By Cem Dilmegani, Co-founder and CEO, AppliedAI.com
“Why do banks exist?” Professor David Beim would ask as a thought experiment during our finance classes at Columbia. Knowing the primary purpose of an institution is important because that purpose defines the institution; once the means to serve that purpose changes, the institution needs to change. That is the case with banks.
Banks help manage differences between cash flows and expenses. This is a problem that involves decision-making under uncertainty, and artificial intelligence (AI) is providing us with a new means to solve that problem. Additionally, all non-core areas of banking are also undergoing transformations with the advance of AI. AI-powered banks will provide faster service at lower cost; however, this will require the transformation of their personnel. They will need to be skilled in building and maintaining models, augmenting their capabilities with AI. This will be a societal transformation, as the financial-services and insurance sectors have employed more than six million people in the United States (as of 2015).
History of automation in banking
Value added by the financial sector, % of GDP from The Organisation for Economic Co-operation and Development (OECD) Economic Policy Papers: Finance and Inclusive Growth
The suggestion that automation reduces banking jobs was not proven correct most of the time. For example, analysis by Deloitte of the last 140 years of employment data shows employment increases due to technology. Banks are also no strangers to tech innovation. They have thrived with the computing revolution since the 1980s. They increased value added by the financial sector as percentage of gross domestic product (GDP) from approximately 5 to 8 while also expanding their headcounts. The headcount figures for the industry don’t decrease in significant amounts except during recessions, and then they bounce back after the recessions.
All finance and insurance employees in the US from the US Bureau of Labor Statistics
We can examine a case in banking innovation in more detail: the impact of automated teller machines (ATMs) on tellers. As James Pethokoukis (AEI – American Enterprise Institute) explains, ATMs did not turn out to be job-killers. ATMs helped tellers focus on sales and marketing while leaving simpler transactions and cash handling to ATMs. Therefore, it was possible to open branches with fewer tellers, leading banks to open more branches serving a larger share of the population. More branches required more tellers, ensuring an increase in the number of tellers.
Impact of ATMs on tellers from “Learning by Doing: The Real Connection between Innovation, Wages, and Wealth” by James Bessen
So what’s different this time around? Why can’t we have more customer-service specialists after chatbots are introduced, just like we had more tellers after ATMs were introduced? First, we need to envision the end state of banking. Let’s assume that we can produce AI that’s on par with the human mind. AI experts surveyed in numerous studies (Seth Baum, arXiv) believe that this will happen by approximately 2060, or possibly before that. If we accept expert estimates and assume we will produce AI that is as capable as the human mind, then would we employ humans in banking? There could be a few cases such as commercial banking in which relations matter—and relating to a human will possibly be easier than relating to a robot. Or the strategy and top-management functions that require “creativity” could be fields in which we add value to machine work. That future does not look like one in which we will employ millions of people in banking, especially if people’s skill levels do not change.
If we believe this vision of the future, there needs to be a time when we start reducing the banking workforce. In the case of tellers, that day has come. As digital-channel adoption increases, banks are closing branches, which results in fewer tellers employed. The Bureau of Labor Statistics predicts an 8 percent decline in the number of tellers until 2024.
Number of branches in the US from Federal Reserve Bank of St Louis
The areas in which AI will impact banking jobs
We can think of banking compromising six core activities and overall compliance. These activities exist in different functions, such as commercial and retail banking, and they are sometimes merged into single units making them difficult to identify. They are: marketing, sales, decision-making, operations, customer service and technology. On top of these activities a common compliance function ensures that banks’ actions follow the necessary rules and regulations.
Any banking product first needs to be marketed to the right segment. Digital and mobile has been transforming marketing in all industries. AI is enabling companies to reach the right customers, with the right products and messages through the right channels. Retargeting companies are using machine learning to identify the right customers. They detect customers who have previously expressed interest in certain topics and target them on the same topics. The next best action systems provide personalized product recommendations enabling customers to reach the right products. AI-enabled marketing vendors aggregate disparate datasets to determine which screen your customer is using at a given time, helping you identify the right channel. With all these advances, a key responsibility of marketer is to focus on building an effective marketing stack making full use of data both inside and outside the company.
Once the right customer is reached, he or she needs to decide to buy the product. With better digital interfaces, sales are moving online. And banks are ready for that move since they can now offer 24/7 available, patient, knowledgeable robo-advisors. Fewer outbound call-center reps and tellers will be needed in retail banking. Sales activities in commercial banking, especially for small and medium-sized enterprises (SMEs), are increasingly moving to inside sales for which AI can analyze calls and even suggest responses to sales reps, increasing their effectiveness.
Once the customer shows his or her intent to purchase, the bank needs to decide its offer. Is this customer likely to default, or not? Decision-making, the core of banking, has seen waves of automation and will be further automated. If banks do not roll out the most advanced credit-scoring systems leveraging their vast data, fintech startups will. Companies such as OnDeck Capital are disrupting SME lending with their fast and easy-to-use services. Numerous other fintech companies are disrupting retail lending in a similar manner. And banks are taking notice, as seen by JPMorgan’s deal with OnDeck Capital.
Operations ensures delivery of products or services after a customer accepts the bank’s offer. As stated in the beginning of this article, ATMs transformed operations. Tellers needed to deal with cash much less frequently so they had more time to focus on marketing activities such as calling customers with new offers, or sales activities such as cross-selling to customers with whom they are working. Operations activities continue to be automated. With the rise of Robotic Process Automation (RPA), repetitive, manual back-office activities will be quickly automated. A decade ago such automations would have required system upgrades that would have taken years, but now smart bots can work with legacy systems and deliver automation benefits quickly.
Customers will have questions and issues through their lifecycles. An excellent customer service is required to provide an excellent customer experience. Chatbots can provide fast, accurate results to simple customer queries already, achieving significant savings. Numerous AI vendors work with banks to provide conversational interfaces; there are even conversational AI platforms such as Kasisto or Personetics that focus exclusively on serving banks.
Finally, the technology needs of these activities need to be satisfied. From monitoring security with autonomous systems and luring unwitting hackers with decoys, AI systems have already transformed cybersecurity. Additionally, a wide range of tools are becoming available to solve business problems. AI libraries developed by tech giants such as Google or upcoming startups can be leveraged to make custom solutions that read, see or speak at levels acceptable to use in some banking applications.
All these activities will require less human input with increased automation. And until we get there, compliance is one of the most important use cases for banks. Deep-learning and topological data analysis are helping banks reduce the number of cases that humans need to examine, making fraud-detection and anti-money-laundering efforts largely automated.
Future of banking jobs
Even after the realization of automation, there will remain the need for an AI-ready workforce. Personnel will be needed to lay out strategies, manage workforces, build systems and models, have face-to-face customer interactions, resolve inquiries and make decisions that can’t be handled by bots. However, those tasks form a scope much smaller than that of today’s banks. Smaller scope does not necessarily mean smaller workforces. We could have more engineers and data scientists building systems, more strategists envisioning the bank’s strategy. But those skills require more work experience and education than an average banking employee currently has. Unless current employees dramatically improve their skills with banks’ help, banks will need to start operating as much smaller entities.
It’s hard to know which of these scenarios will become reality, but given the competitive nature of banking and increasing encroachment of fintech players, banks do not have too much time. Many will become much smaller in terms of headcount. Some will push themselves to retain and train a significant portion of their workforces. But inevitably, survivors will have transformed their businesses and workforces to become AI-enabled banks.