Banner Image

Bank of the Future

Cognitive Automation: Tools that Think

April 18, 2017

Smarter data management is emerging as a strategic imperative for banks to function effectively and drive growth. Cognitive technologies have given us a set of tools that can mine a vast collection of complex data in video, audio, text, and structured databases, and extract deep insights to aid decision-making. These technologies are all the rage today because of new leaps in the field of artificial intelligence (AI) leading to advances in computer vision, machine learning, natural language processing, speech recognition, and robotics.Cognitive tools are based on a host of algorithms that are designed using probabilistic logical reasoning, neural network based architectures, fuzzy knowledge representation schemes, and more, to analyze and extract meaning from the unstructured content. Here are three ways in which banks are using cognitive tools:

Personalization and profiling: The different types of data that a bank routinely collects can reveal clues to customers personalities, lifestyles, preferences, as well as their aspirations. These insights can be used to deepen engagement and identify cross sell and up sell opportunities.

Profiling is also useful in the case of corporate customers. Cognitive technologies can aggregate information about corporates from the internet and build profiles related to the context of a query. For instance, when a bank processes a loan application, an AI-based application can pull up earlier loan requests of the customer, if any, from the database, and the clients credit score, and borrowings and repayment information from regulatory filings. This could also be bolstered by social media sentiments about the company in relation to this particular aspect, building a complete profile pertinent to the loan request. These profiles are in turn utilized by financial advisors. New predictive models that can profile customers based on cognitive inputs are also in the offing.

Cognitive assistants for personalized interactions: With the maturing of natural language processing, banks now employ chatbots to conduct human-like conversations with customers. This brings value in cases where the customer has multiple options or has to wade through a lot of technicalities on a banks portal. Bots give a sense of personalized assistance as well as the feel of self-service. This technology enables customers to remain in touch with the banks round the clock. Banks can also explore hybrid systems, where a bot fields queries up to a point and then hands over to a human agent for further personalized responses. Bots can also proactively publish tailored information about banking services to customers. The caveat here is that the bot has to be trained on good data sets. For instance, the chatbot should be trained in answering all the possible queries a customer can have regarding a product or service, which the bot is expected to support. Customer journeys with the bank should be thought out clearly. This puts the bank staff in the customers seat, and is a good way to understand customer pain points.

More the customer interactions with conversational systems, more the data available for the machine learning algorithms to learn, build, and enhance the psycho-demographic profiles of customers. These profiles help in providing a rich, personalized experience that can seamlessly blend long-term behavioral aspects of individuals with their short-term or immediate requirements.

Automating compliance: A relatively new application of cognitive technologies is real-time detection of regulatory violations. This could prove to be a competitive advantage given that violations attract intense regulatory scrutiny and huge penalties. Of course, the application for this has to be built with deep interpretation capabilities of compliance rules and regulations embedded in unstructured documents. Keeping track of amendments and exceptions is a tedious activity, which is now being automated through intelligent algorithms that employ deep learning as well as traditional machine learning methods. While these algorithms tend to learn better from human annotations, attempts are on to see whether there can be more efficient methods to learn from observations of human activities.

Cognitive computing offers a solution to the too much data, too little insight conundrum that has long haunted banks. Data can be used at a completely different dimension: decision-making on routine queries, fraud detection, and rich customer experience. While cognitive tools require investment and training, the benefits can come as a pleasant surprise.

Lipika is a chief scientist at TCS Research and Innovation and heads analytics and insights practices. Lipika holds a PhD in computer science and engineering from IIT Kharagpur. Her research interests are in the areas of NLP, text and data mining, machine learning, and semantic search.