November 29, 2019
by Karen Krivaa, GigaSpaces 29 November 2019
Nimble fintech start-ups and established tech giants like Amazon and Alibaba are upending retail banking. Traditional banks need to provide the best financial terms, and superior customer experiences to stay in the game.
They can fight back by deploying deep learning solutions armed with NLP to respond faster to customer inquiries, proactively recommend services, analyze contracts quicker and improve their ability to comply with regulations. Here are just a few examples:
Upselling products and services
Chatbots armed with NLP can make upsell and cross-sell suggestions based on a customers’ current activities combined with their personal banking history. For example, Toronto based Sun Life has created the virtual assistant, Ella, to inform customers of an upcoming change in the status of their account such as “Your child’s benefits are about to expire.” An automated service can also use NLP to provide human call center agents with personalized recommendations for new service offerings that is derived from a textual analysis of the latest customer interactions combined with their loan, mortgage, and investment history.
Responding to customer inquiries
Customer service representatives can leverage deep learning solutions including NLP to reduce the average call handling time. NLP techniques can analyze text to understand the customer’s intent to route the call to the most relevant agent. By analyzing previous requests and responses using text classification, smart agent assistants (similar to those utilized at FinanzInformatik), can present the five most similar cases in milliseconds. Bots can also be used to use text to speech conversion to respond in a way that is closest to a human voice. Bank of America's bot named Erica has assisted one million users in three months and can provide information about specific transactions with a particular merchant, and the current amount of a customer’s total credit and debt.
Financial digital advisors
An investment firm can build a personalized investment strategy by analyzing patterns in how customers spend, invest, or make financial decisions based on an analysis of their transaction history using NLP. The software can then perform sentiment analysis using NLP to scan news outlets and social media such as CNN, BBC, USA Today, NY Times, etc. to predict future price movements for the relevant investments before making a specific recommendation.
Time consuming, competitive mundane work can be eliminated by using a model utilizing NLP to interpret, record, and correct digitized contracts at high speed. The accuracy of the model’s outcome is remarkably high because of the repetitive nature of contracts. COIN, machine learning software that utilizes NLP, launched by JP Morgan Chase, is able to extract 150 relevant attributes from 12,000 annual commercial credit agreements in seconds. Previously, the bank’s legal team spent around 360,000 hours manually reviewing commercial loan agreements.
Speed up compliance
Financial advisory services are highly regulated and financial firms are required to monitor the performance of their advisors to ensure compliance. A customer complaint, such as charging a fee that was not disclosed upfront, could attract a fine or a bad regulatory rating from supervisory agencies. NLP software can run through several millions of incoming customer service requests and categorize text to populate a table of interactions that should be reviewed by the compliance team. In addition, NLP can be used to prove that banks are meeting GDPR regulations by deleting personal data as per customer requests.
Financial institutions can provide superior customer experiences by deploying machine learning and deep learning solutions that include NLP. There are, however, several factors that can create a barrier for running machine learning models in production. Models need to be fed with fresh and historical data reliably and continuously in order to be retrained. Bottlenecks ingesting and processing data can slow down models and sabotage their results.
A distributed in-memory computing platform can act as a unified data speed layer to operationalize models by minimizing data movement to practically zero while leveraging the speed of RAM. The platform can support high ingestion rates of millions of IOPs, store any type of data (structured, unstructured and semi-structured) and seamlessly run analytics models with extremely low latency.
Machine Learning and deep learning armed with NLP can improve the efficiency and effectiveness of banking services. By using technology to accelerate the process of storing, moving and analyzing data, personalized services are becoming even more relevant and effective, providing a significant differentiator for financial institutions and a better customer experience for everyone.
About the Author(s)
You May Also Like