Co-authored by Ana Carolina Cruz Aguilar and Mile Dragosavac.
Anti-money laundering (AML) is a complex and regulated field involving composite data and intricate workflows. With tighter regulations and a prevailing reliance on manual processes, the heat is on for banks to get their risk management acts together. Artificial intelligence (AI) and automation can play a key role in transforming this sector.
In this two-part blog, we will understand the current state of this sector, learn more about machine learning (ML) and robotic process automation (RPA), and finally, study the applications and value additions of these technologies to the know your customer (KYC) area.
The Need for AI and Automation in the KYC/ AML Sector
Banks are struggling to comply with regulations for identifying and verifying clients such as the Fifth Anti-Money Laundering Directive, BSA, FinCEN and the FINRA Rule 3310 coming into force in the EU and U.S. Additionally, banks rely on manual processes adversely affecting timelines that can run into several weeks.
‘A recent study confirms that more than 51% of financial institutions regulate KYC and AML processes manually.1’
Financial institutions cope by engaging a large number of analysts to review and investigate suspected cases of money laundering. In case of a corporate client, it entails probing through a web of transactions and paperwork to analyze the relationship between companies and their ultimate beneficiary owners, to flag unusual patterns of behavior. Needless to say, this is expensive and time-consuming.
‘AI, when applied to workflow automation of KYC/AML processes can decrease the duration of control processes, minimize human errors and significantly reduce costs.’
Despite cynicism, AI need not replace human interventions but can augment human intellect. In order to recognize how AI can help in this sector, let’s begin by understanding these technologies.
Machine Learning
While companies understand the need for AI, the surrounding buzz has diluted the actual understanding of its applications. The confusion and the skepticism comes from the use of “artificial intelligence” as a generic and often interchangeable term for automation, machine learning, deep learning and RPA.
‘ML is a sub-set of AI that learns from repeat experiments to improve overall performance.’
ML does this by adapting a particular mathematical function of data using a metric for performance measurement. Once a suitable algorithm has been selected by experts, the parameters it contains must be selected properly. The function and metric are linked so that there is a relationship between the choice of parameter and the quality of the models estimation. By successively changing the parameters, ML constantly improves the metrics under the appropriate mathematical conditions until an optimum is reached (i.e. when a change does not lead to an improvement). Thus the machine has learned to select the best parameter values iteratively.
‘Due to its iterative approach ML can process millions of data sets without error.’
In the context of KYC/AML, the transactions that the ML model considers suspicious must be confirmed or rejected by experts, giving them control.
With the help of deep learning or neural networks (a sub-set of machine learning), extraordinarily results have been achieved recently. Deep learning in particular has become a dominant approach to the recognition of images, speech and text content. Other applications, such as robotic process automation, do not originally include artificial intelligence. If this is the case, however, we are talking about cognitive RPA. This would bring the field RPA graphically closer to the middle.
Machine learning:
|
Robotic Process Automation (RPA)
We have defined AI and ML but how are these two terms related to RPA?
‘RPA is the automation between different applications of repetitive tasks, previously carried but by humans, but with greater efficiency and accuracy.’
There are a staggering number of RPA tools available on the market today, making the access to this technology easier than ever. Some RPA tools are evolving to cover hitherto Machine Learning functionalities, such as unstructured data processing (e.g., optical character recognition) Predictive analytics and some basic judgment-based automation are now standard components with bots even being delivered as cloud services.
From process standardization and re-engineering, together with some basic automation of existing processes, to collecting and structuring data in a systematic way, RPA allows data analytics capabilities and basic actions and recommendations. All this together with cognitive automation and industrialized application of complex machine learning algorithms makes decisions almost human-like.
The following diagram illustrates the different maturity stages of this automation journey. Becoming lean should set the base for automation by standardizing, re-designing and optimizing processes. Once this is achieved, collecting and structuring data becomes routine, enabling the execution of more complex processes which involve the application of ML algorithms, an area which is also known as cognitive RPA or cognitive automation in general.
RPA key benefits:
|
In the second part of this blog, we will learn how these technologies can be applied in the context of KYC and AML.
Want to learn more ? Our experts are on the front lines of this discussion every day with our clients, and can share some rich perspectives with you on how to approach your digital future.

Gerardo Salonia
Principal, Infosys Consulting
Gerardo Salonia is a principal within our financial services practice in Germany with a focus on compliance, AML and KYC areas. He has extensive consulting experience within the e-commerce and financial services domain. Gerardo specializes in disruptive technologies and large-scale digital transformation of organizations. He holds an MBA in business administration from the University of Mannheim and has a risk management certification from the Goethe Business School – Frankfurt University.

Ana Carolina Cruz Aguilar
Senior Consultant, Infosys Consulting
Ana Carolina Cruz Aguilar is senior consultant within our Artificial Intelligence & Automation practice at Infosys Consulting, Germany. Her focus lies in helping organizations leverage the full potential of the latest digital technologies. She also has extensive banking experience, mainly in the credit risk controlling area. Carolina holds an M.Sc. in Mathematical Sciences from the National University of Mexico and is certified Data Scientist from The Data Incubator Reply program.

Mile Dragosavac
Senior Consultant, Infosys Consulting
Mile Dragosavac is senior consultant within our Artificial Intelligence & Automation practice at Infosys Consulting, Germany. His focus lies in business intelligence topics leveraging data driven decisions using machine learning. He also has extensive banking experience, mainly in the credit risk controlling area and multi asset management. Mile holds an MBA in economics from the University of Mannheim and is certified applied Data Scientist from University of Michigan.