AI Time Journal
Pavel Gimelberg is the Principal, Head of Intelligent Automation at EPAM, where he is involved in helping insurance enterprises automate their business processes from claims automation, claims review and communication with customers.
In this interview, Mr. Gimelberg provides insights on how disruptive technologies such as Artificial Intelligence and Robotic Process Automation (RPA) can be leveraged by insurance companies to automate their business processes to drive efficiency.
Mr. Gimelberg is also a speaker at the Insurance Analytics & AI Innovation Asia Pacific conference on June 26-27th in Hong Kong.
Which processes does EPAM’s technology automate or streamline for insurance companies?
Some of our most recent projects focused on:
- Claims automation, particularly in claim registration and first notification of loss, building case files through information extraction and structuring from inbound documentation (e.g. medical documentation, loss adjuster reports, legal service provider updates)
- Claims review, validation and decision-making
- Customer communication
- Fraud analytics, such as risk assessment, scoring and triage
We look at these processes through three lenses: 1) client experience improvement, 2) business process reengineering to achieve efficiencies, and 3) intelligent automation to enhance customer experience and achieve process efficiency.
What is RPA? Why is RPA important?
The term Robotic Process Automation (or RPA) is a bit misleading as there are no physical robots involved in this technology. In essence, RPA involves software scripts that execute on virtual or physical machines and operate mostly on the UI level to mimic operators’ behaviour. The importance of RPA lies in two main areas. First, RPA technology provides much faster end-to-end automation for clients compared to traditional IT automation and therefore makes operational automation feasible that was previously out of reach. Second and most importantly, RPA changes the mindset of operations people, as they become masters of their processes, more involved in automation and realise their great efficiency ideas.
In which domains do you see the biggest impact for RPA in the coming years?
We believe that every company in every industry can benefit from RPA, but there are two types of companies that can see the greatest benefit. The first is high-volume, low-margin operational industries where there’s an environment in which operators fight to save each minute of work. By streamlining their processes, making them RPA-friendly and then subsequently automating, we can unleash efficiency benefits of up to 90% per process. The second industry type involves those with large amounts of working capital and long cycle time, such as wholesalers. For these industries, RPA can help to reduce cycle time. Reduction of cycle time by even 1% can unfreeze millions of dollars in assets and create great competitive advantage.
How many data scientists and/or machine learning engineers do you have on your team?
We have over 100 machine learning engineers in our Intelligent Automation practice who are dedicated to process automation for our clients. EPAM has over 1,000 data scientists, machine learning engineers and big data specialists in our larger Data Analytics practice who are focused on bringing value to our clients by building robust solutions in data organisation, retrieval and analytics and providing valuable insights.
Which technology stack are your experts on and are you leveraging the most?
We specialise in digitally orchestrating all technologies so that our clients can benefit from an integrated modern technology stack across their platforms and solutions. Our experts love to solve complex problems that impress our clients. We have strong expertise in software development, enterprise applications, data analytics, machine learning and intelligent automation.
In which industries, other than insurance, are you applying your technology?
EPAM’s heritage lies in the software development industry where the company started 25 years ago. We call ourselves developers’ developers because we’ve helped some of the largest global software companies develop their software solutions. Over these 25 years, we’ve continued to add capabilities and expand our knowledge in various industries, including Financial Services, Travel and Hospitality, Retail, Media and Entertainment, Automotive, Life Sciences and Healthcare. In addition to our industry expertise, we have strong horizontal delivery practices across all of our locations, which gives us an edge in applying ideas and technologies across industries. This cross-industry knowledge gives us the ability to solve complex problems as we think outside of the box/industry.
What are your team’s biggest achievements in the last 12 months?
Our biggest achievement is always when we “wow” our clients. Sometimes it’s when our key engineers come up with an idea to look at a problem differently and deliver outstanding results. Recently, we worked with a client to deliver a combination of machine vision technology for data extraction and enhancing the extracted data, which vastly improved accuracy rates and provided huge efficiency gains for the client compared to how the work was previously done. Another example is in intelligent automation, where we were able to deliver end-to-end automation in an extremely tight schedule of just a few weeks, while similar projects in other companies would have taken months.
Where do you see the biggest potential of applying AI in the insurance industry in the coming years? Which trends do you think are the most significant?
The insurance industry is facing a significant amount of disruption as new start-ups are threatening traditional players. There are two main areas where AI has potential. The first is using AI to transform the core of the insurance business by dramatically changing the customer experience, such as processing claims in a matter of seconds or applying non-conventional practices to better calculate risks so companies can outprice their competitors. The second area is process improvement, where there are three distinct applications of AI:
- Machine vision technologies to help extract data from semi-structured forms to process them automatically
- Machine learning technologies to automate complex decision-making
- Natural language processing or conversational agents to enable bots to handle customer interaction
All three of these applications combined with traditional RPA provides unlimited scalability, reliability, consistency, and compliance in our automated process, vastly improving process efficiencies in traditional insurance companies.
What is your experience/opinion of applying Chatbots & Conversational AI in insurance? What could be the benefits/challenges?
This is a very interesting question. Chatbots and conversational AI are some of the areas that are more difficult to place properly on Gartner’s technology hype cycle. Back when the technology gained popularity in 2015, there was a lot of hype around the capabilities of chatbots that everyone was interested to try. Around 2016-2017 was considered the year of the chatbot. Most chatbot implementations struggled to get past the POC phase and the ones that managed to move into production are still heavily abused by users. The best chatbots we’ve seen do not have text interaction but instead guide the user through a set of limited choice questions, buttons or carousels. Therefore, it might even be in the very early stage before the hype or already past the hype in the trough of disillusionment.
It’s our belief that the biggest benefits of chatbots are in revenue generation. With chatbots, a company can essentially tap into the whole market without needing to proportionally grow its salesforce. Revenue-generating cases for chatbots have essentially unlimited potential, while costs are predictable and controllable. On the other hand, internal optimisation through chatbots, such as internal support functions or customer support chatbots, can hit the wall of ROI quite quickly. There’s only so many straightforward simple use cases that can be handled by bots.
The current technology behind chatbots is not strong enough to handle complex conversations and we need at least one more breakthrough in natural language understanding to get chatbots up to the task. Otherwise, by now, we would have seen chatbots that are much stronger than pizza order-taking.
The original article can be found here.