108.07台灣銀行家雜誌第115期 / By David Stinson
What Have American Banks Learned from the Outsourcing Fad?With the Rise of AI, Businesses Must Be Careful Not to Automate Their Profit Centers Away
The call center industry has now survived one assassination attempt. In the 1990s, it was the poster child for corporations hoping to profit from labor arbitrage by moving their operations overseas, but the last decade has seen a reshoring trend. The research firm Everest Group reports that the number of call centers with significant (US) onshore presence increased from 35% in 2010 to 53% in 2015. More important than the numbers themselves is the underlying justification. Reshoring is fundamentally not being driven by politics, but rather by customer experience (although incidentally, this history probably contributed to the trade tensions which would arise several years later.) As Indian customer service started to become a joke on late-night TV, companies started seeing call centers as a profit driver rather than just a cost center. The lesson from this management experiment is that cultural connection is a source of value. This lesson will be worth keeping in mind as the industry prepares for its next threat: technology. AI is now becoming mature enough to be applied in banking. Its most attention-grabbing application in banking is chatbots, which can simulate conversation with a human representative. This technology – still in its early stages – is being applied in a variety of consumer-facing areas, including account management and wealth management (robo-advisory), and also other personal finance applications like expense and tax management. Almost all major banks are now experimenting with similar technology in various ways. These developments have caused customer service specialists to wonder if their time has come, this time. They are right to worry. Technology is continuously transforming all sorts of financial operations, including even those performed by white-collar staff. However, many of the most promising developments are internal. AI is being used to scan documents for regulatory purposes, and analyze investment targets, among a variety of other areas. One could make a strong case that a bank’s relationship with its ordinary customers should be last part of its operations to be automated. Can a Machine Feel?The business logic of a cost-saving process is the same whether it is outsourcing, AI, or traditional IT improvements: eliminating an in-house process also means eliminating the profits arising from that process. With onshore agents costing about US$ 20 per hour in the US in the 1990s, it was understandable, and in some cases quite appropriate that companies would seek the 50% savings from overseas operations. In the banking industry in particular, however, a paradoxical effect occurred as the phone customer service industry developed: the same technologies that allowed banks to interact with customers over the phone ended up increasing costs, as customers did more of their business over that medium. In fact, it was not just costs which were increasing, but also value to the customer.This history holds important lessons for the current story of AI development in banking customer service, but first let’s establish the cast of characters. From a journalistic perspective, it’s fun to write about artificial intelligence as a silicon analog of a human, as the name suggests. From a consumer perspective, however, it’s simply a tool, and it will probably remain an inferior alternative to human contact, even in the long term. The notion of Alexa fully replacing human representatives brings to mind the notion of artificial general intelligence (AGI), which known as a “hard” problem in the field. It is not just a matter of implementation; even the parameters of the problem resist easy definition. Instead of the call center agent in this analogy, the AI is the telephone: a piece of hardware, which may create new job categories in the end.It would however be too sanguine to assert that technological progress will not threaten existing customer-facing jobs. India – the low cost alternative – makes an appearance in this story as well. From a technological perspective, this character is not AI, but rather it represents the maturation of the IT revolution from the 1990s. It can take a full generation between the emergence of a new technology and its application in every area of business. As more functions become available on web and mobile banking, based on integration of internal procedures, customers will be able resolve problems themselves. They may not even realize that such interactions would have previously involved human contact. “India” however has the same core vulnerability as in the previous story: it lacks cultural awareness, as well as the ability to solve particularly complex problems. These features can be summarized as emotional intelligence. It is only helpful to a certain point for a company to add documentation and features to customer-facing interfaces. For particularly serious problems, we want to speak to a human – and now. A company which creates barriers to human interaction risks creating the impression that it wants to avoid engagement. The US tech sector, which once aspired to fully automated management of relations with almost all of its stakeholders, is one of the worst offenders in this regard, and is now facing a backlash. The Empathy EconomyTo understand the potential changes that futuristic technologies can bring to business, it is often useful to refer to fields outside of traditional finance and economics. The term “emotional labor” was created by the sociologist Arlie Hoschild in 1983 to describe work in service sectors like hospitality, public administration, social work, and more, which require staff to feign certain affects or regulate their emotions. Some research on emotional labor has focused on how certain tasks involving emotional intelligence tend not to be adequately recognized or compensated. Rather than replace human operators, one important role of this generation of AI will be to allow the humans to monetize this emotional labor. Customers may not be aware of how AI technology is transforming the work of agents. The system can summarize the customer’s situation before they reach the human agent, and help direct their inquiries. It can also be used to monitor the agents for the tone of their speech. This sounds invasive towards employees, but it also creates the possibility that the above problem could be substantially solved – turning emotional labor into value for the company and the consumer.The main difference between emotional labor and the older routine, “rational” labor is that there is no simple, linear correspondence between input and result. One might compare this difference to the distinction in computing between structured and unstructured data. Structured data (like a web form, or any sort of transaction data) involves a limited number of variables and data sources. Unstructured data, meanwhile, could be transmitted from any channel and involve any format. When we express emotions, we tend not to prefer low-bandwidth channels like web forms.One example of these non-linear consequences involves the increasing use of social media to resolve customer disputes. Customers are finding that tweeting a problem publicly often gets results when other channels have failed. This possibility demonstrates the risks involved in the self-service customer interaction model. Social media also creates corresponding opportunities as well, in ways that will continue to be explored. Some of the customer service positions lost to automation will end up being upgraded to PR positions; some of the work will remain the same, except that the stakes will be higher. One of the current focus points of customer service automation is multi-channel integration, helping agents track customer communications across external platforms and internal systems.These two types of technologies will set the stage for the upcoming competition between cost savings and value creation. This analogy between outsourcing and automation is not perfect. Previous moves offshore may have been partially driven by a desire to weaken unions, given the contentious history of labor in the US. Meanwhile, with chatbots, many of the most exciting developments may occur outside of banks’ platforms and brands. Open banking regulations in the US still lag those in Europe, despite a growing consensus on the importance of anti-monopoly regulation. The fate of standardized interfaces for financial information will likely determine the balance of technological development between banks and external platforms.Furthermore, the issues of privacy and information security have become exponentially more complex since the 1990s. AI feeds on big data, and it may seem reasonable to feed it as much as possible. In fact, there is no reason that a privacy breach by a machine would be less likely to provoke a backlash than one by a human. This goes back to the theme that banks must resist the temptation to use technology to create distance from their customers. The clear lesson from this analysis is that human contact is superior form of communication, especially for non-routine issues. As it becomes rarer, it may even become a sort of luxury. Customers are expressing increasing dissatisfaction with over-automation, and some companies are even making a point of limiting its use in customer-facing channels. The American telecom provider T-Mobile recently promised “No bots. No bouncing. No BS!!” Some banks will undoubtedly attempt the low-cost customer service strategy. It is also important to remember that business process offshoring is also still alive, despite the most recent reversal. Yet it is becoming clear that this strategy is not necessarily simpler to execute successfully in the long term. It may in fact be more complex.