The financial industry was one of the first to appreciate the advantages of artificial intelligence (AI) and adopt it. The largest and most successful credit institutions have already developed official AI strategies, many have their own AI fintech departments. According to the research company Autonomous Next, by 2030, banks will be able to reduce costs by 22% using artificial intelligence technologies. The savings of financial institutions may reach $1 trillion. At the same time, a critical problem for large banks is the lack of qualified AI developers and data processing specialists. This may slow down the technologies development in the fintech sector. At the same time, artificial intelligence can help the financial industry to cut costs.
According to a Business Insider report, 80% of banks understand that AI technology can bring them massive benefits. This is shown in a survey of financial services professionals conducted by OpenText, a fintech software developer (Business Insider refers to the survey results). Many of them already actively use artificial intelligence, for example, in chat bots or to fight fraud. However, the larger the size of the bank's assets, the more willing it is to implement fintech solutions based on artificial intelligence.
Business Insider refers to data from UBS Evidence Lab, which interviewed 203 IT specialists in 175 banks in the U.S. According to the survey, 75% of banks with assets over $100 billion implement fintech strategies using AI, among banks with assets less than $100 billion there are only 46% of those who implemented an AI. In addition, 15% of banks with assets less than $100 billion don’t plan to implement fintech solutions based on AI at all, in the sample of banks with assets over $100 billion this share is only 5%.
Banks can save $447 billion by 2023 if they use AI. Savings of $416 billion out of this amount will go to the front office and middle office. In the front office, biometric technologies and personalized offers based on artificial intelligence will reduce the time of interaction between clients and employees, and in the middle office, the use of AI will reduce the risk of cooperation with dishonest clients who launder money. But enough of statistics, let’s take a look at the 4 most remarkable cases of AI implementation in fintech.
Visa's most advanced artificial intelligence tool is called Visa Advanced Authorization, which is a predictive analysis fintech application to spot fraudulent transactions in real-time. It is also worth noting that Visa Advanced Authorization has been the anti-fraud detection fintech solution of the company for over a decade.
Visa Advanced Authorization is capable of detecting data flows within specific transactions that correspond to the higher risk of fraud for the transaction. The company claims that the technology is able to detect up to 500 unique transaction risk attributes. The fintech software is expected to identify whether a transaction occurs in the store where the card owner is making purchases.
Visa also mentions the software's capability to recognize whether a transaction is performed by a high-value retailer such as electronics dealers or jewelry stores. The time frame and the sum of money spent are also taken into consideration, and all this data is being compared with all aspects of the client's spending structure. This may include any variations in the structure of consumer spending during vacation time.
The company claims that the fintech software uses identified risk attributes or fraud-related aspects of client behavior to assess transactions. In this system, number 1 represents the lowest risk and 99 represents the highest risk. If the transaction result exceeds a certain threshold, the system will reject the payment to prevent fraud.
Visa states that it is possible for Visa Advanced Authorization to compare transactions with client records for two years. This allows them to facilitate the recognition of rarer, but not fraudulent transactions and spending habits, as well as detect legitimate transactions that may have appeared risky in the past. Visa has made this possible in order to decrease the number of false positives produced by the fintech software.
Advanced Authorization technology is also backed by Visa Mobile Location Confirmation. This fintech feature lets customers access to a mobile device and give Visa access to their geo-location. This location data is then used to prevent a user's card from being rejected by Visa Advanced Authorization simply because they are on a trip and buying something extraordinary or from an unfamiliar spot.
Additionally, to the basic data points within a transaction, such as the product bought and its price, the fintech solution also holds authentication data such as biometric data or Europay Mastercard Visa (EMV) chip data. Biometric verification data includes fingerprints as well as selfies for face recognition.
Thus, a machine learning model can compare biometric data with biometric data in a file to securely authenticate customer credentials. The model can also use the data to detect whether the face is an inanimate copy or an image of the customer's face.
After the machine learning model has been taught to recognize data points that together are correlated with fraud, it can start comparing consecutive transactions to identify patterns and spending habits. Here, the fintech software can start highlighting risk attributes or transaction details that are associated with the risk of fraud, such as transactions suspiciously far removed from the geolocation of the last customer transaction.
In 2019, Visa introduced the results of its analysis, according to which Visa's Artificial Intelligence (AI) based system, Visa Advanced Authorization (VAA), during the year, has helped issuers prevent fraudulent transactions worth around $25 billion.
In 2018, Visa has handled more than 127 billion transactions between retailers and financial institutions on VisaNet and used AI to analyze all transaction data, with each transaction taking about one millisecond. Financial institutions could approve legitimate transactions by quickly detecting and preventing fraudulent behavior.
According to Visa, for 2019 the share of fraudulent operations on a global level in the Visa system reached a historical minimum and is less than 0.1%. This result was achieved through Visa's approach to business building - investing in human resources and technologies such as AI, providing cardholders and banks with powerful tools, resources and controls to manage risk, and establishing management processes to help businesses and regulators maintain flexibility in decision making.
Bank of America and Erica, their virtual assistant
Bank of America was among the first fintech companies to offer its clients mobile banking 10 years earlier. In 2016, they presented Erica, a digital assistant who was positioned as the most innovative payment and fintech services tool in the world.
By incorporating AI into its mobile fintech solution, Bank of America seeks to reduce the burden of routine transactions to free up its customer service centers to cope faster with more challenging cases, thereby considerably improving overall client service. Bank of America Virtual Assistant Erica is a chat-bot mobile fintech application that can help clients manage their bank accounts and monitor their spending habits. The Bank made the chatbot accessible to all customers of consumer banking after an internal pilot project in 2017 and a wider trial run in Rhode Island several months later.
The Bank of America claims that Erica has the following features:
- Notification to customers that their spending habits may result in zero balance within one week. This is most likely done by using predictive analytics to analyze average monthly expenses and the client's purchase history.
- Alerting customers about changes in their credit rating, which can supposedly be determined using the FICO points system in the fintech app.
- Informing customers of recurring or overdue payments.
- Seek account information and past transactions on request. This is most likely done using natural language text processing software (NLP).
- Tagging payments when they are more expensive than expected. This is most likely due to anomaly detection, which is a type of artificial intelligence fintech software for identifying abnormalities in the data stream in real-time.
- Lock and unlock credit and debit cards according to the customer's wishes.
As of May 2019, customers had more than 400,000 unique methods to ask financial questions to Erica and this number continues to increase.
The Bank keeps developing the AI fintech solution and has introduced a number of new features to Erica, such as proactive notification when funds are being returned to the seller.
David Tyree, Head of Advanced Solutions and Digital Banking at the Bank of America, said the combination of technology and "a highly personalized approach" provides "a more intuitive and efficient banking experience for our customers across all channels.
In the third quarter of 2019, digital clients entered their BoA accounts 2 billion times and used digital assets to make 138 million account payments, the bank said.
Barclays is a United Kingdom bank, which ranked 20th among the 100 major banks of S&P Global. Like other top banks, Barclays has started using AI for a variety of purposes. The bank cooperates with AI providers to a greater extent than with its own development of AI fintech applications, which is consistent with the general trend of AI implementation in financial services.
In July 2019, Barclays has declared a partnership with AI Simudyne, a startup focused on what the company calls "agent-based modeling". This can be defined as modeling various banking situations using transactional data to build simulated clients in a bank's ecosystem. This allows Barclays to simulate the banking and credit markets for detailed predictions.
Barclays states that the adoption of Simudyne fintech software has helped them to deliver more reliable offers to their customers, such as new financial services. This is likely because the software allows the client to gain a full insight into the opportunities and risks associated with lending to that client or making certain investments.
Mortgage analysis: simulating the relationship between clients and lenders, simulating the housing and lending markets in combination with client behavior.
Stress tests: Users can scale their simulations to represent most of the market, or reduce them to show detailed information about specific situations. Users can check how accurate their simulations are by continuously scaling them up until the system can fill in the results.
Risk Identification: simulating situations to reveal new investment risks and better understand how investments can lead to more risky situations, such as low liquidity.
Simudyne's agent-based modeling fintech software works on the basis of predictive analytics, an AI approach to predicting based on customer history data. For instance, a bank may select what happens if it chooses to lend to three clients with different credit scores. When running this simulation, they can choose how much layout data to attach to each simulated client.
This dummy data may include details of previous investments and loan applications. The software can then presume how likely it is that each client will fully repay their loan.
A bank may also generate a simulation of its applicants' relationship to the investment with one another, as well as with other banks. Clients are organized by their interactions and relationships with banks, credit unions, and other organizations such as the IRS. The ML algorithm of fintech solution uses these factors to assess the probability of a return on investment or loan repayment.
Barclays also states it runs the Simudyne fintech solution through the cloud, enabling them to create simulations that span an even broader set of data. This is because Barclays can now access all of its corporate data from any data lab. This means that they can simulate big changes in banking trends and start preparing for changes that are likely to occur early.
For example, Barclays can see an influx of underserved loan applications in a short period of time. Then risk managers in the bank can run agency simulations, trying to decide which of the applicants to grant a loan to. This may indicate that almost half of the applicants are in debt to other organizations and therefore are less likely to fully repay their loan to Barclays.
Morgan Stanley is a US-based financial institution primarily known for its financial advisory services. Morgan Stanley's asset management fintech application, launched in 2018, is called Next Best Action. Jeff McMillan, director of analytics and data at Morgan Stanley, said the fintech application enables wealth managers to do a more in-depth analysis of their customers, presumably because they can access more data about their preferences and market dynamics.
MacMillan also emphasized how AI has increased the extent to which wealth managers can reach out to their customers. Instead of sending emails to their clients one by one when the market is in sharp decline, they can also send messages to hundreds of their clients simultaneously.
Next Best Action, is part of the bigger wealth management platform Morgan Stanley, WealthDesk fintech solution, based on Aladdin from BlackRock. The company has introduced WealthDesk as a tool to bring wealth managers' workflows under one digital environment, allowing them to have full access to all the tools they need on one fintech platform rather than on several.
WealthDesk allows Morgan Stanley's wealth managers to access information about customer accounts that are also held with other financial organizations. This could give them a more integrated view of their customers' assets, allowing them to make better investment decisions for them.
By the end of 2018, Morgan Stanley announced that between 10% and 15% of their more than 15,000 financial advisors were utilizing WealthDesk. This fintech solution is now a crucial part of Morgan Stanley's growth strategy for its banking operations, which generates half of its earnings.
As you can see, now artificial intelligence technologies have become truly available for business and this opportunity can and should be used. Who knows, perhaps, if you do it quick enough, the next article on this topic will be about your case. But before you do, read our articles on AI strategy development and the most common mistakes in AI implementation to be prepared properly.