Artificial Intelligence in Commercial Banking
Artificial Intelligence in commercial banking explores how commercial and corporate banks are thinking, innovating, and implementing artificial intelligence and cognitive solutions to serve customers better and manage risk.
While much of the frontpage stories about Fintech and the digital wave transforming banking centers around consumers – robo advisors, crowdsourcing, and innovations in payments – the digital and cognitive technologies have been quietly changing the staid old world of commercial banking. Cognitive technologies and artificial intelligence in banking and financial services are at a nascent stage and we expect AI to be a prime driver of the banking industry’s future.
One of the myths in the corporate and commercial banking circles is that relationship trumps everything. Indeed, the relationship is essential, but today a high-tech enabled high touch is the order of the day. So, it is not that robots and chatbots will completely replace commercial bankers, but the intelligent machines can help the human interactions become more insight-driven and meaningful.
Corporate treasurers may appreciate a once-in-a-while “How you’ve been?” calls from a relationship manager from a bank, given their busy schedules they may prefer if the interaction is data-driven and results oriented.
Beyond the front-end customer interactions, artificial intelligence and cognitive technologies – machine learning, conversational AI, natural language processing, natural language generation, deep learning, computer vision et al – can transform the commercial banking sector in a holistic manner and afford new opportunities in business development, products and services, risk management, operations, and client service.
Let us examine how banks are leveraging and should leverage artificial intelligence in various facets of the commercial banking value chain.
7 High-Value Use cases of Artificial Intelligence in Commercial Banking:
Now, let’s dive into each of these seven high-value use cases on how AI is helping commercial banks. These seven AI case studies in the commercial banking sector are just the tip of the iceberg.
AI-enabled Customer Segmentation
Machine learning and artificial intelligence are changing how commercial banks can engage in intelligent customer segmentation to find the right customer at the right time with the right message with a propensity to buy a specific product or service.
As we all know, segmentation is about dividing customers into a set of homogenous groups that are similar in specific ways concerning a set of parameters that are relevant to the business, and meeting the needs of that customer group appropriately with products and services.
Segmentation is no longer a simple function of identifying companies that have revenues over “X” and employees over “Y” and in industry “Z.” Today, with big data – streams of unstructured, structured, and semi-structured data that flows from internal and external sources – commercial banks can analyze thousands of parameters across millions of data points and come up with a new genre of customer segmentation.
“Basic segmentation models in banking do not provide sufficient insight into customer needs, channel preferences or desire for receiving advice from banks. Marketing strategies based on standardized assumptions — rather than deeper insights about life stages, behaviors, attitudes, interests, lifestyles and other psychographic factors — are destined to fall short.” – Ernst and Young
For example, Lattice Engines, an AI-enabled engine has launched a data cloud for customer segmentation. Using the new generation segmentation engine, Lattice’s customers are able to create unlimited segments using a combination of the Lattice Data Cloud, which provides more than 16,000 attributes on 200 million companies globally, and their own internal data, including CRM and ERP data, transaction data, product usage data, support and service data, etc. The segments are then made available in company’s execution systems (e.g., marketing automation, CRM, or ad targeting platforms) to enable teams to run hyper-targeted campaigns across the entire buyer’s journey – from awareness to engagement to conversion to expansion.
“Merging AI, external data and a customer’s own internal data sources to create detailed segments is the only way for marketers to get an edge when it comes to driving superior customer experience,” said Shashi Upadhyay, CEO at Lattice Engines.
Commercial banks are using machine learning techniques over a vast data set and can develop micro-segments that are similar to a psychographic and buying behavior perspective, even if their demographic profiles seem an odd coupling rather. It may be that the fruit card vendor down the street and a mid-size manufacturing CFO share similar personality and behavioral traits that may make them ideal customers for mobile transfers. Based on this intelligence from thinking machines, banks can fashion the appropriate services and orchestrate messaging, pricing, and timing to be most effective.
KYC (Know your Customer) and Fraud Prevention:
Banks have spent years trying to the tackle the tricky concept of KYC (Know your customer), particularly in the recent decades as money laundering on behalf of criminal enterprises, has become prevalent in a global world and criminals always seem to be a step ahead of the countermeasures.
Today, cognitive technologies can provide a boost to commercial banks KYC efforts. The primary job of artificial intelligence and machine learning is to identify high-risk customers and high-risk behavior for enhanced due diligence. Machine learning algorithms can scan through billions of diverse data points to identify patterns that will be impossible for a human to fathom.
ML-based complex link analysis is a explores deeper associations among a vast data set of different types – both historical and contemporaneous. These machine learning methods are vital for providing human investigators a synthesis of information in bite-sized nuggets to allow them to comprehend complex webs of clues and evidence and inferring conclusions that are not apparent from any single source. This link analysis and pattern recognition are where the machine learning algorithms excel.
For example, recently FICO (Fair Isaac and Co) has acquired Tonbeller developed anti-financial crime solutions which use unsupervised Bayesian learning algorithms and other techniques to discern deeper aspects of customer behavior which in turn drives further investigations into the suspicious activity.
Hitherto, commercial banks went through the traditional process of obtaining a credit score, a Dun and Bradstreet profile, and poring over financial statements to determine the creditworthiness of corporate customers seeking a loan or a line of credit. This rudimentary approach to loan underwriting and assessing creditworthiness has resulted in one of two problems: a) too stringent of criteria to understand the business seeking a loan and hence higher loan rejection rates. Alternatively, b) without nuance, loosening of credit underwriting standards to an extent there are numerous bad loans on the banks’ books.
Again, techniques such as big data, machine learning, and artificial intelligence have transformed the field of loan underwriting and determining the creditworthiness of a corporate customer.
For example, Lenddo, an AI technology company introduced machine learning based algorithms to look beyond the traditional underwriting methods to foster a new way to determine creditworthiness, in part on thousands of non-traditional sources that look beyond a credit score and a financial statement.
Another company, Zest Financial introduced ZAML (Zest Automated Machine Learning) as a way for banks to look beyond the basic underwriting criteria. The ZAML platform is an end-to-end underwriting solution that allows banks to leverage machine learning and include thousands of data points, at scale, and with speed and full transparency. With ZAML, banks can accurately assess thin-file and no-file borrowers—such as millennials—that traditional underwriting systems overlook.
The same machine learning algorithms can also monitor the borrowers’ credit on an ongoing basis generating scores almost in real time. Machine learning platforms consume vast amounts of data – unstructured and structured, social and corporate – and recognize complex, nonlinear patterns making highly effective and accurate risk models possible. Thanks to both supervised and unsupervised learning, these ML-based risk models learn with every new data point they acquire, improving their predictive power over time.
These risk models can be deployed at an entire loan portfolio level, specific segments, or an individual customer. The interplay of factors at a portfolio level will allow the banks to adjust the loan loss reserves and take mitigation steps at a portfolio or personal account basis.
One of the vital deficiencies in commercial banking is the amount of unstructured data that flows through the system. These unstructured documents and data are not only ubiquitous but also are hard to manage. At times, a slew of team members manually types in unstructured data into a system and sometimes into multiple systems. The manual entry is typically limited to a few data elements and is prone to errors. Lack of traceability to the source documents also limits the ability to audit and reconciliation.
Again, no wonder, artificial intelligence comes to the rescue. Natural language processing (NLP) can extract data, meaning, and insights and provide it for consumption by downstream systems as well as by human beings. The NLP techniques are evolving to the point that they are capable of not just dependent on position, frequency, coordinates, but can “read” “understand” and extract meaning with context. This is a huge leap.
Commercial banks can use the NLP-based extraction in various processes such as extracting and normalizing data from said private company financial statements. The private company financials are often in PDF format, and the specifics of each report may vary by industry, size, and complexity of the firm, and the financial line items of relevance. The concept, called “Financial Spreading” is enormously helpful for commercial banks. The spreading of the data elements not only eliminates manual data entry but also provide traceability for making auditing and reconciliation easy.
Commercial banks are littered with contracts which are often stacked as papers in a dark room or digitized as PDF documents or images. Hence, monitoring the contract terms and enforcing adherence is rather very difficult. Sometimes, it takes a small army of expensive attorneys to sift through the contractual terms and legalese. However, the aforementioned NLP and its companion technology, the NLG (natural language generation) can extract the critical contractual covenants and provide them in a way that human analysts can act upon. The systems can also generate alerts based on pre-set governance norms.
One of the most critical areas that banks can leverage artificial intelligence is of course in engaging the customers. Chatbots and conversational AI are evolving fast and can participate in multi-threaded conversations. Today, with training it is possible for a chatbot to walk a corporate customer thru the process of factoring, for example.
Commercial banks can deploy chatbots directly to the customers as well as within the branches and call centers to allow the relationship bankers and customer service personnel to query the chatbot on the fly and converse with the customers.
Of course, chatbots are not necessarily a reflective term as today voice-enabled assistants have mushroomed all over. Many banks are using Amazon’s Alexa to develop so-called skills.
Artificial intelligence can help commercial banks predict which customers are about to leave. Using machine learning and predictive analytics, banks can develop models to analyze the customer behavior through interactions and transactions with the bank and external behavior on social media, search patterns, and other streams. And this potential churn information can alert commercial bankers to take remedial measures.
AI-powered churn prediction helps to analyze omnichannel events and identify dropping customer engagement. For example, if the system notices churn-indicating behaviors like lower usage time, it can send users relevant offers, push notifications, and emails to keep them engaged. – Markus Lippus of MindTitan
The models tend to get better as more information flows through the system and allow commercial banks to prioritize which customers to contact first. More importantly, the ML models identify vital drivers that are causing customer churn and hence corporate banks can essentially address the underlying issues and not operate at a surface level.
Artificial Intelligence in commercial banking is here to stay, and the use cases will only get broader and the outcomes better. Commercial banks using AI are likely to survive and thrive the unprecedented assault on banking by Fintechs and non-traditional players like Amazon, Google, and Walmart for example.
The infographic below summarizes the use of AI and Machine learning models in banking.