How Fintech Companies Are Using AI And Deep Learning

How fintech companies are using AI and Deep Learning to provide platform solutions to businesses during the Pandemic

In Fintechs, in the evolving post-lockdown scenario, digital seems to be the only viable strategy. However AI is being used as mainstream technology in Fintechs even before Lockdown. In many operations from  fraud detection, to digital eKYC, customer service call centre, credit decisioning using digital review, engaging chatbot, support chatbot, even in day to day portfolio management, AI is proven to add value in significant ways, not always as a replacement of existing system, but more often as a reliable assisting system.

Digital fraud is a new threat in digital world. In fraud detection AI helps to find anomaly detection, finding a pattern from millions of transactions that is unusual and more prone to fraud and raise prompt flag in real time for further investigation.

Digital eKYC seems the way forward, in spite of the regulatory hurdle and data privacy issues, AI seems proved digital would be only way for eKYC in near future. In this case from the live video multiple snapshots are taken with different facial expressions, which is validated against photo given in KYC document with a threshold match probability confidence. It can significantly improve authentication accuracy, that too in few seconds. At the same time AI helps to authenticate the KYC document, using OCR to validate authenticity of different entities in the document with very high accuracy rate compare to human manual process.

AI NLP is proven to read financial documents and capture the numbers in a standardise format, reducing time for credit processing, however when work in this space is still ongoing, it can still be used as an assisted system.

AI speech processing has a major use case in Customer service call centre, where a follow up call is automatically classified and scheduled on time, taking insights and sentiments from the call. AI can as well create summary of each call and share that  for management review and escalation. In old process a sample of the recordings were manually monitored , with AI entire call recordings can be monitored in little time.

In MSME lending, digital reviews of MSME can add value in credit decisioning on top of bank statements and credit history. A web scraping can get certain number of digital reviews of a MSME and NLP feature extraction can be used to build credit decisioning features. It can also be used to find customer sentiment and monitoring that over a period of time. As well it’s a proven technique to find competitor strength and weakness and align internal strategies accordingly.

Chatbot is another use case of NLP deep learning, with multiple use cases, such as engaging customers visiting homepage or solving most frequent customer queries without human intervention. A good chatbot can be a differential customer experience.

A very interesting use case in Deep learning Convolution neural net is to build thousands of trainable features from limited variables which gives significant better results compared to conventional machine learning techniques. A technique proved its potential in lead prioritization, SMS campaign prioritization as well identifying different marketing segments. It reduces marketing cost and significantly increases ROI, which could have not achieved using conventional machine learning models.

Another interesting use case of deep learning is using Recurring Neural Net using Long Short Term memory (LSTM) to predict accurate time series trend in sales, count of leads coming in different channel ,count of customer service call or customer complain in next week, even predicting when more bad or fraud customers will hit LOS and tighten acquisition rules before they hit.

When Fintechs are embracing AI & Deep learning in recent times like never before, one point to keep in mind that the journey will not be overnight, but it will be evolving over time. So more than complete replacement of existing systems, the success criteria should be to find an assisting AI system which can add value. Same time as Fintechs are highly regulated, regulators need to be more open on embracing new technologies. Together,  it will bring significant transformation in customer experience.

Covid pandemic poses an unique challenge, it’s a never before event for Fintechs, which demands a set of never before solutions for going forward. When physical distancing will remain a challenge in near future, digital eKYC probably be the only way forward. An integrated automated credit evaluation using hundreds of data points with video eKYC , with decisioning and disbursal within minutes will be the norm. When existing models with limited features and long historical trend will fail to hold predictive power, thousands of features created using deep learning convolution neural net from limited data could be a reliable solution. Post corona, with more and more digital AI in mainstream, Fintechs will change, and it will be change for good.

Launch Your Fintech Credit Risk Career

An online Credit Risk course aimed at helping you launch a career in Credit Risk. Students from 43 countries and counting, meet the 100+ alumni’s who cracked career in Credit Risk. Contact at +91 8806930487 or drop a mail at scientistexpress@gmail.com for a free demo class.