The artificial intelligence (AI) landscape had evolved significantly from 1950 when Alan Turing first posed the question of whether machines can think. Today, AI technology has reached a stage where it can reshape economies and improve human lives, promising a better quality of life, productivity, efficiency, cheaper and more accurate predictions, recommendations, decision making, and cost-saving.
AI applications are experiencing a rapid uptake in several sectors where it is now possible for companies to detect patterns in large datasets and improve problem-solving and decision making.
In transport, for instance, autonomous vehicles equipped with virtual driver systems, high definition maps and optimized traffic routes, promise cost, safety, and environmental benefits. In healthcare, AI systems help diagnose and prevent disease and outbreaks early on, discover treatments and drugs, propose tailored interventions, and power self-monitoring tools.
Our focus in this article is financial services that leverage AI to gain many significant benefits, such as fraud detection, credit assessment, cost savings, trade automation, and compliance.
Large fintech companies such as JPMorgan, Citibank, State Farm, and Liberty Mutual and start-ups such as Zest Finance, Insurify, WeCash, CreditVidya, and Aire are some of the forerunners in rapidly deploying AI. They combine different AI and machine learning (ML) techniques to improve customer experience, identify smart investment opportunities, and possibly grant more credit with better conditions to the customers.
This article presents an overview of AI applications in the financial sector, covering credit scoring, lending, algorithmic trading, cost reduction, customer experience, and compliance.
1. Credit scoring
A credit scoring system is a statistical analysis that makes credit-worthiness assessment and recommendations to grant loans to people. The credit-scoring algorithms use a mathematical model to formulate recommendations based on machine-based inputs (historical data on people’s profiles and on whether they repaid loans) and human-based inputs (a set of rules). In other words, the system evaluates whether a borrower may default on her/his debt obligations.
In traditional credit-scoring, analysts make mere hypotheses based on minimal, available information about the customers or customer segments. However, in recent years, AI algorithms have enabled the analysis of vast quantities of data collected from credit reports, providing fine-grained analysis of the most relevant factors and their relationships. Neural network and deep learning credit-scoring techniques, analyzing customer segments, and large datasets in new ways can improve the accuracy of predictions by up to 15%.
FinTech lending platforms allow consumers within seconds to shop for, apply, and obtain loans online. To do that, they leverage various data sources (such as insurance claims, social media activities, online shopping information from marketplaces, shipping data from postal services and browsing patterns, in addition to traditional credit data and payment history), to predict the probability of default before making lending decisions.
Of course, this was made possible by AI, which can combine the power of both traditional FICO sources, as well as alternative data. Experts estimate that the FICO score often has an accuracy of 68.3%, while an algorithm based on alternative data has a 69.6% accuracy rate. Combining both types of data improves the accuracy rate of up to 73.6% in loan success.
Ant Financial, formerly known as Alipay, uses algorithms to process the huge amount of transaction data generated by small businesses. This allows Ant to lend more than USD 13.4 billion to nearly 3 million small businesses. The algorithms analyze transaction data automatically on all borrowers and all their behavioral data in real-time. It can process loans as low as several hundred dollars in a few minutes. Every action taken on the platform – such as transactions and seller-buyer communication – affects a business’ credit score. Simultaneously, the algorithms evolve over time, improving the quality of decision making with each iteration.
3. Cost reduction
The use of AI benefits financial institutions in terms of cost reduction in their operations. The deployment of AI in the front office (client interaction), middle office (support for front office), and back-office (settlements, human resources, compliance) can save financial institutions an estimated 1 trillion dollars by 2030 in the US, impacting 2.5 million financial services employees.
Advanced AI tools can increasingly reduce the need for human intervention. Financial data and account actions are integrated with AI-powered software agents within the front office. These agents can converse with clients using advanced language processing in platforms such as Facebook Messenger or Slack.
AI can facilitate risk management and regulatory processes in the middle office, while in the back-office, AI can broaden data sources to assess credit risk, take insurance underwriting risk, and assess claims damage.
4. Legal compliance
The financial institutions are well known for the high cost of meeting the standards and reporting requirements for regulatory matters. New regulation in the United States and the European Union over the last decade has further increased the cost of regulatory compliance for banks.
In recent years, banks have spent an estimated USD 70 billion each year on regulatory compliance and governance software. This expenditure reflects the costs of having bank counsel, paralegals, and other officials verify compliance with the transaction. Costs for these activities were expected to increase to almost USD 120 billion by 2020.
Deploying AI technologies, particularly language processing, can decrease banks’ compliance costs by approximately 30%. It will reduce the time needed to verify each transaction considerably. AI may assist in interpreting regulatory documents and in codifying compliance rules.
The Coin program created by JPMorgan Chase, for instance, reviews documents based on business rules and validation of data. The program can examine documents in seconds, which would take 360 000 hours of work for a human worker to review.
5. Fraud detection
Fraud detection is another big application of AI for financial companies. Banks always kept track of account activities and patterns. However, advances in machine learning are now starting to enable near real-time monitoring, allowing the identification of anomalies immediately, which triggers a review. The ability of AI to continuously analyze new behavior patterns and automatically self-adjust is uniquely essential for fraud detection because patterns evolve rapidly.
In 2016, the Credit Suisse Group AG launched an AI joint venture with Silicon Valley surveillance and security firm Palantir Technologies. To help banks detect unauthorized trading, they developed a solution that aims to catch employees with unethical behaviors before harming the bank. Fraud detection based on biometric security systems is also gaining traction in telecom, greatly enhancing the bank’s security measures.
6. Algorithmic trading
Algorithmic trading uses computer algorithms to automatically decide on trades, submit orders, and manage those orders after submission. The popularity of algorithmic trading has grown dramatically over the years. Today, it accounts for the majority of trades put through exchanges globally. In 2017, JPMorgan estimated that just 10% of the trading volume in stocks was “regular stock picking.” Increased computing capabilities enable “high-frequency trading,” whereby millions of orders are transmitted every day, and many markets are scanned simultaneously. Besides, while most human brokers use the same predictors, the use of AI allows more factors to consider.