic trading. These models are different because they not only analyze large amounts of data, but they are truly autonomous – they learn and improve over time to the point where they can compete with humans. This “intelligence” comes from sophisticated machine learning techniques such as evolutionary computation (based on genetics) and Bayesian networks.
Artificial intelligence tools collect large amounts of data from global sources, “learn” from them, and make predictions accordingly. This data consumption is exhaustive. It extracts information from financial transactions, news reports, books, social media platforms (e.g., tweets) and even TV shows (e.g., Saturday Night Live).
It is important to understand how AI has made great strides in this area. Unlike traditional technology intervention techniques (which allow humans to determine financial strategies), AI can now guide the game.
One example of these AI-powered trading systems is Aidiya , an AI hedge fund based in Hong Kong that allows users to trade all stocks via AI. It is worth noting that startups are not the only ones interested in AI trading technology. Previously, prominent names , such as Goldman Sachs, Wells Fargo, Citigroup, Morgan Stanley, Merrill Lynch, Bank of America and JP Morgan Chase, took an active interest in the Kensho Project – an AI trading platform.
Another fast-growing application of artificial intelligence in finance is fraud detection, which is understandable given the huge amounts of money involved. The cybercrime industry has stolen approximately $600 billion from businesses around the world, accounting for 0.8% of global GDP. Using criminals to exploit modern technology, cybercriminals have become more savvy and clever. The integrated fraud detection and prevention market is expected to grow by more than $40 billion by 2022, according to Statista.
So, how can artificial intelligence help? To master this skill, modern machine learning cybercriminals can use a combination of supervised and independent techniques to build models with predictive accuracy and capability.
Supervised learning uses annotated data (which humans evaluate and identify as fraudulent activity) and learns complex patterns from corporate datasets. Meanwhile, the unsupervised learning process will process those datasets that were not previously identified and infer the data structure on its own. Other fraud detection techniques include regression and classification. They can analyze data and determine whether transactions are fraudulent.
The standard supervised algorithms used to solve these problems include the following.
– Decision trees help introduce a set of rules that learn normal customer behavior while being trained by fraudulent instances so that they can identify anomalies and warn about permissions.
-Human brain-based neural networks can learn and adapt to customer behavior to detect real-time fraud.
Examples of unsupervised learning algorithms include.
– K-means clustering splits a data set into a group of similar data points (called clusters) for anomaly detection.
– A local outlier factor determines the local density of data points, thereby identifying regions where similar densities exist. Data scientists can use the concept of locality to flag ends with unusually low densities, called outliers. This application can be used to detect fraudulent transactions.
Regulatory compliance is an important function of finance, especially in an economic crisis such as the current one. Regulatory compliance is linked to enterprise risk management and deals with risk functions such as operational, market and credit risk.
RegTech is an advanced function in FinTech that focuses on compliance. Here, artificial intelligence is advantageous in the continuous monitoring of company activities. In this way, it can provide valuable real-time insights and prevent regulatory violations from the outset. In addition, this form of monitoring allows firms to free up regulatory capital and use automation to reduce excessive compliance costs, with large financial firms spending $70 billion a year on compliance.
A well-known company in this space is IBM. Not long ago, IBM acquired Promontory, a 600-employee RegTech startup. The acquisition has enabled IBM to drive numerous AI-based solutions to manage financial compliance. For example, IBM is using its proprietary AI tool Watson AI in conjunction with Promontory’s RegTech expertise to deploy real-time voice conversation analytics to ensure compliance. Part of this includes translating speech-based conversations into text and then using natural language processing for text classification. The consequence of this process is the formation of categories that detect potential violations.
Other AI applications include automated reading and interpretation of regulatory documents, particularly for determining meaning. London-based Waymark is already offering this service to financial firms.
While there are many other applications of AI in the financial sector, there is another side to the story. The industry needs to correct practical issues to enhance AI implementation.
One of the biggest concerns remains the availability of suitable data. While R and Python can read any form of data from Excel spreadsheets to SQL / NoSQL datasets, AI-driven solutions run slower than an organization’s ability to accurately organize its internal data. Often, data is stored in different repositories across departments, often in separate systems where regulatory and internal political dilemmas limit information sharing.
Likewise, another dilemma is the lack of skilled employees who not only have expert knowledge in AI, machine learning and data science, but also have extensive experience in building and implementing AI-centric solutions in the financial industry.