by Jelani Harper
SAN FRANCISCO – There’s no shortage of applications of the various dimensions of artificial intelligence in financial services. Typically, forms of rules-based algorithms, machine learning, and natural language processing are deployed for mandates such as Know Your Customer and Anti-Money Laundering measures.
However, there are a number of everyday situations in which AI’s vaunted automation is able to deliver practical business value in ways that decrease time to insight, reinforce customer service, and aid business users in their jobs. By deploying manifestations of AI in core workflows, organizations are able to expedite them and, consequently, enable employees to concentrate on more meaningful tasks.
According to Vizru CEO Ramesh Mahalingam, in automotive insurance claims adjudication processes, “$.30 for every dollar of a premium an insurance company receives, it goes to the claims processing. On an average it’s about 220 steps that today insurance companies take. It takes about eight weeks to two and a half months to approve a claim. But [AI ] systems can do it in about 45 seconds.”
In this respect, the deployment of AI in financial services merely represents the next phase of a longstanding evolution of helping organizations increase productivity.
“Overall, if you follow the evolution of business tools, automation and innovation have always won,” UJET CEO Anand Janefalkar denoted. “Replacing recursive mechanical tasks in an intelligent manner, which used to be called automation and sometimes now is called AI, essentially is what we’re doing.”
Natural Language Processing
There’s several ways in which NLP is leveraged by organizations within finance to streamline and accelerate processes. One of the most pervasive is to augment the omni-channel experience in which customers interact with organizations in a variety of mediums, including text, voice, chat, email, and more.
Janefalkar mentioned a use case in which an organization is leveraging aspects of NLP with contact center software “on 14 different languages across the world, and using specific permutations in the EU to be compliant with those regulations in addition to GDPR and stuff.”
In the auto insurance claims adjudication use case Mahalingam referenced, NLP is critical for the implementation of an intelligent omni-channel user interface, enabling customers and policy holders to initiate claims processes.
An effective way to input NLP into this process is as a means of fortifying the overall intelligence of bots, which underpin that user interface. According to Mahalingam, in this use case, what people conventionally term chatbots are actually omni-channel bots, which “automatically can receive information from multiple channels and translate it intelligently” to the rest of the claims adjudication system to triage and correctly route information. NLP enables these bots to understand the intent and context of the user’s communication, and is an integral aspect of the user interfaces for financial institutions.
There are plentiful ways in which everyday applications of machine learning assists financial service organizations. For customer service, for example, “machine learning and predictive algorithms…ensure that… wait times that we’re reporting are very, very accurate,” Janefalkar explicated.
Although this application of machine learning is relatively basic, it’s still essential to assuage customers who contact organizations for pressing issues related to financial services. Moreover, this application of machine learning enables financial organizations to predict their wait times based on current, as opposed to historic, data.
These cognitive analytics also enable organizations to facilitate a host of additional functionality with their contact centers in financial services. “Neural networks from our calls and chats that are originating not from an IVR call but are originating from, say, iOS and Android SDKs as well as web SDKs, we can do a lot more,” Janefalkar mentioned. “We can do, based not only just on phone number but email address, locale and location, we can do what was the last state of their account on their app before they started this.”
Machine learning is also one of the fundamental technologies underpinning the deployment of intelligent botsfor the auto insurance adjudication use case Mahalingam referenced. The crux of deploying bots for this use case—or most others for such intricate work processes—is to utilize a bot for each discreet aspect of the process. Thus, for the claims adjudication use case, “It’s deep machine learning,” Mahalingam clarified.
“The AI workflow is basically ML driven workflows. It runs through its own data modeling, it projects at this ratio when underwriting, it projects the probability of getting approved, all that stuff is driven by an ML engine.” In this use case, organizations might rely on separate bots for translating languages, serving as user interfaces, accepting or declining declarations, and ultimately issuing judgment. Buttressing these bots with advanced machine learning techniques greatly increases the propensity of these bots to aid in this process to do “auto adjudication,” Mahalingam revealed. “Typically, chatbots can only accept declarations; they cannot adjudicate.”
The ability to actually issue resolutions, as opposed to simply issue diagnostics for business problems or workflows, is a critical factor for equipping intelligent bots with advanced machine learning—particularly with solutions that incorporate their own machine learning engines. Regardless, whether simplifying and expediting complex workflows such as auto insurance adjudication, accurately predicting wait times based on current data, or routing incoming contact based on natural language, there are a host of ways in which AI is impacting financial services today.
Jelani Harper is an editorial consultant servicing the information technology market, specializing in data-driven applications focused on semantic technologies, data governance and analytics.