How to make A.I. transformation more likely to succeed
What are we learning about artificial intelligence in financial services?” asked Ms. Lael Brainard, one of the seven members of the Board of Governors of the US Federal Reserve. “My focus today is the branch of artificial intelligence known as machine learning, which is the basis of many recent advances and commercial applications,” the governor told her audience in Philadelphia, Pennsylvania. “Due to an early commitment to open-source principles, AI algorithms from some of the largest companies are available to even nascent start-ups… So it is no surprise that many financial services firms are devoting so much money, attention, and time to developing and using AI approaches.”
JPMorgan Chase is reportedly devoting some USD 10.8 billion to its tech budget in 2018. Europe’s largest bank, HSBC, is spending USD 15 billion on new technology. And the biggest spender of all, Bank of America, has set an annual global budget of nearly USD 16 billion for technology and operations. That figure is at least USD 3 billion more than Intel, Microsoft or Apple spent on research and development in 2018. As Andrew S. Grove, the long-time chief executive and chairman of Intel Corporation, told a Stanford researcher in 1991, “Don’t ask managers, ‘What is your strategy?’ Look at what they do!
Because people will pretend.” Whether they are pretending or not, the resource allocation patterns suggest that banks are now effectively IT companies.
What Grove saw as the actual strategy of a firm is the cumulative effect of day-to-day prioritizations or decisions made by middle managers, engineers, salespeople, and financial staff – decisions that are made despite, or in the absence of, intentions. And that is where the problem lies. Money for new investments accounted for only 27% of bank spending on information technology in 2017. According to Celent, a research and consulting company based in Boston, the rest – close to 73% of spending – was allocated to system maintenance. Of the nearly USD 10 billion JPMorgan Chase spent on IT in 2016, only USD 600 million was in fact devoted to fintech solutions and projects for mobile or online banking, although CEO Jamie Dimon warned shareholders in his letter that “Silicon Valley is coming.”
This knowing-doing gap is no simple pretension by senior leadership. Financial institutes we have spoken with have (1) all organized employee seminars inviting motivational speakers to talk about innovation; (2) established corporate venture funds to invest in innovative startups; (3) practiced open innovation, posted challenges online, and run tournaments with external inventors; (4) organized “design thinking” workshops for employees to re-think customer solutions outside the mainstream; and (5) installed Lean Startup methodologies that allow employees to fail fast in order to succeed early. So widespread is the innovation process, and yet, managers continue to face unyielding organizations whose core business is being encroached on by Google and Amazon, if not Tencent or Alibaba or some other digital upstarts. “Tell me one thing that I should do but haven’t done,” hissed an executive the moment we mentioned Google Venture. It seems that no matter how hard these in-house innovation experts try, their companies simply will not budge. The ships are not just big; they cannot turn. Why?
Seizing a window of opportunity is not necessarily about being the first, but about getting it right first.
Too many innovation experts are focusing solely on the nuts and bolts of everyday implementation: gathering data, tweaking formulas, iterating algorithms in experiments and different combinations, prototyping products, and experimenting with business models. They often forget that the underlying technologies – AI in this case – never stay constant. Seizing a window of opportunity is not necessarily about being the first, but about getting it right first. In this instance, that means getting it right for banking clients. Doing so takes courage and determination, as well as vast resources and deep talent. But the banking industry is not where Silicon Valley comes first – the auto industry is.
How likely is it that your industry will be disrupted by the Valley?
No automaker today would speak to investors without mentioning “future mobility.” BMW is “a supplier of individual premium mobility with innovative mobility services.” General Motors aims to “deliver on its vision of an all-electric, emissions-free future.” Toyota possesses the “passion to lead the way to the future of mobility and an enhanced, integrated lifestyle.” And Daimler, maker of Mercedes, sees the future as “connected, autonomous, and smart.” In contrast to the personally owned, gasoline-powered, human-driven vehicles that dominated the last century, automakers know they are transitioning to mobility services based on driverless electric vehicles paid for by the trip, by the mile, by monthly subscription, or a combination of all three. In the past, mobility was created by individual cars automakers sold; in the future, mobility will be produced by service companies operating various kinds of self-driving vehicles in fleets over time. At the BMW Museum, anyone can witness the gravity of this vision first-hand, articulated by its chairman of the board.
Walking up the spiral ramp of a rotunda inside the BMW Museum, one sees flashes of pictures about BMW history that display in variable sequences, slipping in and out of view like mirages. At the very top of the museum is a “themed area” of about 30 stations demonstrating an emissions-free, autonomously driven future. These are not only a vision, but a real project, begun in earnest in the autumn of 2007 by then-CEO Norbert Reithofer and his chief strategist Friedrich Eichiner. The two men tasked engineer Ulrich Kranz, who had revived the Mini brand in 2001, to “rethink mobility.” The task force soon grew to 30 members and moved into a garage-like factory hall inside BMW’s main complex.
“I had the freedom to assemble a team the way I wanted. The project was not tied to one of the company’s brands, so it could tackle any problem,” Kranz said in an interview with Automotive News Europe in 2013. “The job was to position BMW for the future—and that was in all fields: from materials to production, from technologies to new vehicle architectures.” And so Kranz and his team decided to explore uncharted territory that included “the development of sustainable mobility concepts, new sales channels, and marketing concepts, along with acquiring new customers.” The starting point for “Project i” was, in other words, a blank sheet of paper.
“We traveled to a total of 20 megacities, including Los Angeles, Mexico City, London, Tokyo, and Shanghai. We met people who live in metropolises and who indicated that they had a sustainable lifestyle. We lived with them, traveled with them to work, and asked questions,” Kranz recalled. “We wanted to know the products that they would like from a car manufacturer, how their commute to work could be improved, and how they imagined their mobility in the future. As a second step, we asked the mayors and city planners in each metropolis about their infrastructure problems, the regulations for internal combustion engines, and the advantages of electric vehicles.”
Once the findings came back, Kranz expanded his team by seeking out “the right employees both internally and externally.” The result was BMW’s gas-electric i8 sports coupe and all-electric i3 people mover, which shimmered under white lights at BMW World, where the company’s top automotive offerings are showcased. The i3 had almost no hood, and the front grille was framed by plastic slits that looked like a pair of Ray-Bans. It came in a fun-looking burnt orange. The front seats were so vertically poised, with the dashboard stretching out, that they exuded a “loft on wheels” vibe. Like the interior, made of recycled carbon fiber and faux-wood paneling, the electric motor of the i3 was geared to urban dwellers in megacities who yearned for a calm, relaxing drive.
What made BMW all the more remarkable was its timing. Almost two years before Tesla’s Model S was introduced, BMW had presented the battery-powered car as a revolutionary product, and committed to build it and deliver it to showrooms by 2013. When the BMW i3 went on sale, Tesla’s Model S had spent just over a year on the US market. The 2014 i3 went on to win a World Green Car award, as did the 2015 model, the i8. In short, BMW was fast and early.
Then something terrible happened – or really nothing happened.
The i3 is now five years old, and the i8 is four. The BMW i brand includes the services DriveNow and ReachNow (car sharing), ParkNow (to find available parking), and ChargeNow (to find charging stations). But, besides being featured in occasional press releases, Project i has given way to other BMW sports cars in prime-time TV advertising spots. There is no news from Project i, except that project members are reportedly leaving. Ulrich Kranz, the former manager, joined former BMW CFO Stefan Krause at Faraday Future, and after a short stay, they started Evelozcity in California, where they recruited another i-model designer, Karl-Thomas Neuman. And Kranz is not alone. Carsten Breitfeld, former i8 development manager, is now CEO of Byton, where he also enlisted a marketing expert and a designer from the BMW team.
How much Project i has cost BMW, we will never know. But if, according to BMW figures, the carbon-fiber production and the autobody works for the i3 set the company back some half a billion euros, the entire project could easily have cost two to three billion – a sum that would have been enough for the development of two to three series of a conventional VW Golf or Mercedes S-Class. Two to three billion euros is also more than fifteen times the USD 150 million Apple spent to develop the first iPhone, which launched in 2007. With so much bleeding, the new CEO Harald Krüger talked of Project i 2.0, a plan to integrate the BMW i sub-brand back into the parent company, and refocus distribution efforts on “classic” products.
In 2018, BMW USA reported just 7% of its sales were cars with a plug, which included all its hybrid offerings. Meanwhile, Tesla reported booming sales of its Model 3, which has become one of the USA’s top 20 most-sold vehicles in the third quarter of 2018. Tesla was ranked fourth in luxury car sales during the same quarter. At the time of writing, Tesla has surpassed BMW and Daimler to become the world’s second most valuable automaker in terms of market capitalization, trailing only Toyota.
Then something terrible happened – or really nothing happened.
Did Tesla and other start-up companies steal BMW’s idea and run with it? No, it is what is called the Zeitgeist, a German word meaning “spirit of the time.” When the time is ripe, the ideas are “in the air.” Competition invariably emerges, and companies have to improve their ideas to stay ahead. They need to come up with demonstrations that excite potential customers, potential investors and, more importantly, potential distributors.
BMW’s shift in its distribution of the i sub-brand echoes what Kodak did. Kodak built the first digital camera back in 1975 and was the first to put out a competent product, but then ended up folding its consumer digital and professional divisions back into the legacy consumer film divisions in 2003. Meanwhile, Nikon, Sony, and Canon kept innovating in the subsequent decades, with features like face detection, smile detection, and in-camera red-eye fixes. We all know what eventually happened to Kodak.
How to become future-ready
BMW is by no means a laggard in innovation. At IMD business school in Switzerland, we track how likely a firm is to successfully leap toward a new form of knowledge. For automakers, it is the shift from mechanical engineering, with combustion-engine experts, to electric and programming experts of the same kind as those who build computers, mobile games, and hand-held devices. For consumer banking, it is the shift from operating a traditional retail branch with knowledgeable staff who provide investment advice to running data analytics and interacting with consumers the same way an e-commerce retailer would. The pace of change may differ between industries, but the directional shift is undeniable.
The IMD ranking measures companies in each industry sector using hard market data that is publicly available and has objective rules, rather than relying on soft data such as polls or subjective judgments by raters. Polls suffer from the tyranny of hype. Names that get early recognition get greater visibility in the press, which accentuates their popularity, leading to a positive cascade in their favor. Rankings based on polls also overlook fundamental drivers that fuel innovation, such as the health of a company’s current business, the diversity of its workforce, the governance structure of the firm, the amount it invests in outdoing competitors, the speed of product launches, and so on. According to an objective composite index like this one, BMW is among the best. Table 1 shows the ranking of the top 55 automakers and component suppliers. The methodology of the ranking is described in the appendix.
But the index also points to the general conservatism of large companies. Most radical ideas fail, and large companies cannot tolerate failure. It does not matter whether you call BMW’s strategy “throw everything at the wall and see what sticks” or a groundbreaking, iterative approach to mobility; if the only way to innovate is “to put a few bright people in a dark room, pour in some money, and hope that something wonderful will happen,” Gary Hamel once wrote, “the value added by top management is low indeed.”
But it is not just about cars. The dilemma experienced by German automakers is strikingly similar to the ones facing executives in banking and a host of other industries these days. Just as Detroit is confronted by Silicon Valley, Wall Street can see the future of banking everywhere it looks. Turning to China, it sees Alibaba, whose AliPay has become synonymous with mobile payment, and Ant Financial, Alibaba’s finance subsidiary, which is now worth USD 150 billion – more than Goldman Sachs. Looking homeward, it sees that start-ups like Wealthfront, Personal Capital, and Betterment have all launched robo-advisors as industry disruptors. In retail checkout lanes, it sees Square or Clover or Paypal Here taking in credit card payments on behalf of millions of small-time merchants. It sees that the future of banking is not only about Big Data analytics, but also about calling on and bundling a group of financial services that happen in real time and with little human interaction. A smart infrastructure that automatically interacts with customers, continuing to improve its algorithm and adjust its response without human supervision as it handles data gushing in from all around the world at millions of bytes per minute, is tantamount to one giant leap forward for every banking incumbent.
Deep-learning-based programs can already decipher human speech, translate documents, recognize images, predict consumer behavior, identify fraud, and help robots “see.” Most computer experts would agree that the most direct application of this sort of machine intelligence is in areas like insurance and consumer lending, where relevant data about borrowers – credit score, income, credit card history – is abundant, and goals such as minimizing default rates can be narrowly defined. This explains why, today, no human eyes are needed to process any credit request below USD 50,000. For these businesses, the question of where and how to deploy AI is easy to answer: find out where a lot of routine decisions are made, and substitute algorithms for humans.
But data can be expensive to acquire, and investment conventionally involves a trade-off between the benefit of more data and the cost of acquiring it. For many traditional banking incumbents, the path to AI is anything but straightforward. Managers are often tasked with considering how many different types of data are needed. How many different sensors are required to collect data for training? How frequently does the data need to be collected? More types, more sensors, and more frequent collection mean higher costs along with the potentially higher benefit. In thinking through this decision, managers are asked to carefully determine what they want to predict, guided by the belief that this particular prediction exercise will tell them what they need to know. This thinking process is similar to the “re-engineering” movement of the 1990s, in which managers were told to step back from their processes and outline the objective they wanted to achieve before re-engineering began. It is a logical process, but it is the wrong one.
Consider the process of shopping at Amazon. Amazon’s AI is already predicting your next purchase under “Inspired by your browsing history.” Experts estimate the AI’s success rate at about 5%, which is no small feat considering the millions of items on offer. Now imagine if the accuracy of Amazon’s AI were to improve in the coming years. At some point, the prediction would be enough to justify Amazon pre-shipping items to your home, and you would simply return the things you did not want. That is, Amazon would move from a shopping-then-shipping model to shipping then shopping, sending items to customers in anticipation of their wants. The complication lies in when Amazon should introduce this AI-driven fulfillment service. With the underlying technology improving, Amazon might choose to launch such a service just a year ahead of the competition, when the AI prediction is not yet perfect, and suffer a hit on returns and a dip in profitability. Why? Because launching the service slightly sooner will give Amazon’s AI more data sooner than the competition, which will mean its performance will improve slightly faster than that of others. Those slightly better predictions will in turn attract more shoppers, and more shoppers will generate more data to train the AI faster still, leading to a sort of virtuous cycle. 
In fact, this data intelligence is the only first-mover advantage that matters. It grows from a positive feedback loop. The more data that is used, the more valuable the business becomes, since getting relevant data in quantity is always difficult and expensive. This is why Google Maps becomes more accurate as more people use it: the underlying algorithms have more data to work with, so the apps become even more accurate. Google has made two decades’ worth of investments to digitalize all aspects of its workflow, but not because it has a clear notion of what it wants to predict. It had done so before a clear notion of AI fully emerged. This is the groundwork that must be laid before a well-defined strategy for effective AI can be established, which means the conventional budget allocation will not work for banking incumbents seeking to scale their footprints in the age of AI. They have no choice but to follow a disruptive playbook, but with a twist.
How understanding disruption helps strategists
In the early 1990s, Professor Clayton Christensen of the Harvard Business School noticed an interesting pattern among companies facing the emergence of a new technology. When technological progress was incremental, even if the increments appeared in rapid succession, powerful incumbents always triumphed. Companies that were endowed with vast resources, extensive networks of suppliers, and a loyal customer base were able to command great advantages over their rivals, as would be expected. This is what made IBM a formidable player in the computing industry and General Motors a bellwether organization in the automotive industry.
And yet, there is a class of technological changes in which the new entrant – despite far fewer resources and no track record – almost always topples existing industry giants. This special class of technological changes, Christensen noted, does not have to be sophisticated or even radical.
Take transistor television as an example. When RCA first discovered transistor technology, the company was already the market leader in color televisions produced with vacuum tubes. It naturally saw little use for transistors beyond novelty, and decided to license the technology to a little-known Japanese firm called Sony. Sony, of course, could not build a TV out of transistors, but it did manage to produce the first transistor radio. The sound quality was awful, but the radio was affordable for teenagers who were delighted by the freedom to listen to rock music away from the complaints of their parents. Transistor radios took off. Still, the profit margins were so low that RCA had no reason to invest further. It was busy making serious money and investing every R&D dollar on improving vacuum tube color TV.
Sony, meanwhile, was looking for the next big thing. It launched a portable, low-end, black-and-white TV at a rock-bottom price, targeting low-income individuals. Called the “Tummy Television,” it was tiny enough to perch on one’s stomach – the antithesis of RCA’s centerpiece of middle-class living rooms. Why would RCA invest in transistors to make an inferior television for a less-attractive market? It did not.
The real trouble began when Sony finally pushed the transistor’s performance to produce color TVs based entirely on the new technology. Overnight, RCA found itself trying to catch up on a technology that it had ignored for the past three decades, which it had ironically pioneered and licensed out. Christensen called this type of technology – inferior at first but immensely useful later – disruptive, a term that has since been immortalized in the business lexicon of executives, consultants, and academics.
What we see today in the financial industry are new entrants leveraging digital interfaces and AI decision-making processes that involve minimal manual work to target an underserved market segment. Their technologies cannot satisfy high-end banking customers yet. But like the desktops that displaced minicomputers, or the angioplasty that displaced open-heart surgery, AI and digital automation will inevitably improve and, one day, these new solutions will be able to meet a substantial part of the need among big clients. The implication is that there will always be space for manual-intensive, human-centric operations, but that space will shrink substantially in the future.
One logical solution is for banking incumbents to create a separate unit and launch “speed boats” that adhere strictly to the playbook of digital disruptors. These will target an underserved market, and provide security services on a digital platform, with minimal human intervention. Initiatives like this are meant to develop a new set of capabilities – advanced analytics, dynamic product deployment, linking to third parties to fill a sudden surge in market demand – initially targeting a new segment that does not interfere with the mainstream business of the current banking operation. Over time, such new businesses will develop crucial capabilities that will mature enough to be transplanted back into the mainstream. This approach prevents the often-heard refrain of IT large-scale transformation: overtime, overbudget, and with underwhelming market results. In a way, it is RCA launching Sony’s transistor radio, but keeping ownership of it to get future technologies ready.
And here is one last twist. Scaling up a disruptive business will always be costly. The company will suffer financial loss for years, if not decades, and in the foreseeable future, will be unlikely to achieve the level of profitability of the core business. BMW has been profitable for a very long time; Tesla is still operating at a loss today, as is Uber.
From Amazon to Square to Ant Financial, profitability is not the most important metric for managers; user base and market share are. This is why banking incumbents need to consider an alternate investment structure, allowing third parties, venture capitalists, and even competitors to take an equity stake. Such a structure seems controversial, but is not unprecedented. Alibaba does not own all of Ant Financial. After exiting China, Uber now owns a minority share of its Chinese rival Didi. This is also the new strategy of GM’s CEO Mary Barra, and it paid off handsomely in May 2018 when SoftBank announced a USD 2.25 billion investment in Cruise Automation – the self-driving unit of General Motors, headquartered in San Francisco. The investment pushed Cruise, originally purchased by GM for USD 581 million, to USD 11.5 billion. It takes more than just vision, belief, passion and experimentation in AI to transform a company. It takes a pocket so deep that it requires other people’s money to act on that aspiration. It is an unconventional approach taken during an unconventional time.
One last flashback…
Adjacent to the Mercedes-Benz museum in Stuttgart, Germany is one of the largest Mercedes dealerships in the world, which we also visited during the autumn of 2018. Its cavernous main hall is preceded by a restaurant, a café, and a shop hawking Mercedes-Benz merchandise. We saw a vertical banner stretching from the ceiling to the floor along the glass panels on one wall. “Ready to change,” the banner cheered. “Electric intelligence by Mercedes Benz.” It referred to Concept EQ, a brand of electric plug-in models first unveiled in Stockholm on 4 September 2018. There were three EQs on display next to an exhibition kiosk that did not work, but instead displayed an error alert and tangled cables spilling from the back that had come unglued.
Then, on the top floor, visitors were gawking at a Mercedes-AMG, known for its “pure performance and sublime sportiness.” Here was a vision of a forward-looking sports car with all its driving pleasure fully realized. The risers and the wrap-around LCD walls only accentuated the carbon-fiber composite of the chassis, gleaming in matte black. But we also noticed that the CO2 emissions rating of this Mercedes AMG GT 63 S, with its 630 horsepower, was an F.
Table 1: Top 55 automakers and component suppliers
|General Motors Company||98.357||2|
|Ford Motor Co.||82.265||4|
|Toyota Motor Corporation||82.235||5|
|Nissan Motor Co., Ltd.||81.442||6|
|Bayerische Motoren Werke A.G.||71.473||7|
|Robert Bosch Gmbh||47.094||15|
|Fiat Chrysler Automobiles N.V.||43.215||16|
|Brilliance China Auto Holdings||42.935||17|
|Cooper-Standard Holdings Inc.||36.989||22|
|Baic Motor Corporation Ltd.||35.015||23|
|Skoda Auto, A.S.||34.876||24|
|Guangzhou Automobile Group||33.444||25|
|Yamaha Motor Co., Ltd||32.383||26|
|Fuyao Glass Group Industries||31.058||27|
|Hyundai Motor Co., Ltd.||29.133||28|
|Jaguar Land Rover Ltd.||28.849||29|
|Suzuki Motor Corporation||27.926||31|
|Byd Company Ltd.||27.702||32|
|Geely Automobile Holdings Ltd.||27.568||33|
|Magna International Inc.||27.077||34|
|Mitsubishi Motors Corporation||24.689||35|
|Chaowei Power Holdings Ltd.||24.134||36|
|Mazda Motor Corporation||22.551||37|
|Tata Motors Ltd.||21.093||39|
|Beiqi Foton Motor Co., Ltd.||20.672||40|
|Kia Motors Corporation||17.535||41|
|Isuzu Motors Ltd.||17.462||42|
|TS Tech Co., Ltd.||17.074||43|
|Haima Automobile Group Co.||13.603||44|
|Aisin Seiki Co., Ltd.||11.655||46|
|Saic Motor Corporation Limited||10.135||47|
|Mahindra & Mahindra Limited||8.539||48|
|China Faw Group Co., Ltd.||6.358||50|
|Anhui Jianghuai Auto Group||5.043||51|
|Jiangling Motors Corporation||4.127||52|
|Dongfeng Motor Group Co., Ltd.||2.925||53|
|Chongqing Changan Auto Co.||0.181||54|
|Great Wall Motor Co., Ltd.||0||55|
Source: see Appendix
This article was originally part of a report by The Credit Suisse Research Institute on AI and the future of work.
Howard Yu is the author of LEAP: How to Thrive in a World Where Everything Can Be Copied (PublicAffairs, June 2018), and LEGO professor of management and innovation at IMD. In 2015, Yu was featured in Poets & Quants as one of the Best 40 Under 40 Professors. He was shortlisted for the 2017 Thinkers50 Innovation Award, and in 2018 appeared on the Thinkers50 Radar list of 30 management thinkers “most likely to shape the future of how organizations are managed and led.” Yu received his doctoral degree in management from Harvard Business School.
 For an excellent analysis of this thought experiment, please refer to Prediction Machines: The Simple Economics of Artificial Intelligence by Ajay Agrawal, Joshua Gans, and Avi Goldfarb. https://www.amazon.com/dp/B075GXJPFS.
Research Information & Knowledge Hub for additional information on IMD publications