Commentaries and insightful analyses on the world of finance, technology and IT.

« July 2017 | Main | March 2018 »

September 18, 2017

Potential of AI in Fraud Detection

Technological innovation is the new normal in today's world. How many would have imagined a drone could be used for advanced pitch analysis in cricket matches, which has been tried out at the recently concluded Champions trophy in England. Though Artificial Intelligence (AI) has existed for decades yet it is being widely accepted and tried out in the Banking industry now, more than ever. In the Banking Industry, AI is already making inroads be it in Customer Service like introducing Chat-bots, or in AML, or be it Fraud Detection. AI has been and is helping Banks in being more proactive in their approach, like identifying a potential fraudulent transaction even before it happens.


What Potential does AI hold in fraud detection for Banks?

Banking is one industry that deals with a huge amount of data, which could be leveraged to do advanced analysis and come up with meaningful insights. There is a range of activities being carried out by banks in using AI for customer service, for personal financial services, etc. However, there is huge scope for banks to utilize AI techniques in fraud detection. For example, Banks have started using AI techniques in identifying the fraudulent transaction patterns in a card and use the data to prevent frauds. Banks could have a software which could raise a red flag if a customer has accessed his account from 7 to 8 different IP addresses in a span of a week, however the customer could be an artist/actor/tourist who could be doing shopping while he is travelling. So here, an AI software would be vital in looking and analyzing the spending pattern closely. The AI technique here makes the machine to think like a human. Another potential use of AI would be to analyze the user profile based on transactions done and then try to determine whether there is reasonable suspicion. This way the Banks can avoid a fraudulent transaction even before it occurs. Some banks have already started replacing passwords with voice recognition for some of their services which uses AI. Now, this can only be achieved if the system is fed with the historical data of the fraudulent patterns as well as the historical data of verified transactions. Recently MasterCard announced an AI fraud detection service, which helps the FI's reduce the false decline and increase the accuracy of real time approvals for genuine transactions. This would be a great relief for customers. Nothing is more frustrating for the customer than to have their transactions declined for no fault of theirs. The use of AI, which could analyze their spending pattern in order to identify fraudulent transactions, would be truly beneficial to the banks in better understanding their customers. Having said that, every technology comes with its own limitations and that should not deter the banks from trying out AI for fraud detection.


Banks that do not keep updating or taking steps to track suspicious transactions are at a greater risk. Banks have been slow in adopting AI for Anti-fraudulent activities. And it is just a matter of time before more banks adopt the AI techniques for not just fraud detection but aversion too.


September 14, 2017

Future of ADM in the Digital World

Background - Technologies Enabling Digital Transformation


There is a lot happening in areas of robotics, Artificial Intelligence, machine learning, and IOT as businesses turn digital and there is much discussion around how future IT would look like with fast evolving digital landscape. Application development and maintenance (ADM) engagements have been mainstay of IT outsourcing. This blog covers how ADM engagements landscape would change in the wake of digital transformation.

Before one makes an attempt to chart out future of ADM engagements in the next decade, it would be good to summarize the futuristic technologies that have been enabling digital transformation.


  • Pervasive Technologies such as Connected Autonomous Vehicles (CAVs) and predictive analytics enabling customer experience.

  • Cognitive Intelligence and Machine Learning are being applied to enable many business and technology functions. Future applications will be built with intelligence to learn and change cognitively, rather than execute on fixed instructions.

  • Wearables and fashion electronics have become popular and businesses offer multi-channel customer connect across various channels including wearables.

  • Disintermediation platforms have removed the middleman and have provided direct connect with the new partners.  Blockchain technology is providing a secure platform for partners to conduct business seamlessly.  

  • Robotic Process Automation (RPA) is quickly changing the concept of workplace.  The future workplace would offer increased co-existence of the robots, virtual personal assistants, conversational systems and humans


Application Programming Interface (API) & Micro services, Data Analytics and Cloud are changing the technology architecture of IT systems.

From Financial Service perspective, Banking on Cloud, Blockchain, Fintech, and Open Banking are trends shaping the industry.


Changing Expectations of Our Clients

In the digital world, clients value a combination of domain and technology skills, and are focused on outcomes rather than wanting to pay based on effort spent.  In this context, the service providers would need to -


  • Offer outcome management capabilities as against pure technical capabilities.

  • Offer single interface to deal with Business Technology rather than providing multiple experts to clients to reach out to. For convenience we call these professionals as "Business Technologists" as against "System Analysts" who brought in IT architectural view or "Business Analysts" who brought domain/industry view.


Changing ADM Engagements Landscape


Discrete design, development and delivery methods will fall short for Digital "Business Technology" projects.  The business technology engagements would require "Design" thinking.  "Design" thinking would remove the boundaries between software and infrastructure development. "Design" thinking process is more customer centric and iterative. It would assist in developing creative solutions when the problem itself is inadequately defined. 


Digital projects would be perpetual in nature and would require ongoing development.  Maintenance work would shrink. 

Digital Enterprise applications will be multi-layered. The schematic representation provides a snapshot -


Front-End

  • Increased self-service layer with applications such as mobility solutions, future branch

  • Limited assisted service applications that can work with third party applications.

  • Third-Party applications such as "Open Bank"

Security Layer - Enterprise security framework that can control access at organizational level

Applications Layer Customer Relationship Management (CRM) and Business Process Management (BPM) layers 

API Layer - API layer and Enterprise Service Bus (ESB) connecting the Applications (CRM, BPM) with products/Services layer.

Products/Services Layer -Products such as Core Banking Platforms, Payment Engines, Anti-Money Laundering, Loans packaged products as well as products offered on Service are part of this layer.

Enterprise Data Lakes - Offers common data across above layers

IT project delivery would follow Agile/DevOps principles - design, development, testing, infrastructure and deployment would preferably come from a single self-organized team to deliver projects.  Cloud native applications can be developed easily using containers/micro-services.  Containers would bridge the gap between new services and legacy applications.

There is a myth that the legacy applications cease to exist in light of new Digital applications.  The legacy applications would continue to co-exist along with new digital applications as customers need to retain the existing IT while they introduce digital transformation. 

The three plateaus of IT that the businesses witness would continue to exist, but the proportions would see a shift with legacy applications maintenance getting optimized, and making way for ongoing modernization and more development.  Digital development of today, considered complex would become the business-as usual.



Legacy is a relative term.  What is "Legacy Modernization" and "Digital" today would become Business as Usual "Legacy" in near future.


Conclusion - Vision for ADM Engagements - Key Priorities


Based on changes in ADM landscape discussed above, some of key priorities include -


Broaden and Deepen ADM reach - Examine each portfolio against our experience/service offerings and penetrate in white-spaces through extreme offshoring and extreme automation propositions. Distributed agile and SAFe methods should be deployed to enable extreme offshoring.  Extreme automation should be considered across onboarding, transition, development, production release, maintenance, and production support.  Also, areas such as Augmented/Virtual Reality (AR/VR), though sound engineering oriented, would easily leverage the programming skills in ADM team and be of interest to our programmers. 


Upskilling/Reskilling - Technology adoption, learnability and understanding of Business and technology is critical for ADM than ever before. Business analysts would need to be well-versed with technology and Developers, Testers and Managers would need to be more well-versed with business/domain knowledge. This expectation fully aligns to Infosys Zero Distance Philosophy - Every developer, project manager, analyst and architect should be at "Zero Distance" - to the end user (Desirability), to the underlying technology (Feasibility) and therefore to the value (Viability)".

The trainings should be on-demand and can be leveraged through partnership.  In addition, there should be increased focus on certifications/skill assessments as part of the training. 


Introduce Infosys Artificial Intelligence Platform (NIA) to ADM clients - Creating a NIA power-programmer team is a key priority.  Programmers should be upskilled in Machine Learning and Artificial Intelligence areas to implement NIA for ADM Clients. Some of the FS-ADM specific use cases of NIA include -


  • Understanding customer purchase behavior across retail channels

  • Data relating to credit history, KYC, fraud prediction, customer churn prediction, etc.

  • Analysis of tax relief at source and exception reports

  • Policy document knowledge and multi-channel chat-bot support


Roll-out Nextgen Delivery Model (NGDM) to enable at any scale, closer connect with clients and teams, enable multi-shift/multi-zone presence with clients.  The computing infrastructure could be state-of-the art with all machines web-cam enabled.


Offer multi-channel and chat-bot support - ADM service line should create a platform for multi-channel/chat-bot support for rapid development and deployment across multiple platforms


Author:
Renu Rajani, Vice President, FS ADM, Infosys Limited.

Supporting Authors:
Sastha Prasad Viswanathan, Group Project Manager, FS ADM, Infosys Limited.
Viral Thakkar, Senior Principal Technology Architect, FS STAR, Infosys Limited.