EU Monitor

Global financial markets

Artificial intelligence in banking

June 4, 2019 A lever for profitability with limited implementation to date
Author Artificial intelligence (AI) is a significant step forward in the digitalisation and
Orçun Kaya transformation of modern businesses. In short, it refers to computers’ capability
+49 69 910-31732
[email protected] to acquire and apply knowledge without programmers’ intervention.
Editor Investors are lining up to be part of the imminent change. AI attracted USD 24
Jan Schildbach bn in investments globally in 2018, a twelvefold increase since 2013. US start-
Deutsche Bank AG ups received the most attention, followed by Chinese, which already outpaced
Deutsche Bank Research European AI start-ups.
Frankfurt am Main
Germany Within Europe, Germany, France and the UK are the frontrunners in
E-mail: [email protected] experimentation and in the implementation of AI. In light of intensified global
Fax: +49 69 910-31877
competition, the European Commission proposed a EUR 9 bn budget to fund
www.dbresearch.com AI-related projects between 2021 and 2027.
DB Research Management
Stefan Schneider
Similar to earlier examples of information technology (IT) implementation in
financial services, AI promises great efficiency gains and potential revenue
increases. To date though, AI implementation in banking has been modest. AI is
being tested for real-time identification and prevention of fraud in online banking
as well as in know-your-customer (KYC) processes. Robo-advisors are also
evolving over time to become true AI solutions. Looking forward, regulatory
measures around data privacy and concerns regarding cybersecurity might
create obstacles to AI use in banking. In addition, the highly regulated nature of
banking may cancel out some efficiency gains of AI.
AI’s potential contribution to bank profitability should not be underestimated.
Empirically, AI has a significant positive impact on European banks’ return on
assets (ROA). By increasing labour productivity, AI technologies could
structurally reduce costs in the banking sector. Rapid implementation of AI
technologies is, therefore, central to fighting persistently weak profitability and to
remaining competitive.
Artificial intelligence in banking

Introduction

Huge progress in computer hardware, software and internet technologies have

irreversibly changed our societies. It is now difficult to imagine an economic
agent without computers, internet or mobile devices. The pace at which IT is
evolving offers great opportunities to expand the client base, introduce new
products or improve existing ones and to increase efficiency in a relatively short
period of time. On the other hand, if companies miss out on the current IT wave,
they might be overtaken by events soon.
Among the various IT breakthroughs of recent years, the advancement in AI is
particularly remarkable. In short, AI refers to computers having cognitive skills
similar to humans, which could result in immense efficiency gains for firms and
their clients alike. The financial sector has been one of the early experimenters
with AI technologies, not least due to its likely contribution to stronger
profitability. It is therefore essential to take a closer look at the potential role of
AI in banks’ digital transformation.

Artificial intelligence: A giant step beyond standard IT applications

To date, IT solutions in the business world have by and large focused on

automating repetitive tasks that would otherwise require human involvement.
Machine learning 1
The boundaries for these IT applications have been set by their developers, and
There are various approaches to programming
computers so that they mimic human decision-
by design these solutions have been limited in their capabilities. They have
making. Decision trees, ranking or prioritising largely been static and unable to comprehend or act on their own. With
are among the more established solutions. A technology evolving rapidly, however, this is increasingly changing.
relatively new approach is machine learning
(ML). ML is a subset of AI and refers to Artificial intelligence refers to the ability of computer programs to acquire and
computer programs that recognise patterns apply knowledge without human intervention and involvement. By observing the
and make predictions based on them. Typical
world around them and analysing information autonomously, AI systems draw
examples are internet platforms that
recommend particular products or news stories conclusions and take appropriate actions. They learn from their previous
to users who might like them based on judgements and, depending on the level of accuracy, improve their performance
previous preferences. By continuously over time.
analysing new data and scenarios, ML tools
make adjustments to decision-making AI as a term was first coined at the Dartmouth Conference in 19561 and is not
processes without being specifically new per se. In recent years, though, some breakthroughs in IT have allowed
programmed to do so. They are therefore able
to learn from data. Subcategories of ML are tremendous momentum in AI’s capabilities:
deep learning, as well as supervised,
unsupervised and reinforcement ML.
i) The expansion of internet usage has led to huge amounts of digital
ML tools process vast quantities of data
information being generated and stored. In about 10 years, the amount of
through neural networks. In short, these are data generated worldwide grew some 17 times. Forecasts point to another
processes which classify data on successive fivefold increase between now and 2025. This large volume of information,
layers. In doing so, they rely on the once cleaned and structured (i.e. big data), is at the core of data-driven
probabilities of possible outcomes. They make
decisions based on the most likely outcome,
decision-making.
even though it might turn out not be the perfect
ii) There has been a colossal increase in the processing power of computers. A
choice in the end. However, neural networks
involve a feedback loop. Depending on the standard measure of that, the number of transistors, has increased 10 m
accuracy of the outcome from previous trials, times since the 1970s. The speed of central processing units, another
they update their approach to perform better element contributing to processing power, rose by a factor of 6,750 over the
the next time.
same period.2 This enables algorithms to process information at much faster
Source: Deutsche Bank Research
rates and contributes to the accuracy of their decision-making.
iii) Other developments – such as the reduction in data storage costs,
advancements in data mining processes or an increasing number of IT
experts – have further fuelled the feasibility and capability of AI. While the

1
See McCarthy, Minsky, Rochester, Shannon (1955).
2
Moore’s Law states that the processor speed of computers, or the number of transistors on an
affordable central processing unit, doubles approximately every two years.

2 | June 4, 2019 EU Monitor

Artificial intelligence in banking

hard drive cost per gigabyte has come down from around USD 5,000 in 1990
to some USD 0.025 today, the number of IT specialists grew by 50% in the
euro area between 2007 and 2017, for example.
Big data as input, data identification methods such as machine learning and the
greater affordability of these tools have been the driving factors behind AI’s
recent rapid success in understanding languages, recognising objects and
sounds, and observing and solving problems autonomously.

Artificial intelligence investments on the rise

Thanks to its rapid evolvement in recent years, AI is being experimented with

Strong growth in VC investments in AI
start-ups globally 2 and implemented in several areas. Due to measurement issues, however,
quantifying its deployment is hardly a straightforward task. Indeed, firms might
USD bn
deploy AI to increase efficiency in their processes, which is not directly
25 observable for analysis. Moreover, it is sometimes difficult to differentiate
between more standard IT solutions and sole AI applications. To partly
20
overcome these drawbacks, information on venture capital (VC) investments in
15 AI start-up firms may be useful.3 In 2018, AI start-ups received a staggering
USD 24 bn globally, up from less than USD 2 bn in 2013. Growth in VC
10
investments over the past two to three years has been particularly strong. AI
5 firms have also increasingly become acquisition targets. Over the last 20 years,
a total of 434 companies in the AI sector have been acquired, 220 of them since
0 2016 alone.4
13 14 15 16 17 18*
*estimated Of the total VC volume in 2018, almost USD 15 bn went to AI start-ups in the
Sources: OECD, Deutsche Bank Research
US, and another USD 6.5 bn went to Chinese firms. In 2017 and 2018, the
number of VC deals flattened out. Yet the average volume of VC investments
surged, an indication of VC flowing into more mature AI firms whose capital
Patents in AI: China outpacing the EU
needs are larger than those of typical seed stage start-ups. In China, for
3
example, SenseTime Group, a computer vision and deep learning technology
% of total global patents in AI* developer, raised USD 1.6 bn in VC funding in 2018. With the new capital, the
40 value of the company rose to over USD 6 bn, making it the world’s most
valuable AI unicorn. In the US meanwhile, it is primarily large tech firms which
30 invest in AI start-ups.
20 To VC investors, AI appears to be a truly transformative technology with
significant potential, like the internet and mobile revolutions in past decades.
10 How do AI start-ups use the funds they receive? First observations indicate that
they hire new AI talent (which proves to be costly and difficult to find) and
0
10 15 10 15 10 15 10 15 expand their services. Investors might therefore need to wait a while before they
US EU-28 JP CN see meaningful returns on their investments.
*AI-related patents are patents in cognition, meaning and
understanding, large-capacity and high-speed data storage,
high-speed computing, large-capacity information analysis.
Artificial intelligence and intellectual property rights
Sources: OECD, Deutsche Bank Research

A technological field usually is more useful and has a greater value for the
economy in the years to come if there is a substantial increase in the number of
patents filed in this particular field.5 There were some 20,000 patent applications
in AI-related technologies in 2016, double the figure of 2010. Around 50% of
that was accounted for by AI patents in computer vision. This technology is
mostly used in self-driving cars and shows how intense competition in this area
currently is. Of the total AI patent applications in 2015 (the latest available data

3
The OECD identifies AI start-ups as firms whose business model focuses on i) “artificial
intelligence”, “machine learning” and “machine intelligence”; ii) “neural networks”, “deep learning”,
and “reinforcement learning”; and iii) “computer vision”, “predictive analytics”, “natural language
processing”, “autonomous vehicles”, “intelligent systems” and “virtual assistant”.
4
See WIPO (2019).
5
See Inaba and Squicciarini (2017) for a detailed explanation.

3 | June 4, 2019 EU Monitor

Artificial intelligence in banking

with respect to countries), the US accounted for about one-third, a more or less
stable share since 2010. Within the US, it was the tech giants who filed the
largest number of AI patents. China made up 25% of the applications in 2015,
up from 10% in 2010. Japan and the EU-28 each had a share of 14%, both
AI activity concentrated in a small set of
EU countries
down from around 20%. China increasingly seems to be replacing the EU and
4
Japan in AI research and development with potentially significant implications in
% of total AI patent applications in the EU between the future.
2010 and 2015
Within the EU, half of all AI patent applications originated in Germany and
France. Together with the UK (16%) and Sweden (8%), four countries account
21
31 for the lion’s share. Given that patents create a legal monopoly, they introduce
important first-mover advantages. Considering potentially large economies of
7
scale, countries unable to implement AI now might be at risk of remaining
8 behind for a long period of time.
17
16 In light of intensified global competition in IT in general and AI in particular, the
European Commission proposed a budget to fund research and innovation
DE FR UK SE NL Other EU
projects in Europe in March 2019. Horizon Europe is the successor of Horizon
2020, which has a volume of EUR 77 bn to be spent between 2014 and 2020.
Sources: OECD, Deutsche Bank Research
Horizon Europe aims to allocate EUR 100 bn between 2021 and 2027. One of
the main sub-categories of Horizon Europe is the Digital Europe Programme,
which aims to invest EUR 9 bn specifically in high-performance computing and
ATM to bank branch ratio in Europe* 5
data, AI, cybersecurity and advanced digital skills projects. Even though Horizon
4 Europe represents an important step forward in enhancing AI technology in
Europe, its ability to drive successful AI projects remains to be seen. Indeed, its
predecessor received 115,000 innovation and research proposals between
3
2014 and 2016, yet only 14,000 proposals were selected for funding, a very low
success rate. The high rate of over-subscription is evidence of strong demand
2
for funding. But the large number of rejected applications points to some
underlying problems. Alternative solutions, such as enhancing IT literacy at early
1 ages or improving IT infrastructure, might be necessary to increase the number
of high-quality AI and innovation projects.
0
87 91 95 99 03 05 09 13 17

*data includes Finland, Norway, Denmark, Sweden, Belgium,

Earlier examples of IT implementation in banking
Spain, the Netherlands, Switzerland, Italy, the UK, France,
and Germany till 1999 and all EU countries later on.
Banks are usually early adopters of IT opportunities. This is true not just for the
Sources: World Bank, Humphrey et al (2003), Deutsche back office, where modern technologies have been used for a long time (e.g. to
Bank Research
R
process payments), but also for the front-end. An example are automated teller
machines (ATMs), one of the earliest IT applications in banking. These devices
Internet banking shows persistent growth 6 replaced the repetitive tasks of bank employees in cash withdrawal and account
balance checks. They made it easier for clients to access standard banking
Users in % of adult population
services while also making banks more efficient. Since the first ATM was
80 installed in London in 1967, they have become standard devices in branches. In
Europe, their numbers have grown to three ATMs per bank branch in 2017, up
60 from one ATM per four bank branches in 1987. With bank employees relieved of
routine cash-handling tasks, they were able to take on other services such as
40 relationship banking (i.e. catering to clients’ individual needs) and offering other
bank services like credit cards, loans and investment products.6
20
Online banking is another example of banks adopting client-facing new IT.
0 Starting from the late 1990s, the use of internet as a medium for banking
08 10 12 14 16 18 services has increased immensely. Direct or internet banks with very few or no
EU DE FR UK physical branches emerged. Virtually all banks started providing online banking
services. In 2018, more than half of the adult population in the EU used internet
Source: Eurostat banking to check their account balances or transfer funds. In some countries,
such as Denmark, internet banking penetration rates are particularly high (90%).
In Germany, 59% of individuals used internet banking in 2018, up from only

6
See Autor (2015).

4 | June 4, 2019 EU Monitor

Artificial intelligence in banking

35% in 2007. For clients who have little time to visit a branch, online banking
has become the main tool for standard services.

Mobile banking in Germany

The way that bank clients access the internet has changed as well. Germans,
7
for example, are increasingly using their mobile devices for internet banking,
% of respondents and some 40% of them have a banking app on their mobile phones. Moreover,
60 one-fifth of them also use their apps for mobile payment services. This is
particularly popular among younger, more educated and internet-savvy
40 individuals. With banks and their clients meeting on virtual platforms and ever
more people using online services, banking is becoming less and less branch-
20 dependent.

0
User Non-user User Potential Non-user Artificial intelligence in banking
user
Banking app Mobile payments For banks, data is essential to almost all business lines, from traditional deposit
Sources: Postbank Digital Study (2018), Deutsche Bank taking and lending to investment banking and asset management. Autonomous
Research
data management without human involvement therefore offers great
opportunities for banks to improve speed, accuracy and efficiency. Potential AI
applications in banking can be classified into four broad categories: 1)
customer-focused front office applications, 2) operations-focused back office
applications, 3) trading and portfolio management, 4) regulatory compliance.7 At
least for now, banks by and large are still experimenting with AI technologies
rather than fully implementing them in their processes. Customer- and
operations-focused AI solutions seem to be undergoing more intensive
exploration than others:
i) AI is being tested for real-time identification and prevention of fraud in online
banking. Indeed, credit card fraud has become one of the most prevalent
forms of cybercrime in recent years, which is exacerbated by the strong
growth in online and mobile payments.8 To identify fraudulent activity, AI
algorithms check the plausibility of clients’ credit card transactions in real
time and compare new transactions with previous amounts and locations. AI
AI implementation in banking 8 blocks transactions if it sees risks.
-Credit scoring
ii) AI is also being tested in KYC processes to verify the identity of clients. AI
Customer- -Insurance policies
algorithms scan client documents and evaluate the reliability of the
focused -Client-facing chatbots
information provided by comparing it with information from the internet. If AI
-Know your customer
algorithms identify inconsistencies, they raise a red flag and a more detailed
-Capital optimisation KYC check by bank employees is performed.
Operations- -Model risk management
focused -Stress testing
iii) Another area where banks are experimenting with AI technologies is
-Fraud detection chatbots. Chatbots are digital assistants that interact with clients by text or
voice and aim to address their requests without the involvement of a bank
Trading and
-Trade execution employee.
portfolio
-Portfolio management
management iv) Banks are also exploring AI to visualise information from legal documents or
annual reports, for example, and to extract important clauses. AI tools create
-Regulatory technology
models autonomously after observing the data and back testing to learn from
Regulatory -Macroprudential surveillance
their previous mistakes to improve accuracy.
compliance -Data quality assurance
-Supervisory technology v) Some existing financial technology tools evolve as true AI solutions over
Sources: FSB (2017), Deut sche Bank Research time, too. Good examples include robo-advisors that enable full automation
in certain asset management services and online financial planning tools that
help customers make more informed consumption and saving decisions. As
these financial technology solutions mature, they increasingly use techniques
that search data and find patterns in them autonomously.

7
For a detailed overview on how AI is being implemented in the individual categories of table 8,
please see FSB (2017).
8
See Mai (2018).

5 | June 4, 2019 EU Monitor

Artificial intelligence in banking

In their quest to become more efficient, banks mostly seem to be exploring AI

applications to replace activities which are costly, laborious and repetitive. The
focus is on operational risk management gains like fraud detection or improved
KYC and on opportunities for cost reduction like chatbots or robo-advisors.
A real-life application of AI: Deutsche Bank’s Alpha-Dig platform 9
Using AI and machine learning tools, it is possible to quantify geopolitical risk and predict its effect
on financial markets. For example, Deutsche Bank’s Alpha-Dig platform (Alpha-Dig) infers context
from news media, social media, and other natural language articles, and then builds a picture of a
country’s political risk profile.
Alpha-Dig first uses algorithms to mine global financial news as a proxy for how much media
attention there is towards certain countries’ risks. The process uses Natural Language Processing
and machine learning techniques to infer context in a news article and ensure that positive and
negative indicators are gleaned from it. As a second step Alpha-Dig overlays learnings from
Wikipedia whose articles are largely accurate and easily readable for machines. To adjust for
potential biases, the platform uses readership data to see what topics are trending. Once data from
the mainstream financial news is enhanced with learnings from Wikipedia, Alpha-Dig can create a
picture that shows how political issues have become more or less important over time. Among
other statistical methods, Z-scores can be calculated, which look at the average amount of daily
geopolitical news for a topic in the recent past and see what proportion of all geopolitical news is
consumed by that topic. If a particular political event is receiving attention that is greater than two
standard deviations more than normal, it is labelled an ‘outlier’ event. Obviously, no system can
accurately forecast geopolitical implications all the time. But via tools like Alpha-Dig, an objective
measure that can assist investors in what are notoriously difficult times can be created, thanks to
advancement in AI.
Source: Deutsche Bank Research. Konzept. January 2019, p. 34-39.

Impediments to the use of artificial intelligence in banking

Despite its immense potential, some external factors might slow down AI
implementation in banking. To begin with, the EU’s General Data Protection
Regulation (GDPR), which came into force in 2018, contains preventive clauses
on automated decision-making. This affects not only the financial industry but all
sectors in general. Article 22 of the GDPR states: “The data subject shall have
the right not to be subject to a decision based solely on automated processing,
including profiling...” This is particularly problematic for AI tools whose decision-
making by definition is solely automated. To overcome the restrictions under
Article 22, human involvement at some stage might be a solution. I.e. at the end
of the AI chain, the final decision could be given to humans. In addition, Article
13 of the GDPR involves disclosure provisions. For example, if an AI tool rejects
a bank account or loan application, the client has the right to know the logic
involved in this decision. Article 13 does not necessarily require the source code
of the AI algorithm to be revealed in detail. Yet some information on the input
parameters of the AI tool has to be disclosed. In any case, the intervention of
human programmers might be required in order to fully comply with these and
many other data privacy rules, a setback for the expected efficiency gains of AI.
Another likely impediment to AI use in banking is the potentially malicious
manipulation of big data. For example, hackers might try to flood systems with
fictitious data (fake social media accounts, websites, news) to influence AI
decision-making. As a result, AI tools might come up with biased decisions and
discriminate against certain individuals, or hackers could even take control of AI
systems. With AI systems being connected to each other, malevolent issues
might intensify. Even though AI itself has a relatively high level of accuracy in
detecting cyber-attacks and malware, the continuous surveillance and
monitoring of programmers might be necessary to address cybersecurity issues.
The introduction of regulatory sandboxes, where the safety of new AI tools is
tested in a real-world environment, might be beneficial in this context.

6 | June 4, 2019 EU Monitor

Artificial intelligence in banking

In the eyes of some observers, AI – and especially neural networks – have

opaque reasoning and function as black boxes.9 These concerns arise from the
sometimes complex AI algorithms and the inability of humans to visualise and
understand these patterns. What aggravates the complexity problem is the fact
that AI algorithms update themselves over time and become more connected. It
is important to remember that AI predictions and decisions might be very close
to those of humans in the end. Unlike humans though, AI by its nature is unable
to communicate its reasoning. This complicates the use of AI, considering that
processes in banking have to be fully traceable backwards even if the decisions
made are reasonable and justified. If there is a problem with a decision, it needs
to be clearly detectable at which step the error has occurred. The entire
decision-making process has to be compliant with regulatory and supervisory
rules and fully transparent. While it may also cancel out some efficiency gains,
the involvement of human programmers and overseers might be a solution to
reduce issues around the opaqueness of some AI algorithms. Despite these
potential impediments, banks are dedicated to experimenting with AI, which
could have significant profitability implications.
AI and bank profitability: An almost linear
relation 10
2010-15 averages, ROA in % on y axis, Artificial intelligence and bank profitability
number of AI patents normalised* on x axis

0.6 AI might contribute to bank profitability in two ways: First, by taking over
0.5
repetitive tasks from bank employees, autonomous AI software could reduce the
SE
demand for less-skilled labour and improve the efficiency of remaining bank
0.4
staff. This is crucial, as employee compensation usually represents a large
0.3 BE FR
ES NL share of banks’ cost base. Second, AI implementation could also contribute to
0.2 AT UK
DK revenue generation. For example, it might help banks to develop new products
0.1 DE
and offer tailor-made products better suited to client preferences. Nevertheless,
0 IT quantifying the link between the use of AI and bank profitability is hard, not least
-0.1 due to issues around identification and a lack of micro data. From an aggregate
20 30 40 50 60 70
angle, though, AI patent applications10 and ROA in the banking sector in
*Per 1,000 patent applications
European countries show an almost linear relation, with a correlation of 80%.
Sources: ECB, OECD, Deutsche Bank Research Banks seem to be more profitable in countries where the level of AI patent
activity is higher. At the same time, it is somewhat difficult at this early stage of
AI diffusion in banking to assess the potential operational risks and associated
What explains banks' ROA in Europe? 11
costs for banks which might stem from the increased use of AI.
% of variation in ROA explained by...
There are various factors that determine bank profitability which need to be
7 accounted for in an empirical setting. In our panel regression, we control for
macro and banking sector-specific indicators, as well as time-fixed effects, in
29
addition to AI. In our sample, we covered data from ten EU countries and a time
64 span from 2010 to 2015. The dependent variable of our analysis is the ROA of
the banking sector in individual countries. Our results reveal that macro factors
such as GDP growth and inflation are the most important contributors to bank
Macro factors
profitability. These explain two-thirds of the variation in bank profitability.
Banking-sector indicators Banking sector indicators such as cost-income ratio, equity-to-assets ratio and
AI implementation proxy non-performing loans explain some 30% of the variation in ROA. AI patents
positively impact ROA at statistically significant levels and explain 7% of the
Results of panel regression. The dependent variable is the
biannual ROA of the banking sector. Explanatory variables
variation in bank profitability. It is important to remember that there are large
are GDP growth, inflation, cost-income ratio, non-performing overlaps between standard IT solutions and sole AI applications and patents.
loans, equity-to-assets ratio, total assets of banks, share of AI
patents and time-fixed effects. The time dimension of the For example, large-capacity and high-speed data storage, as well as high-speed
analysis is from 2010 to 2015. The countries included are
Austria, Belgium, Denmark, Germany, France, Italy, Spain, computing patents, have a broader impact far beyond AI. In this vein, it may be
Sweden, the Netherlands and the UK. fair to argue that bank profitability is positively related to stronger use of AI in
Source: Deutsche Bank Research specific and stronger use of IT in general. Obviously, the causality could also
work the other way around, with more profitable banks investing more in AI.

9
See Bathaee (2018) for a detailed explanation.
10
We normalise the number of AI-related patents (i.e. divide them by the total number of patents in
a given country) to eliminate potential outlier effects, such as an overall increase in patent activity.

7 | June 4, 2019 EU Monitor

Artificial intelligence in banking

However, in banking, IT implementation may be largely driven by (customer)

demand and not necessarily by supply. Moreover, strong competition makes
modern technology a priority even for those banks that are not profitable. The
fact that AI impacts bank profitability means it might help European banks to
address one of their core problems of recent years: persistently weak
profitability.11 By increasing labour productivity, AI technologies could
structurally reduce costs in the banking sector.

Concluding remarks

AI has the potential to revolutionise several aspects of our everyday life.

Especially in the US and China, AI has attracted large investments in recent
years and is increasingly being implemented. In Europe, the picture is
somewhat mixed, with a couple of countries having an active AI landscape while
others lag behind. Being aware of its potential, European policymakers have
introduced measures to increase AI activity in Europe. AI also has the potential
to fundamentally change aspects of financial services. To date, though,
implementation in banking has been modest. Looking forward, regulatory
measures around data privacy and the highly regulated nature of banking might
create obstacles to AI implementation. Still, AI’s potential contribution to bank
profitability should not be underestimated. In an environment where competition
in banking is growing ever more intense – thanks to data-driven financial
services providers such as financial technology (FinTech) start-ups and large
technology firms that are challenging traditional banking business models –
rapid implementation of AI technologies might be pivotal for banks to remain
competitive.
Orçun Kaya (+49 69 910-31732, [email protected])

11
See e.g. Schildbach (2017).

8 | June 4, 2019 EU Monitor

Artificial intelligence in banking

Literature

Autor, David H. (2015). Why Are There Still So Many Jobs? The History and
Future of Workplace Automation. Journal of Economic Perspectives, 29 (3), 3-
30.
Bathaee, Yavar (2018). The Artificial Intelligence Black Box and the Failure of
Intent and Causation. Harvard Journal of Law & Technology, 31 (2), 889-938.
FSB (2017). Artificial intelligence and machine learning in financial services:
Market developments and financial stability implications.
Humphrey, David B., Magnus Willesson, Göran Bergendahl and Ted Lindblom
(2003). Cost Savings from Electronic Payments and ATMs in Europe. FRB of
Philadelphia Working Paper No. 03-16.
Inaba, Takashi and Mariagrazia Squicciarini (2017). ICT: A new taxonomy
based on the international patent classification. OECD Science, Technology and
Industry Working Papers, 2017/01. OECD Publishing, Paris.
Mai, Heike (2018). Card fraud in Germany: Few incidents, but high costs.
Deutsche Bank Research. Talking Point.
McCarthy, John, Marvin L. Minsky, Nathaniel Rochester and Claude E. Shannon
(1955). A proposal for the Dartmouth summer research project on artificial
intelligence.
Postbank Digital Study (2018). Der digitale Deutsche und das Geld.
Schildbach, Jan (2017). Where do European banks stand? 10 years after the
start of the financial crisis. Deutsche Bank Research. EU Monitor.
WIPO (2019). Artificial Intelligence. WIPO Technology Trends.

© Copyright 2019. Deutsche Bank AG, Deutsche Bank Research, 60262 Frankfurt am Main, Germany. All rights reserved. When quoting please cite
“Deutsche Bank Research”.
The above information does not constitute the provision of investment, legal or tax advice. Any views expressed reflect the current views of the author,
which do not necessarily correspond to the opinions of Deutsche Bank AG or its affiliates. Opinions expressed may change without notice. Opinions
expressed may differ from views set out in other documents, including research, published by Deutsche Bank. The above information is provided for
informational purposes only and without any obligation, whether contractual or otherwise. No warranty or representation is made as to the correctness,
completeness and accuracy of the information given or the assessments made.
In Germany this information is approved and/or communicated by Deutsche Bank AG Frankfurt, licensed to carry on banking business and to provide
financial services under the supervision of the European Central Bank (ECB) and the German Federal Financial Supervisory Authority (BaFin). In the
United Kingdom this information is approved and/or communicated by Deutsche Bank AG, London Branch, a member of the London Stock Exchange,
authorized by UK’s Prudential Regulation Authority (PRA) and subject to limited regulation by the UK’s Financial Conduct Authority (FCA) (under
number 150018) and by the PRA. This information is distributed in Hong Kong by Deutsche Bank AG, Hong Kong Branch, in Korea by Deutsche
Securities Korea Co. and in Singapore by Deutsche Bank AG, Singapore Branch. In Japan this information is approved and/or distributed by Deutsche
Securities Inc. In Australia, retail clients should obtain a copy of a Product Disclosure Statement (PDS) relating to any financial product referred to in this
report and consider the PDS before making any decision about whether to acquire the product.

ISSN (Online): 1612-0280

9 | June 4, 2019 EU Monitor