In July 2017, BNP Paribas finalised its acquisition of Financière des Paiements Électroniques (FPE), a fast-growing fintech startup that focuses on simple, online retail banking services. Since its launch three years ago, FPE has opened more than 630,000 accounts. All its success notwithstanding, the startup is anything but disrupting big banks –it has been acquired and now is safely ensconced inside BNP.
This is hardly atypical. At Credit Suisse, we sought to understand exit strategies in fintech. We discovered that the most common ones for startups were acquisitions or quick imitations by incumbent banks. PayPal, once a disruptor itself and now an incumbent, is investing substantially in buying up potential rivals, particularly those in the global payments and merchant transfer space. When not being acquired, fintechs are expanding their reach into unrelated sectors that only seem tangentially financial. Square, a payments provider, is now offering meal delivery with Fastbite in an attempt to compete against Uber Eats. Ultimately, we expect most startups to end up turning into conventional financial institutes, like Zopa, the first peer-to-peer lender with the initial ambition to disrupt the established finance sector but that ended up paying for a banking license to provide traditional banking services.
All this is consistent with a broader trend: more companies have disappeared because of mergers and acquisitions than any other reason. Between 1978 and 2012, the number of companies less than a year old as a share of all businesses declined by 44%.2 And in the financial sector, as we shall see below, individual startups face additional hurdles beyond what others faced in the e-commerce Internet world.
The theory of disruptive innovation, introduced by Harvard Professor Clayton Christensen in 1995, has proved itself a powerful way of thinking about innovation-driven growth. The core principles of the theory are that successful disruptive technologies 1) target unserved or underserved segments; 2) initially perform relatively less well against existing customer needs; and 3) underprice the existing offerings. As a precondition for disruption, incumbent players initially don’t respond to the newcomers’ efforts due to the low margins, whereas from the existing customers’ perspective, the lower price of the new offering does not compensate for its inferior performance. In this environment, disruptive innovators can develop their ideas and test their business model without major interference from established players and their demanding clientele.
However, among the largest and most promising fintechs that Credit Suisse has investigated, there is little evidence that newcomers are targeting underserved customer segments. Instead, all startups in our sample have rolled out their initial offering to already-served banking clients, often targeting the incumbents’ core customer base.
This should not be too surprising. Technology-driven innovation requires a significant level of IT expertise, which is why most of the world’s leading fintech startups are launched in close proximity to talent pools with advanced technological know-how. Fintechs develop and offer services tailored predominantly to their own geographical region, which is usually an economically mature, well-banked space. (The only exception would be China.) So, when it comes to segmentation, fintechs don’t have much choice but to enter an already-serviced space and fight for the customers of incumbent players. And the incumbents react fast. Before long, the startups are either acquired or face direct competition from the incumbent banks.
Is that enough for established banks to stop being worried?
Three waves of machines
Of all the breakthrough technologies applicable to finance, the explosion of machine learning and artificial intelligence seems the most promising. One of the most radical improvements in recent years has been how machines learn. For data scientists and machine-learning experts, March 2016 was a momentous month. AlphaGo, a computer program developed by Google, beat world champion Lee Sedol at the ancient Chinese board game of Go by a score of 4 to 1.3 Before AlphaGo played the board game Go against humans, Google researchers had been developing it to play video games: Space Invaders, Breakout, Pong, and others.4 Without the need for any specific programming, the general-purpose algorithm was able to master each game by trial and error, pressing different buttons randomly at first and then adjusting to maximise rewards.
Game after game, the software proved to be cunningly versatile in figuring out an appropriate strategy and then applying it without making any mistakes. That’s why AlphaGo represents not just a machine that can think – like IBM Watson – but also one that learns and strategises, all without direct supervision from humans.
This general-purpose programming was made possible thanks to a deep neural network: a network of hardware and software that mimics the web of neurons in the human brain.5 Reinforcement learning occurs in humans when positive feedback triggers the production of the neurotransmitter dopamine as a reward signal for the brain, resulting in feelings of gratification and pleasure. Computers can be programmed to work similarly, and the positive rewards come in the form of scores when the algorithm achieves a desired outcome. Under this general framework, AlphaGo writes its own instructions randomly through many instances of trial and error, replacing lower-scoring strategies with higher-
scoring ones. That’s how an algorithm teaches itself to play anything, not just Go.
It is easy to imagine a world where self-taught algorithms play a much bigger role in coordinating economic transactions; AlphaGo simply shows us what may be possible in the near future. With instantaneous adjustment, automatic optimisation, and continuous improvement all quietly managed by unsupervised algorithms, the redundancy of production facilities and wastage in the supply chain should become headaches of the past. Freed from the pressure to vertically integrate and with far fewer resources needed for organisational coordination, smaller players will be able to specialise in best-in-class services and deliver extremely customised solutions in real time when specific demands arise.
At Credit Suisse, we can observe first-hand how intelligent computing has improved operational efficiency. One of the most complex areas where innovative technologies play a key role in improving operational efficiency and reducing risks is compliance. In the second half of 2016, Credit Suisse’s compliance department launched an initiative crucial from both business and regulatory perspectives. The project aimed to enable a holistic view of all its client relationships – “Single Client View” – and consisted of designing and implementing a technology-based solution capable of consolidating client-related data from all regions into a single platform. The two key challenges were making data from different systems compatible and accessing it in real-time from any location to effectively detect and mitigate potential risks. The project was launched in December 2016, with the initial scope of client data covering natural persons in our retail and international wealth management businesses booked out of Switzerland. Next, the team focussed on continuously adding additional client data, improving the data quality, enhancing functionality, and rolling out the technology globally to more users. We are now able to review and assess client information significantly faster than before, adopting a multi-jurisdictional view and simplifying the projection of complex client relationships. Importantly, the system assigns each client a global risk score to assist in internal decision-making and to allow us to use advanced analytics.
Another example is compliance with regard to politically exposed persons, where we were able to speed up internal assessments by 60% while reducing related costs by as much as 40%. Machine learning accelerates and streamlines investigation reviews by over 80% – while covering 20% of the information. Furthermore, when it comes to employee surveillance, we now have the capability to screen a substantial number of activities and check them against an extensive list of potential risks.
Still, this development of artificial intelligence (A.I.) is akin to the early phases of electricity, when it first arrived and replaced steam power in manufacturing. At the turn of the century, most textile factories were still powered by flowing water and waterwheels. Factories that installed steam engines had to accommodate pulleys, belts, rotating shafts, and complex gear systems. In fact, the configuration of such a factory was built around a rigidly imposed, centralised steam engine, sacrificing all possible workflow efficiency. Interestingly, when manufacturers began to adopt electricity, engineers couldn’t even fathom an alternative layout like the modern-day assembly line. Rather, they grouped the electric motors into a big cluster, forgoing the benefits of decentralised power in optimising the workflow. It took almost another two decades before the manufacturers truly reaped the full benefits of electricity.6
Today, most incumbent banks tend to frame A.I. as a cost-cutting measure, substituting human labor in administrative processes. Though this is important, its biggest potential is likely to be so profound that it will transform financial institutions as we know them. The true disruptors won’t be startups; rather, they will be deep-pocketed technology behemoths, spreading their tentacles into the world of finance as cross-boundary disruptors.
“Your margin is my opportunity”, Amazon’s CEO Jeff Bezos reportedly once said. The biggest threats to big banks are not fintech startups but rather Amazon or Google, who shift the industry value chain. These and other cross-boundary disruptors, including Alibaba in China, have been leveraging the existing data they have harvested through e-commerce and then specialising in new services targeting customers outside of big banks. Unlike a startup, which is forced to enter an existing space to gather customer information, these tech giants have already acquired huge volumes of data based on commercial activities outside of finance. Consequently, they naturally turn to new segments outside of traditional financial institutions. This doesn’t mean that big banks will disappear. But if they aren’t prepared, they could be reduced to utility companies: ubiquitous, reliable, but certainly not key, with margins so low that they are hardly attractive to anyone, including investors. This is why all big banks must push A.I. to the next level, while time is still on their side.
Howard Yu is an IMD Switzerland Professor of Strategic Management and Innovation. He specialises in technological innovation, strategic transformation and change management. He is a two-time prize-winning case writer awarded by the European Foundation for Management Development. He received his doctoral degree in management from Harvard Business School. Prior to his doctorate, he worked in the banking industry in Hong Kong.
Urs Rohner has been the Chairman of the Board of Directors of Credit Suisse Group AG since 2011 and was its Vice Chairman from 2009 until 2011. In 2004 he was appointed a member of the Executive Board of the Group and served as General Counsel and as COO. Mr. Rohner graduated with a law degree from the University of Zurich. He is admitted to the bars of the Canton of Zurich and the State of New York.
This article was first published by the European Business Review.