Banks and insurance companies spend a lot of time indulging in data inputs from hand- written cheques and forms. Needless to say, such manual task often leads to a dip in productivity. However, certain organisations are trying to come up with a solution regarding the same. Chennai-based organisation, Marketsof1 is one such company, which is trying to reduce that time consumption by 80% with the help of technology such as Artificial Intelligence (AI). In a conversation with Aparajita Gupta, Mahadevan Jayaram, Co-Founder, DeepQuanty Artificial Intelligence Labs explains the challenges and probable solutions it can offer to the sector.
Excerpts from the interview.
The vital role that the Banking, Financial Services and Insurance (BFSI) sector plays in building a nation requires no explanation: be it offering critical loans, keeping risks at bay or in general smoothing out capital flows. However, the sector also deals with extremely confidential information and hence laws, rules and regulations about privacy of information restricts inter-institutional transfer of data; that is, a bank won’t share with another bank or NBFC, the banking history, identity or any such information.
And yet the same industry seeks colossal data about a person or institution before taking any step in creating or allowing flow of capital. Hence, one is able to acquire credit or money only after engaging in monumental paperwork. The paperwork then is transcribed, analysed and finally a decision is taken. It takes time.
Consider a typical loan application form that is usually handwritten, accompanied by a couple of years' statements of accounts, six months' statement of bank accounts, KYC documents and such. It can easily take about a week to process the same.
The situation is no different for insurance application forms, insurance claim settlement processes, bank saving account, current account, credit card applications forms, gold schemes and even chit funds.
We at Marketsof1 automate the process of transcribing and extracting information / intelligence in the above process by applying cutting-edge Computer Vision based on Artificial Intelligence (AI) and Machine Learning techniques.
In order to ensure that we save as much time as possible, we have certain features that can help us save a lot of time. For e.g. one of our features titled SnapChek extracts data from a cheque. While another feature named, FormEasy extracts both handwritten and printed data; it also checks if all the required fields are filled. On the other hand, ZapSkore extracts all information from submitted bank statements, tallies debit and credit entries, applies intelligence to extract the number of cheque defaults, average quarterly balances, incomes, expenses, etc. Our EasyKYC extracts all relevant information from a KYC document such as driving licence. Depending upon the complexity of the document and the data extracted, we expect a time reduction up to about 80%.
AI and Deep Learning are the backbones of all object recognition and pattern recognition. So is it for any of DeepQuanty’s products. Products have been developed using tens of thousands of real and simulated documents. High-end computers have trained to see and read such documents just as young children gaze at billions of images to read, understand and comprehend visual data. Just that, computers learn to do them much faster.
Documents are first converted to images. Images are then processed for removing noises such as smudges and disturbances. The processed image is then fed in our home-grown applications that extract data. The applications typically run on high-end processors. Consider a page of bank statements. It may take even seasoned professionals about half a day to extract key information to decide whether to provide a loan or not. ZapSkore can crunch that time in great measure. Or consider automatically processing a cheque or an application form for any bank product - there is considerable gain in time.
SnapChek is specifically designed to process bank cheques. Most cheques in India are still handwritten, while others are printed. Banks spend time in processing both inward and outward cheques. SnapChek, enables extraction of date, amount in words and figures, compares the two, extracts IFSC Code, MICR and Account Number. It also compares the signature in the cheque with card and signature stored in core banking.
Currently, we are perfecting the next version of FormEasy and ZapSkore to tackle different product requirements. For example, version 1 of FormEasy works when the consumers write only within the stated boxes. Version 2 works even when such an assumption is relaxed. ZapSkore v1 works when the statement has little 'noise' such as stamps marks, cross-lines etc. Version 2 relaxes the same.