Technology is now used in thousands of unseen ways to bridge the gap between human needs, products and services. From the smartphone which suggests eating spots you might like based on your location to paperless banking and even smart factories – anticipating human needs has always helped businesses to ramp up sales, reduce waste and give services tailor made for customers.
A large part of this is thanks to artificial intelligence (AI)-driven automation. By implementing robotics to optimise mundane processes such as data entries, cash deposit, passbook updating and salary uploads to name a few, the banking sector has increased its efficiency manyfold in the last decade. Nowadays automation is pushing the boundaries of technology, from selling financial products to customers, to loan processing, to creating a much more productive relationship between customers and machines.
Used along with Machine Learning (ML) algorithms, newage software solutions have created the idea of “Intelligent Enterprises.” The term refers to a management approach that applies technology to the challenge of improving business performance, and is driven by a blend of technology-based services and a higher understanding of human behavior. But how are intelligent enterprises different from traditional knowledge-based business solutions and what are its benefits for companies?
Benefits Of AI And ML-Powered Banking
Nowadays the biggest banks execute AI to make vast changes in workforce, customer experience and expenses. In fact according to Accenture 2018 survey held with bank executives, the upside of AI is highest in three areas:
- Building customer trust and confidence (71%)
- Cost and operations optimization (63%)
- Improving compliance with regulations (62%)
One of the biggest challenges in banking is having to process, control and analyse millions of unstructured and fluctuant data. AI automation, not only processes big data simultaneously; it also leaves no room for human errors and doesn’t let anything drop beneath the radar.
However to get the best from the technology, banks need to look beyond their talent pool to understand where AI can be most helpful in working with humans and solving real-life problems without minimal fuss. For this banks are now reaching out to startups which are heavily leveraging AI and machine learning to scalable solutions and innovative use cases.
Yes Bank Datathon
In a pioneering effort towards integrating automation and machine learning further into its banking operations, India’s 4th largest private bank — YES Bank — concluded its 100-day long accelerated YES Datathon on December 23. The Datathon is a pioneering corporate-startup partnership, to identify the top 20 solutions for its own banking requirements. The chosen concepts now be taken live within a month.
Despite this being the first Bank-led data science hackathon programme, aptly called Datathon, 6,000+ data scientist/engineers and developers from across the country and some from outside like Sri Lanka participated in the programme.
Speaking to Inc42, Amit Shah, Group President & Chief Fintech Officer, YES BANK stated, “The entire programme — YES Datathon — was conducted in multi-rounds, with 200 teams out of 1600+ participating teams shortlisted for the second round. This was further brought down to 50 teams for the finale round held on December 22-December 23, 2018. We are overwhelmed by the kind of solutions, these teams brought to the table.”
Some of the leading banks — HDFC Bank, ICICI Bank and SBI — have organised various hackathon programmes to meet the talent requirements. Many have also gone for hosting acceleration programmes to build data-oriented solutions.
However, Datathon is an entirely different concept. And, unlike Hackathon which is usually 48-hour long, it’s three-six months-long challenge where candidates have to actually develop PoCs based on the real data.
Speaking to Inc42, Rajat Kanwar Gupta, President, Business & Digital Technology Solutions Group , YES Bank explained, “We always had lots of information and data that we collect from the customers through KYC forms etc. For a long time, these data were only used for risk and other limited use cases. But, we had more information than many of the data-driven companies claim to have.”
He added, “So, starting from 2016, we started digging more into the data science and business intelligence from the existing data that we had. We created a core data science team and started building blocks — data management team, data mapping etc — which needs to be in place before the data analytics part comes in.”
Related Article: How Will Artificial Intelligence Change The Banking Industry?
If the bank set up its own internal full-fledged data team, why to be dependent upon the outsiders’ solutions for banking requirements?
To this, Gupta responded, “We have over 2 Mn customers. The data is much more and so are our requirements. We wanted to have specific and customised banking solutions which be easily adopted on the work floor in short span of time. This is why we felt, hackathon or acceleration programmes which we have separately conducting for years now are not the right approach.”
Gupta also clarified that the data to be shared with the participants had to be anonymised. Further, the data must be available on the server in the form it is used in banks. It can’t be allowed to be shared or copied.
“We needed diversified products based out of big data analytics, AI and ML. Finally, we locked on Datathon,” he added.
YES Datathon: India’s 1st Bank-Led Datathon
Datathon, launched in September 2018 aimed at augmenting YES BANK’s embedded data analytics & ML units to drive rapid prototyping of AI/ML based products, optimize digital product suite, and enhance product/service design and delivery, in an accelerated 100 day period
Over 6,000+ candidates constituting 1,600+ teams applied for the Datathon event where the teams were given a challenge to create some AI, ML, big data analytics based unique banking solutions out of given YES Bank’s curated and anonymised banking data.
Out of 1,600+ teams, 200 top teams were identified from the challenges were given 60 days to create working data models/prototypes which will be trained, tested and deployed by the bank by January.
Shah informed, “The top 200 teams include not only students from top technology institutes like IIT Bombay, Chennai, Kharagpur and ISI Kolkata but also 150+ professionals from organizations like IBM, Walmart Labs, Siemens, Amazon Development Centre, Capillary Technology, TCS, Accenture, Amdocs and Infosys among others, who are taking on the challenge beyond their professional duties.”
During the finale round, “Interestingly, 90% of Datathon’s Top 50 teams are working professionals in the field of data science. This is a unique facet wherein data professionals from a non- financial services background are participating in the challenge beyond the work hours,” he added.
The Top 15 Teams Who Showcased Their PoCs During Finale
Oracle ( Professionals from IBM + 2 students from IIM Bangalore): The team has worked on a ‘master product’ which creates a single 360-degree view of every retail customer, and also provides customized product and service recommendations for every individual customers – (including product propensity/predictive service delivery/service resolution etc.
Finance Data Dons (Professionals from TCS, Chennai ): The team created two products
- Product 1 : The team has created a data model which creates a personal finance management tool for every customer using transaction patterns as a base. The application will have the ability to predict and classify transactions as well as smart investment advisory once implemented
- Product 2: The model will help optimizing Merchant relationships – both online and offline retailers. The team has used ML to cluster POS transactions and identify anomalies in transaction volume and volume. This will help to increase POS/gateways, identify new merchants, provide merchant offers
Django Unchained (Professionals from Vodafone & PwC): The team has created an AI-based application for relationship managers (sales representatives) of the bank which will enable them to measure share of wallet for the bank for every retail customer, predict attrition as well as provide customized products/services.
Reverse Atlas (Students from VIT, Vellore): The team has worked on a unique ML algorithm (which finds usage on streaming sites) which helps identify customer relationships basis transactions and creating relationship trees for every customer basis transactions. This would help identify and target new customers with exact products/solutions/offers as well as deepen current relationships.
Insights ( Team from University of Moratuwa, Sri Lanka): The team has developed a unique model basis, using a modified version of Shapley’s value – called Shap Value to create a customer – product/service propensity model basis product holdings/demographics/transactions. For instance, the propensity of a customer with 2 FD holdings towards an MF investment. This model combined with Team Oracle creates a comprehensive Next Best Action model.
NLP Rockers (TCS): The team used NLP and clustering methods to convert email service requests into service tags across the entire customer bases and subsequent clustering and prioritization. The model then uses predictive analytics to create customer clusters and predict service requests per cluster.
Zessta (Data Science Team for Zessta Software Services): This product works on lowering customer attrition using a clustering algorithm which identifies customers most prone to attrition.
Avensis (mix of students and professionals): Chat bots have been quicker and better but a missing link has been customization. This solution will help YES BOT provide more customized responses to customers/non customers using demographic and account data , while also providing regular services like payment reminders/EMI tracker and relevant investment notification.
Meson Labs (Startup – Meson Labs): This product takes a unique approach to managing finance and investment using neural networks to create human like learning as well as use Google adwords to help classify transactions since transaction markers are often commonly known brands or words. This helps create a relevant map of all transactions for individual customers, comparison of spends and investments and accordingly provide advisory”
Prayaas (3 Students from NMIMS, Mumbai one from ISI, Kolkata): The team has worked on 2 solutions, the first one helps manage and predict individual customer deliveries like cards/statements/tax statements/cheque books – providing proactive service, reducing human involvement and as a result reducing service requests.
Data pirates (Student Team): The model provides an alternative method to score and profile customers , pooling in LinkedIn APIs and other external APIs to help score customers who are currently not scoped by CIBIL.
GSA07 (Professionals from Direct I and Nearby Technologies): A major problem that bank customers face is to connect to the bank to avail a service, address any queries or resolve an issue. To address this concern this model will put in place which anticipates a customer’s concern based on his demographics and transaction data and allows the bank to reach to their customers with a solution before the customer knocks at their door with a problem.
Fakedata ( students and professionals mixed team): The model uses machine learning to create a customer satisfaction or ‘happiness index’, and basis affinity of customers to products or brands (basis transactions) – provides customized offers/offering as well as enables more nuanced and targeted marketing.
Greenity (Startup Greenity Solutions): The team created a predictive ML model, which studies current credit card customer’s transaction patterns/buying behaviour and preferences to create a pattern for predicting target customers for YES BANK credit card.
YES Datathon: What Lies Down The Line?
It goes without saying that given the rapid advances in the AI field, such programmes need to be an ongoing and continuous process of ensuring and optimizing the balance of startups problem solving skills to traditional banking sector.
So at the sidelines of the Datathon, Shah said that YES Bank is now actively partnering with top technology institutes — IITs and BITS— and will also host AI/ML challenges and data engineering workshops to deepen practical and technical knowhow of future technology leaders and widen the data science ecosystem.
The 6 Data Science/Machine Learning challenges hosted in partnership with IITs/BITS will lead up to YES Datathon 2019 to be hosted in June 2019. Datathon will now onwards be prominently used to meet the talent pool requirements also.