AML POLICY
customer relationships should be reviewed
holistically across lines of business, with a
comprehensive approach to quantifying
BSA/AML risk for new and existing custom-
ers, and that “the quantification of risk
[should] encompass a customer’s entire
relationship.” 11
Consequently, a centralized, yet intercon-
nected approach that provides a holistic
view and governance model (e.g., one that
consolidates a function into a central utility
while maintaining communication and
awareness with respective business lines),
allows for better identification and manage-
ment of risk, enhanced communication, and
a setting more conducive to internal and
external testing and evaluation.
Investment in Advanced Technology
With the growth of technology rapidly
increasing, FIs have been employing tech-
niques such as advanced analytics and
automation to assist with performing key
AML functions (e.g., know your customer
activities, transaction monitoring, list
screening). While these strategies continue
to evolve and work well with structured
data sets (e.g., information organized into
specific fields or formats) and prescriptive
functions (e.g., defined rules and scenar-
ios), FIs are also looking at bolder technol-
ogy solutions, such as AI (referred to herein
interchangeably with the terms “cognitive
technology” and “cognitive computing”), to
tackle more complex and analytical areas,
including unstructured data sets (e.g.,
unorganized or non-text information).
Cognitive computing refers to the simula-
tion of human thought processes, using
self-learning systems that include data
mining, pattern recognition and natural lan-
guage processing to mimic the way the
human brain works. 12 This allows a
computer to derive conclusions and com-
plete activities that may require fundamen-
tal human skills and intelligence, such as
looking, reading, writing and integrating
knowledge. Machine learning, which works
by using algorithms that iteratively learn
and adapt in an automated manner as the
models are exposed to new data, is a
method used to enable a computer to learn
without being programmed. 13
One attractive advantage of cognitive tech-
nology is the ability to access, make sense
of, and/or use large, unstructured and
diverse information, including non-tradi-
tional information such as social media
information (e.g., posts, images) and audio
(e.g., telephone conversations) that may be
available in other parts of an FI (e.g., the
marketing department or customer service)
for AML purposes. A more ambitious goal is
the potential to be predictive, rather than
reactive, in identifying fraudulent activity,
such as by recognizing traces of the behav-
ior before it fully occurs through enabling
computers to learn, comprehend and detect
new money laundering schemes and
characteristics.
A glimpse at how cognitive
technology is being applied
within AML
On February 15, 2011, an IBM-developed
supercomputer with AI, named Watson,
capable of answering questions in natural
language, won the first place prize of $1
million on the quiz show, Jeopardy. 14
Five years later, on March 8, 2016, KPMG
LLP, a “Big Four” audit, tax and advisory
firm, revealed plans to leverage IBM’s cog-
nitive computing technology explaining
that “[a]uditing and similar knowledge ser-
vices are increasingly challenged with
tackling immense volumes of unstructured
data. Cognitive technologies such as
Watson can transform how this data is
understood and how critical decisions are
made.” For example, “cognitive technology
is further advancing improvements to sam-
pling processes, in which auditors review
subsets of data to analyze thousands or mil-
lions of actions to draw conclusions.” 15
On November 22, 2016, it was announced
that Promontory Financial Group, a risk
management and regulatory compliance
consulting firm with a global AML and
counter-terrorist financing practice, was
acquired by IBM. Promontory’s profession-
als intend to train Watson via machine
learning. For instance, Watson will learn by
“continuously ingesting regulatory infor-
mation as it is created and through interac-
tion in real-world applications. This
includes solutions for tracking evolving reg-
ulatory obligations, expectations and con-
trol requirements, as well as solutions that
address specific compliance needs, such as
financial risk modeling, surveillance and
insider threat, and financial crimes includ-
ing counter fraud, [AML] and [KYC].” 16
In addition, a 2016 research report from
Celent (a research and advisory firm focused
on business and technology strategies)
examined AI applications, vendor profiles,
and application trends, and noted that AI
“technologies are increasingly being applied
in the banking industry, mainly toward
knowledge management, identity authenti-
cation,…anti-money laundering, and risk
control.” In reference to AI, the report fur-
ther notes that “[w]ith their ability to fully
understand the market, customers, and reg-
ulatory changes through data, banks are in
the best position to apply these
technologies...” 17
“Consent Order,” U.S. Department of Treasury Comptroller of the Currency,” 2015, www.occ.gov/static/enforcement-actions/ea2015-113.pdf
11
“Cognitive Computing,” http://whatis.techtarget.com/definition/cognitive-computing
12
“Machine Learning,” SAS, http://www.sas.com/en_us/insights/analytics/machine-learning.html
13
John Markoff, “Computer Wins on ‘Jeopardy!’: Trivial, It’s Not,” N