AI and the Financial System

13 minutes

How AI is changing the way we lend and borrow money.

What is the financial system?

Picture the scene: a small town is home to just a few houses and families, one local business (a restaurant) and a town committee, or mayor. How do they interact with each other on a daily basis? The mayor provides public services and in turn collects taxes from the restaurant and households, while the restaurant serves food and drink to customers.

But what would happen if something changed this status quo? What if the restaurant owner wanted to open a second branch in town? Or the mayor decided to build a new bridge? Both would need funding, and the usual interactions between the three would not be able to fulfil these financing demands.

This is where financial systems come into play.

The restaurant owner could apply for a loan from a local bank, or the mayor could opt to approach a financial market and borrow the funds from the town’s residents, giving them municipal bonds in exchange. In fact, both the local banks and the financial marketplace can be lenders in this scenario.

This simple process can also work on a much bigger scale. Take a country, for example – where you still have the same three main groups: households, businesses, and the government. Here the system is made up of financial institutions (such as a local bank or insurance company) and financial markets (such as a market for issuing and selling bonds and stocks).

As the diagram below reveals, the primary aim is to create a channel between borrowers and lenders.

But the financial system does not just focus on lending and borrowing – it offers a variety of other financial services as well.

With a local bank you can transfer money, while investors use the financial markets to try and make a profit by investing in currency pairs, commodities such as gold and oil, and more recently, cryptocurrencies. Unfortunately, these financial services can be exploited and are vulnerable to corruption.

To avoid this, financial regulators are crucial. These neutral organisations supervise the operation of the financial system to protect both businesses and their customers. However, even with regulators in place, the financial system can be very complex – which is why AI has become a crucial part of the business world in recent years.

But as technology grows and becomes more advanced, how will it continue to impact the world of finance moving forward?

AI in Financial Institutions

Financial institutions are often referred to as financial intermediaries, because they are the middleman between lenders and borrowers. A local bank is the best example of a financial intermediary, while insurance companies, pension funds, and mutual funds also fall into this category.

The main role of a bank is to collect deposits from savers and then offer those funds as loan products either to members of the public – who might be in the market to buy a new house, for example – or to businesses wishing to expand product lines or open new branches. But banks also provide other financial services to keep their current customers and attract more, such as consultations on different loan packages or investment opportunities, as well as transaction and credit card services.

However, these additional services can distract large banks (who often have millions of customers) from their primary role. They can lead to huge queues when you are trying to contact customer support, or result in delays when desperately trying to report a stolen credit card before your balance is maxed out.

Utilizing Smart Chatbots

Banks are therefore turning to AI to help lighten the load and speed up customer service. Some banks are now developing smart chatbots to automate simple client requests, which use a technology called natural language processing (NLP).

Chatbots are removing the clutter and saving time – instead of the frustration of too many menu options and “on hold” music when calling a bank, these chatbots are delivering information to the public in a matter of seconds. And if they don’t know the answer, they are connecting customers directly to the right customer service agent in the right department.

The sky really is the limit when it comes to AI and its daily interactions with the public. But the big question is this: just how comfortable is the human race when it comes to handing over financial decisions to machines?

For example, would you be happy for a chatbot to calculate whether you can afford a car loan? What would be the difference between a human voice deciding your future, or a computer? Important factors like customer service, warmth, friendliness, accuracy, precision and response time would all need to be weighed up and considered.

The Reliability of AI

There are other merits to AI in the financial world, though. It can save you time when used via a smartphone, where it can reliably scan invoices and pay them, for example. The key word here is reliability – we’re asking a machine to authorise and execute a payment with a bank, and AI is already proving it can make these everyday tasks easy for us whilst also eliminating the possibility of human error and discrepancies.

For financial institutes such as banks, AI technology is simplifying the reading and archiving of payments, freeing employees from cumbersome and boring tasks.

Improving Security and Minimizing Fraud

AI is also helping to stamp out fraud and is making transactions with financial institutions safer and more secure. In the same way that a machine can read and understand reference numbers on invoices, or a smartphone can identify your face at the lock screen, ATM machines could soon be able to confirm your ID in the same way.

Currently, cashpoints can only engage in a simple conversation with the user by asking basic questions, such as requesting a PIN code. But banks are also now working on AI for ATMs that use voice and image recognition. It will be able to identify you from millions of other users, taking account security to a whole new level.

While recording your face at an ATM might seem intrusive, something very similar is actually already happening every time you withdraw money. ATMs are able to record users via direct or overhead cameras to provide potential evidence should a customer or the bank claim (or suspect) fraudulent transactions.

The Impact of Data

Banks are also using AI to improve their primary business by taking advantage of the data generated by customers. For example, they can program AI models to learn from a client’s past transactions and then adapt credit card limits according to that person’s spending behavior.

Banks can also adopt a big brother approach and monitor their clients’ activities on social media to better understand how much they can trust them. When you apply for a mortgage at your local bank, your credit rating will be assessed by standard criteria – such as proof of income, credit history, and employment verification – so that you can be charged what the bank deems as the right interest rate accordingly.

However, in some regions of the world where people may not have a credit history to apply for a mortgage, banks can use AI solutions to go beyond the conventional criteria. They do this by collecting information from social media feeds to better predict the risk a potential client may pose to the bank and answer one major question: are they reliable people who will not miss a repayment?

In these examples, AI is helping banks to better serve society by mitigating risk and making sure customers get the right kind of loans they need. But some big questions remain: what does it mean for data privacy? Can AI identify fake social media profiles? And should all of this be regulated?

AI in Financial Markets

Financial markets are physical or electronic platforms where lenders and borrowers can meet – and a large volume of trading activities happen here every day. While the primary role of these institutions is to act as a conduit between the two parties, they offer many other services as well.

Let’s revisit our example of a town mayor who needs money to improve infrastructure, such as building a bridge. If you buy a municipal bond from the mayor, you are not obliged to keep the bond and can sell it in the market at any time. You can also buy such bonds directly from the market.

But whatever way these trades are conducted, millions of data points are being generated in every single market. You can pretty much trade anything these days – from coffee and oil to Bitcoin and Tesla shares – but while these markets are often not connected, they do impact each other. For instance, trading activities in the oil market can influence the currency market, no matter where they are located.

All these factors together make this industry highly complex to analyze, even for an expert, and so professional market participants such as hedge funds, algorithmic traders, wealth managers, and investment bankers are all trying to leverage AI capabilities in two main ways.

Firstly, some professionals believe that the finer details of the markets make AI a natural choice to build trading strategies, optimize asset allocations, and predict price movements faster and smarter than ever before.

Indeed, AI can uncover complexities neglected by humans. For example, it is nearly impossible for the human brain to simultaneously analyze the dynamics of the oil and currency markets alongside the impact live political speeches in multiple countries and languages referencing these markets has. Also, given the time differences due to the geographical locations of these markets, it is simply impossible for an investor to continuously monitor and analyze the markets at all times of the day and still respond within milliseconds to miniscule trends or special events.

Secondly, AI can help investors exploit all types of information and predict future market movements. Traditionally, most data used by market professionals was numerical, such as prices and trading volume. However, there are other sources of valuable information such as newspapers, users’ activities on social media, or even blog posts that can influence market movements. This is often referred to as market sentiment.

AI solutions, such as NLP techniques, allow the experts to capture both the opinions of other experts, and the emotions of non-experts on social media, to evaluate if such information is linked to movements in market prices.

AI in Financial Regulation

It isn’t all plain sailing, however – if not used correctly, AI can negatively impact customers. For instance, data privacy is a concern when chatbots ask about a person’s financial history or want to change a password. Similarly, you might worry about a person’s savings when a bank gives it as a mortgage to another customer based on an AI’s decision.

Can we really trust the machine? Fortunately, there are processes in place to monitor this. Financial systems are constantly under the surveillance of an independent authority, otherwise known as a financial regulator.

In Switzerland, the Swiss Financial Market Supervisory Authority (FINMA) plays this role. The main task of a regulator is to monitor the system and ensure it’s running safely and securely. Financial institutions, and the traders who are active in the financial markets, are first and foremost trying to maximize their own profits. Therefore it is the regulator’s task to put a limit to the extent to which they take risk.

For example, banks may give loans to many risky borrowers, and it may happen that these clients will not be able to pay back those mortgages – but a regulator ensures the bank always has enough capital to cover such losses.

The regulator’s task is even more crucial when an innovation comes into play. For instance, the great financial crisis of 2007 and 2008 was partly caused by innovative instruments – such as mortgage-backed securities – which were used and trusted heavily by financial institutions and traders.

Similarly, technological innovations like AI could introduce even greater unforeseeable risks to society. As a result, regulatory bodies around the world are widening their scope to take into account all new technical advances, and assess their positive and negative implications.

This is being done in three areas:

Maintaining and Updating Current Regulations

Regulators now have two pivotal roles: to maintain their existing regulations, and to update them to incorporate rulings on the use of AI in financial systems. Accountability and risk are the key themes here.

For example, the current capital requirement framework for banks should be amended so that it covers potential losses from giving loans to borrowers based on their social media records. Risk management policies should also include protocols in case any AI tech makes a wrong decision, such as classing a risky borrower as a safe bet.

Introducing New Regulations

Regulators are having to get out their crystal balls and plan for the future. This means introducing new regulation covering risks that may not have happened before, but could happen in the forthcoming years. And with AI constantly changing and developing, these supervisory issues have to be updated regularly.

Banks need to implement new terms and conditions to protect customers’ data privacy when they use chatbots, or when they share their information – such as social media or spending behavior – with the bank.

While chatbots make the customer support process more efficient for financial institutions, a regulator needs to ensure the data privacy is in line with the bank’s overall strategy. What’s more, regulators are having to clarify who is responsible and liable for any mistake made by AI. Put simply, can a bank trust a machine’s intelligence? For example, a loan applicant could try and fool the AI by creating fake and glowingly positive updates on social media.

Taking Advantage of AI to Shape Regulations

Regulators are also utilising this new technology to enhance their frameworks and ensure they fulfil their anti-money laundering and counter-terrorism financing requirements. Banks are increasingly using AI-based detection techniques to predict suspect and fraudulent behavior – such as the taking over of a customer’s account, credit card, payment fraud, identity theft, and money laundering – among the millions of transactions carried out every day.

This is where the aforementioned crystal ball comes into place: regulators must make sure these systems accurately identify risk and predict crimes in the future.

Moreover, regulators can use AI techniques to detect and prevent insider trading activities in the markets. Insider trading is a big concern for the financial industry and happens when a trade is placed using information that is not available to the public.

For example, suppose that Company A plans to acquire Company B. The plan is confidential and not available to the public, but is leaked to someone who then places a trade to earn a profit. Such illegal trade harms the confidence of investors and deteriorates the sustainability of the markets. But while all trading activities are recorded, it is extremely difficult to detect illegal activities among the billions of transactions performed each day.

That’s where regulators can benefit from AI techniques that help analyze such data quickly, and detect suspect and fraudulent activities.