FINANCIAL ORGANIZATIONS TODAY ARE LIKE A KID IN AN AI CANDY STORE
Artificial intelligence solutions have become mainstream in the financial sector. The range of available solutions is broad and varied, from risk management through fraud prevention, customer experience, resource management, and more. The real challenge now is how to choose the solution most suitable to varying needs and resources. We spoke with Pazit Katz Oz, head of products at Matrix DnA, as well as Idan Beer, head of Matrix’s AI Center of Excellence and a Deep Learning researcher, to find out more…
“The great global migration into digital realms has made it possible to accumulate very valuable information about clients and processes,” says Idan Beer. “With Big Data technologies to sort and match data, and with artificial intelligence to forecast, suggest conclusions, and extract patterns from quantities of data, information has become a multiplier of power and a significant component in the value chain for financial organizations. Today, components based on artificial intelligence are part of almost every process of data structuring, hedging and risk evaluation, interaction with users, resource management, and more.”
According to Pazit Katz Oz, “Solutions based on artificial intelligence are beginning to be adopted by the mainstream of the financial sector. The widespread applications include (among others) risk management, fraud detection and prevention, and customer experience. Banks are now using machine learning to identify atypical behavior patterns among customers – patterns that might foreshadow leaving the bank. Banks are also integrating artificial intelligence that offers financial advice and recommends investment channels to customers, or gives them tips on how to handle household expenses.”
“Actually, because the range of available applications is so broad and varied, one of the central challenges that financial organizations face in procuring AI technologies is the challenge of choosing,” says Idan. “You could compare those organizations today to a kid standing in front of whole shelves full of different kinds of chocolates at the candy store. From such a huge selection, sometimes it’s hard to choose. Each requirement can be met by different solutions on different platforms, with different features. Some allow for work on a large scale and others are for local deployment. And the challenge has evolved into one of correctly choosing and fitting the solution to the requirement. A wrong choice could cost the organization heavily.”
The Most Popular Artificial Intelligence Applications for the Financial Sector
“In a number of areas, artificial intelligence applications have already become integral to banks’ activities,” Pazit explains. “In fraud detection and prevention, for example, AI algorithms analyze massive quantities of data, detect patterns and tendencies, and thus help locate and prevent credit-card fraud and identity theft, and bring suspicions to light regarding money laundering. Trainable AI algorithms for detecting anomalies can enable banks to monitor unusual transactions, such as large sums transferred to unfamiliar accounts, or transactions performed outside of normal business hours. By detecting such anomalies, the algorithms can help banks discover fraud, and even prevent it, in real time.
Another popular application is in the field of credit risk management. An artificial intelligence algorithm may detect a pattern of late payment connected to certain loan requests, relate it to new requests that are similar (with respect to the attributes of the customer and the loan), and thus prevent future defaulting on loans. This type of usage is by no means new, but with progress in technology, open code, and new types of engines, more advanced methods are being developed that enable models to provide broader responses and achieve greater precision and effectiveness.”
An additional area where many banks choose to integrate artificial intelligence technology is the improvement of customer experience and the expansion of services. Pazit adds, “We’ve long enjoyed personalization in the areas of cinema, music, and books, and it’s beginning to gather momentum in finance too. AI applications analyze the customers’ information in order to deeply understand their preferences and requirements for the sake of personalizing the banking experience, and building stronger customer relationships. For example, an AI algorithm could analyze a customer’s history of transactions in order to identify the customer’s preferred category of expenditure, such as products or entertainment. Using that information, the bank can offer targeted campaigns or reward programs suited to the customer’s specific requirements and fields of interest.”
“Another application that’s relevant to the customer experience, as well as saving resources, has to do with improving the efficiency of processes behind the scenes” Idan adds. “For example, the process of checking and approving financial documents. CoDx is a solution by Matrix based on artificial intelligence. It provides an answer for companies interested in implementing a digital transformation, but are obliged to deal with a large volume of photocopied documents and unstructured information, such as certificates, deeds or invoices. The solution handles the significant challenge of changing a company over from processes involving unstructured forms to automation, by using Deep Learning technologies and artificial intelligence. In this way, it provides for savings in personnel, for more efficient functioning of the organization, and ultimately for significantly less time spent on customer service. Our solution, for instance, can receive a document and extract information from it that is essential to the process, such as the type of document, the details of the sender, the sum on an invoice, and so on. Once this type of information is structured, automatic processes can be run – a sort of “express lane” for checking the customer’s entitlements, approving invoices, or even performing financial transactions automatically. Processes that were previously manual, requiring days and demanding the same amount of resources from the customer per invoice, whether for ₪100 or for ₪10,000, are reduced by CoDx to minutes and to near-zero human involvement. The human manages the process but does not do the hands-on work.”
The ChatGPT Revolution in CUSTOMER Relations: Has the Machine Taken the Driver’s Seat?
“Artificial intelligence is now playing a crucial role in customer relations,” says Pazit. “Chatbots provide immediate assistance to customers around the clock, allowing bank employees to focus on more complex tasks. Just as an indication, figures from the Forester consultancy firm show that over the last year, 11% of adult web surfers in the USA have interacted with their bank through a chatbot. And here is where everyone’s new friend, ChatGPT, enters big-time, proclaiming a new standard for human–machine interactions. ChatGPT was recently credited in Forbes with making 2023 the Year of the Chatbot for the banking industry. Unlike chatbots based on rules and templates, which in many cases don’t provide the answer that the customer is looking for, the new chatbots are intelligent digital assistants (IDAs), making for a true revolution.”
Idan agrees and adds: “Up until now, we’ve seen relatively minor changes in the human being’s position within organizations’ core business processes. With technology, people have become more efficient and effective, but have remained in the driver’s seat. However, in recent months we’ve witnessed no less than a revolution, and it is affecting the division of roles between employees and machines. This trend will only increase with time. Instead of being helped by the machine, like a pilot by a co-pilot, the human is recast as a supervisor of the machine, while the machine does the navigating. That trend, which may be becoming a physical reality in the form of automated cars, also applies in the sectors of finance and information systems. The emergence of conversational AI, such as the ChatGPT system, makes the human being entirely replaceable in an organization’s interactions with customers. Systems for predicting and evaluating are becoming more objective – and effective – than employees who can be swayed by beliefs and opinions. The bottom line is that the human being is shifting to the co-pilot’s seat.”
“The applications of ChatGPT in the financial sector range widely. They’re not limited to direct communications in the context of customer service,” says Pazit. “It’s a technology that can integrate into real-time fraud detection, give investment advice, and help organizations and private individuals plan their finances. In other words, the same AI applications that are already being used by many banks anyway – but with greater capabilities, making almost complete automation a possibility.”
ChatGPT Admits: I Might Tend Toward Language That’s Relevant Particularly to English-Speaking Men, and My Responses May Be Offensive
“It’s important to realize that people are not completely dropping out of the equation,” says Idan. “Despite the recent significant advances, the general opinion is that over the coming years, the machines will still require close professional and human supervision. Organizations that offer services to the general population must take into account the biases and other problems that could be built into artificial intelligence, and must use human professionals for monitoring the automatic processes and calling attention to ethical problems. Whereas a researcher of Deep Learning and natural-language processing has tools for critically examining the model and the answers that it provides, a user or organization outside that field may not be able to assess the model and consider whether the answers from it are based on true knowledge, on statistical error, or on bias that’s rooted in a deficient training process. If you ask ChatGPT itself what problems may arise from its use, it will tell you that it can slant toward language that’s relevant primarily to English-speaking men, simply because that’s most of the information it encountered during its training. The system will attest that while it tries to create human-like text without understanding the context or meaning of the words it uses, it may create improper or offensive responses. It’s not a matter of hidden or secret problems, it’s shortfalls that I’ve seen ChatGPT itself admit to.”
“In the new machine age, the job of humans isn’t disappearing, but it’s changing,” Pazit adds. “Where creativity is required – and no less important, where empathy is required – humans have a central role. In fact, you could say that advanced AI capabilities, like those of Chat GPT, enable people to stop doing the repetitive, technical work that makes them robotic themselves, and start being creative.”
