Ratul has more than 11 years of experience in technology and advanced analytics, machine learning and AI systems, primarily focused on consumer and SME lending.
Ratul worked in leading credit bureau Transunion Cibil in India, where he worked more on automation and machine learning using the power of Big data. Earlier Ratul also worked in Global data science leader SAS in its Research and Development office in Pune, India. Ratul holds a B.Tech from National Institute of Technology, Calicut (India).
Analytics India Magazine: How important is Data science & AI within Banking Systems in Malaysia?
Ratul Paul: Malaysia has 15 million credit-active consumers and 2 million Small and Medium Enterprises (SME). That’s a very small market with tough competition among banks to get a greater wallet share. With bad rate significantly high in the subprime market and banks working on a very small profit margin to be fiercely competitive, risk management and optimizing growth strategy are tremendously challenging here. Data science and AI use cases are primarily focused on reducing cost and risk, and increasing sells and revenue.
AIM: Can you elaborate on a specific use case of data science or AI that you worked on
RP: In Malaysia, bank’s client data is strictly protected by a lot of privacy laws, at the same time Banks are very serious about their reputational risk. In such a scenario banks in Malaysia are not proactively open to implementing AI and replacing conventional decision systems and scorecards, because an AI decisioning has a lot of black box layers and internal compliance team still not open to anything that they can’t interpret. In such a scenario, so far only a few AI use cases have shown potential in Malaysia.
1.Fraud detection – A lot of interest has been shown by banks to authenticate Identity Card and prevent fraud and a lot of players built deep learning AI models to authenticate IC. AI technology can authenticate multiple specific features in an Identity Card using optical character recognition(OCR), and that way significantly reduce fraud.
- Online loan application and decisioning – coupled with IC verification, mobile applications have been developed , that can check IC, match with applicants face and then move to the application, scoring and decisioning process, everything within three to five minutes, that way increasing customer experience and improving lag time significantly, as well reducing cost.
3.Auto alert for Consumers- This AI system notifies consumers whenever there is a loan or credit card applied using his ID, Approve limit changes, address or phone number changes in the banking system or when customer miss any payment. The consumer can validate if the changes are expected or fraudulent and take appropriate action accordingly.
- Natural language processing – In NLP front Banks are interested in building more advanced chatbots that can interact with the customer more intelligently and give a better customer experience. Banks are also interested to standardize quarterly financials those are submitted to banks by SMEs in a different format. This standardization has significant potential to reduce cost and lag time while improving customer experience.
- Alternate data – There is also significant interest among different players to merge data from different sources, like telecom, credit bureau and card provider data, together to profile a customer and understand him better with additional data points. That would help in up-sell and cross-sell strategy for banks with a more targeted campaign, Early collection strategy, And late collection for defaulted untraceable customers. when a customer defaults, additional data points can trace the exact whereabouts and profile of a customer and that way makes it easy for banks to collect the defaulted money. However, progress in this front is very slow due to compliance and regulatory policy and strict data privacy law.
RP: In Malaysia – there has been a huge interest among banks to embrace data science and AI. However, banks are more risk averse here than in other geographies and bound by strict compliance laws. So banks still primarily use licensed software (like SAS) and open source programs (like Python) are still not accepted and will not be accepted by banks in near future. Also, strict data privacy laws and the bank’s internal compliance makes it difficult or very time consuming for the analyst to get the data.
While banks understand that the potential of AI and alternate data is huge in adding business value and profitability, but banks want to go slow while changing existing process due to reputational risk and internal compliance and legal policy.
AIM: What are some of the challenges that the industry in Malaysia faces in terms of AI adoption.
RP: All entities in Malaysia are bound by strict consumer data privacy law and usage and scope of data are meticulously defined. So getting data for building any AI model would be difficult and time-consuming because the request would go through multiple legal and compliance team scrutiny. Even an NDA (non-disclosure agreement) sign takes months in Malaysia due to different legal clauses. At the same time adopting new AI systems needs to get strict regulatory approval first with a testing period of minimum 3 to 6 months, next internal legal and compliance team approval. Because of the lengthy process of adoption of new technology, banks are usually very choosy while showing interest in next-generation AI systems.
AIM: How can governments and citizen associations come together for a healthy discussion as well as the implementation of AI?
RP: Government and regulator can come together to ease the compliance process, for good use of data. At the same time, effective and supportive regulation needs to be in place to make the best use of data collaboration between different entities. Since new data points are available now like never before, old regulation guidelines need to be regularly amended to keep pace with the changing scenario.
Citizen association need to educate the common man that their data is protected and used for good purpose only, and it will eventually benefit the common man to get easy credit at need.
AIM: Is AI talent an issue in Malaysia? If yes, how can we resolve this?
RP: Malaysia might not have enough resource to work on advanced AI systems as of now, but many entities are working towards training aspirants, at the same time industry partnered with educational institutes to nurture talents and give right exposure even before joining the industry. Also, liberal immigration policy makes it easy for industry to hire foreign talent.
AIM: What is the biggest trend in data science/ AI that you look forward to in 2019?
RP: Reducing cost and increasing profitability is the driving force for AI in the banking system. Fraud is a big concern in banking and stopping fraud transaction has been a pressing issue for banks in Asia. AI systems that can prevent fraud and at the same time reduce operational cost and generate greater revenue would be in demand in 2019 and coming years. Also, Chatbots will advance significantly to improve customer experience.
Provide your comments below
Bhasker is a Data Science evangelist and practitioner with proven record of thought leadership and incubating analytics practices for various organizations. With over 16 years of experience in the area of Business Analytics, he is well recognized as an expert within the industry. Earlier, Bhasker worked as Vice President at Goldman Sachs. He is B.Tech from Indian Institute of Technology, Varanasi and MBA from Indian Institute of Management, Lucknow.