Bank 4.0 Read online

  Endnotes

  1For example, in China mobile payments and micro-loans.

  Fast startup vs slow corporate? The word “corporate” itself is now often used as an adjective to describe a type or mode of company that, as a FinTech startup, you don’t want to emulate. CEOs of startups often say: “We don’t want to go all corporate”.

  So what do we understand by the word corporate? For many people, when they hear the word corporate it implies: slow moving, bureaucratic, potentially out of date. Not words that we associate with the dynamic FinTech companies that we read about in the press. However, there are two thoughts on this that may give us a clue into what is driving the optical differential in growth, and they both stem from one word: “legacy”.

  Legacy, the gift or transfer of value from the past to the present time, the notion that value is being created and built upon.

  Legacy, the term used to describe out-dated systems and processes that are no longer current and competitive.

  So when we are considering a large corporate, we do have to view it with both definitions in mind. They have, by their very nature, demonstrated an ability to create a scaled, profitable business that has endured for decades, serving multi-generational customers, returning money to shareholders and with a capacity to invest for the future. The legacy of value.

  With this we also typically see an organisation that has its foundations in technology, culture and organisational design that is from a different time. It has an iterative approach to product, technology and organisational systems—which makes it very difficult to transform across all of these dimensions.

  When we look at the growth of FinTechs, what we see is often the development of a business from a first principles basis, and this is true across all business functions, and in all of the inherent organisational processes within the business. It has the ability to create an organisation to fit the current times, the current challenges and to develop current solutions. In all elements of the business, whether that is the design of the product or service, there is no requirement to keep in mind the management of the existing business or customers. They are not constrained by the maintenance of existing revenue streams and the management and migration of existing customers. They are the notional blank sheet of paper. But hasn’t this always been the case for startups?

  And the answer is, yes, it has been—but with a difference. We now live in a time where technology has fundamentally, seismically shifted two of the biggest barriers to entries, or to put it another way, “moats around the incumbents castle”. This is the ability to create product, and the ability to distribute it.

  This technologically-driven and -enabled innovation capability, coupled with an organisation that is using first principles design processes, without a legacy of customers (and even employees) that it needs to consider, has an enormous advantage: speed. Products are envisaged, prototyped, tested, amended, refined, and launched in timescales that may be faster than a large corporate. So is it all down to the technology? Well, it is in part; but there appears to be another, potentially even more impactful, difference between the fastest-growing FinTechs and the incumbent corporates: culture.

  In creating an organisation from scratch, there is an ability to develop a culture that itself is designed to operate at a pace, shape and even a method that mirrors the technological design capabilities of our times. Leveraging agile collaboration tools such as Slack/Trello, in an environment where diaries are managed months in advance and involve complex steering committee and matrix alignment critical for all decisions, is an anachronism. This is why so many large corporates that are trying to transform their innovation capabilities are struggling—the legacy of the culture plays a very large role.

  So we know that we have the ingredients for fast-paced growth in a first-principles based, technologically-driven startup. With the right leadership, culture and persistence they have an ability to develop product at pace that is not constrained by the existing norms of operating a scaled business. The ability to leverage technology to take these products to market is the key difference over the last 10 years. A FinTech app can be downloaded directly to one of millions of individual consumers in a matter of seconds from a device that fits in your pocket. Imagine how inconceivable it would have been 20 years ago for a company like Instagram, which had at the time 13 employees, to take a product to 30 million people? The distribution model of mobile download has a symbiotic relationship with social media sharing capabilities on the very same platform. The marketplace has been levelled at least, and in the early stages may be even tilted in the direction of the startup.

  But a word of caution. Firstly, our image is distorted, the unicorns are getting a disproportionate share of voice. This of course makes sense—the story telling of the company from the garage becoming a global mega-brand is highly compelling. And conversely, the story of yet another startup failing to achieve a funding round is not.

  Secondly, we are seeing a regulatory environment that is struggling to accommodate for, support and govern a range of rapidly emerging payment and financial services companies. As the FinTech startups hit true scale they are entering a phase where the incumbents are able to play to their own strengths. So the transition from a high-growth, even-scaled FinTech to a company that has served the test of time is a tricky one. The giant financial services corporates have mainly been able to weather the storm of economic downturn, and create a true legacy across generations.

  And that leads us to maybe the final conclusion, and one that we see both the challenging entrants and the legacy incumbents embracing: partnership. Allowing the FinTechs to leverage their natural capabilities in terms of pace, early stage growth, and innovation; but then integrating this into the operating models of the incumbents.

  7 The Role of AI in Banking

  A robot may not injure a human being or, through inaction, allow a human being to come to harm.

  —“Handbook of Robotics, 56th Edition, 2058”, Isaac Asimov (1942)

  In 1942, science-fiction author Isaac Asimov introduced the world to his three laws of robotics1. An incredibly prescient visionary, Asimov started the world thinking about the potential challenges sentient or cognitive technology might present humanity. The number one principle for robotics may end up being: create more value than the human you displaced—the primary threat from AI’s may well be technological unemployment as opposed to robot overlords taking over the planet and enslaving humanity. While likely neither malevolent or benevolent, AI still has the potential to do large scale damage structurally where employment and equality are concerned.

  When you look for the organisations making big bets on Artificial Intelligence today2, the lists always include technology majors, but as yet we don’t see many banks investing anywhere near the scale of Microsoft, Google, Apple, Alibaba, Baidu and others. Industrial players like Boeing and Tesla are by necessity making big bets on AI, so it is entirely reasonable to expect that we should see a similar scale of investments coming through financial services, healthcare, etc. However, when we look at AI in financial services right now, the lion’s share of progress appears to be coming from players like Ant Financial and smaller FinTech’s who are able to specialise in these emerging technologies. Ant Financial themselves is reportedly investing more than $15 billion over the next three years on AI and Quantum computing3. On their current valuation that’s about 10 percent of their total market cap.

  There are a handful of banks taking steps in the right direction. JPMorgan Chase spent 16 percent of their budget on technology in 2016, $9.6 billion in total and up from $1.2 billion in 2012, but they have not disclosed how much of that specifically goes into AI research and development. Goldman Sachs Strats division (quantitive strategy/technology) now makes up around 30 percent of GS’ headcount, and they’ve recently been seen aggressively recruiting AI specialists in Machine Learning (ML), Artificial Intelligence (AI), program management and digital product design. BofA, BBVA, Deutsche and HSB
C are talking about their strategic spend in AI, while TD acquired the AI startup Layer 6 in January of 2018, driven by Rizwan Khalfan, their Chief Digital and Payments Officer.

  The ability to anticipate the needs and preferences of individual customers doesn’t exist in banking today, but will be a requirement going forward… There’s such little talent and expertise in the AI space, and for us to be able to partner with organisations like Layer 6, who are considered both best-in-class from a research and a pragmatic perspective, is really the secret sauce.

  —Rizwan Khalfan, TD Bank Group

  Rizwan points us directly at the core problem for the industry at large. AI is an entirely new skillset and banks don’t have any real expertise in the space and, frankly, are a long way from having world-class capabilities that could compete with the tech majors. Given AI is not a core capability, and banks are starting behind the eight ball on both budget and talent, it’s pretty clear that strategic partnerships, acquisitions and such will be essential.

  The advantage of tech majors is that they have both the capital and technology pedigree to be able to focus on AI. FinTech’s are already built from the ground up around technology, they have talent that is more easily adapted to AI R&D and they don’t have process, policy or legacy that could slow them down. All this adds up to the likelihood of banks slipping further behind on AI over time. Thus, it is likely that when AI starts to operationally impact financial services, incumbents will have far less control over the outcome than, say, the impact of regulatory change or customer behaviour might have on AI.

  When discussing AI in banking or financial services, it’s important to define what exactly we’re talking about. Many bankers make the mistake of thinking of AI is something that is a long way off, and when it comes it won’t be focused on banking. These types of algorithms, which allow for leaps in cognitive understanding for machines, have only been possible with the application of massive data processing and computing power in recent years.

  Talking about AI in general today is like people talking about Tokyo like it’s synonymous with Asia. It belies a misunderstanding about different types of AI, and how and where AI will likely impact banking. For example, we’re not going to need a bipedal android with artificial general intelligence to eliminate a plethora of banking jobs. Even today, with nascent developments in AI, we already have the foundations for material changes in the way we staff financial services over the coming decade.

  In the 2000s, UBS moved their trading floor out to its headquarters in Stamford, Connecticut. The trading floor housed more than 5,000 traders holding pride of place in their 700,000 square-foot building. Today the trading floor stands empty, abandoned as a result of automation of the trading arm of UBS’ business. In quantifying the rate of change, Goldman has found that today one computer engineer can replace four or five traders. Today one-third of Goldman’s staff are already computer engineers as they speed up automation internally.

  Figure 1: UBS trading floor, Stamford, CT, circa 2008 (Before AI)—Today these traders are gone.

  Goldman and UBS use complex algorithms that mimic what a human trader used to do—simple machine intelligence with human equivalent decision-making capability for a specific task. One good example of this is the project that UBS and Deloitte created in 2016—a simple, automated program for dealing with their clients’ post-trade allocation requests. The system does an automated review of emails sent by clients detailing how they want to allocate large block trades across funds, then processes and executes the required transfers. This takes the automated system seconds to execute, reducing down from the hour or so it would have taken a human investment banker previously. We simply programmed an algorithm to replicate what a human trader used to do.

  The shifts in capability here really centre around the principle that we are no longer coding a set of rules in an IF-THEN-THAT type syntax into computer code. We build algorithms, databases and learning engines that can observe behaviour, and learn to act accordingly. We are building computers that learn. All we need to do at that point is feed in the data—of which we have plenty. Just ask Facebook.

  AI will essentially evolve through three distinct phases4:

  •Algos and Machine Intelligence—Rudimentary machine intelligence or algorithm-based cognition that replaces some elements of human thinking, decision-making or processing for specific tasks. Neural networks or algorithms that can make human-equivalent decisions for very specific functions, and perform better than humans on a benchmark basis. This does not prohibit the intelligence from having machine learning or cognition capabilities so that it can learn new tasks or process new information outside of its initial programming. In fact, many machine intelligences already have this capability. Examples include: Google self-driving car, high-frequency trading (HFT) algorithms, facial recognition software, insurance assessor apps using image recognition, and credit risk assessment algorithms (e.g. sesame credit).

  •Artificial General Intelligence—A human-equivalent machine intelligence and learning system that not only passes the Turing Test and responds as a human would, but can also mimic human decision-making. It will likely also process non-logic or informational cues such as emotion, tone of voice, facial expression and nuances that currently a living intelligence could (can your dog tell if you are angry or sad?). Essentially, such an AI would be capable of successfully performing any intellectual task that a human being could.

  Examples include: Sophia (Hanson Robotics) and Singularity.io5.

  •Hyperintelligence (Strong AI)—A machine intelligence or collection of strong machine intelligences (what do you call a group of AIs?) that have surpassed human intelligence on an individual or collective basis to the extent that they can understand and process concepts that humans cannot understand.

  We don’t need to wait another 10, 15 or 30 years to see this happen, and the Turing Test is fairly meaningless as a measure of the ability of machine intelligence to disrupt a bank in terms of its day-to-day operations.

  The range of impact of Artificial Intelligence is going to be broad. IBM’s developerWorks team has an excellent primer on the advancements that have been made in Artificial Intelligence over the years, and how these are classified by the industry6. Terminology like cognitive computing, machine intelligence and artificial intelligence are not interchangeable, but do relate to the broader developments in AI that we’re seeing evolve today.

  Figure 2: Various AI disciplines as applied to financial services.

  To simplify the chart opposite, essentially there are two broad areas where AI will impact financial services. These are the interaction/conversational AI layer between the customer and the institution, and internally from a process perspective—anywhere we currently have a human checklist, a transaction or activity against compliance, risk or credit assessment rules, wherever we take instructions and apply those to an application, buy or sell order, or wherever we have a legal or contractual relationship to execute against. Any process a human can learn within a bank that doesn’t require strong dependency on social cues, an algorithm will be able to learn and replace that human in short order.

  AI will massively affect marketing; it will radically change customer service expectations; it will dominate our ability to engage customers on a behavioural basis; it will replace huge swathes of process-driven jobs; and will revolutionize the way we view and operationalize risk in the organisation today. In fact, just taking that last element, it is entirely possible that risk management in financial services will become the exclusive domain of AI within the next 10 years. But this is not going to happen from within an AI department in the bank, not even from the IT department. This is a systemic attack on the core of what we consider the operational engine of modern financial services today.

  This may sound like hype, but the worst case is that banks have three-to-five years before they have to start firing staff because of AI’s impact; and the best case is 7–10 years. In January 2017, a McKinsey & Co
mpany study found that about 30 percent of tasks in 60 percent of occupations could be computerized; while last year, the Bank of England’s chief economist said that 80 million US jobs and 15 million UK jobs might be taken over by robots7.

  Of course, not all jobs are created equally. In 2013, a highly-cited study by Oxford University academics, called “The Future of Employment”8, examined 702 common occupations and found that some finance jobs—bank tellers, loan officers, tax preparers and insurance claim assessors—are more at risk than others, including economists, financial analysts, financial modellers and statisticians.

  Deep learning: How computers mimic the human brain

  Central to the revolution in artificial intelligence is not computers that are programmed, but computers that learn. But how do computers learn?

  It all comes back to processing inputs (data) and mimicking neurons or the brain. In The Economist of 6 May 2017, data was characterized as the new oil of the emerging digital economy. Well, if data is the crude oil equivalent, databases, blockchains and data warehouses are the drilling rigs, and deep learning is the refinery that turns that oil into other useful products. Deep learning is at the heart of the emerging AI boom.

  Deep learning neural networks have been architected to use the same basic learning principles that occur in the human brain. The human brain consists of special cells called neurons, which are composed of several parts, including brain fibres known as dendrites. As you learn, these brain fibres grow. The fibres connect your brain cells to one another at contact points called synapses. The larger your brain fibres grow, and the more brain cells they connect, the more information can be stored in your brain. When you reinforce learning over time or practice skills you’ve learned, the dendrites in your brain grow stronger, forming a fatty tissue layer and doubling connections between key neurons or memories.