Evolution of Technology: Is AI a threat?

Reeshabh Choudhary
9 min readFeb 15, 2024

Introduction

Back in 2020, when the world was grappling with Covid-19 waves across the world, big players of Insurance domain were dissecting how they can continue delivering value to their customers in no contact era. This was a big shift from the usual norm for Insurance industry, which has been dependent on their human work force to generate business and connect with their clients. They were potentially discussing the automation of their CRM wherever possible. KPMG produced a paper “New reality for insurance” in July 2020, discussing which areas of Insurance Value Chain would potentially be reformed and scope of AI/ML integration in their CRM.

Customers prefer products which make their lives easier and help them get things done without any hassle. Insurance domain has always been integrating the best of enterprise technology so far and making lives easier for customers as well as their own representatives. However, now there was a discussion on potential replacement of human representatives with computers or bots. Rather than reaching out to customers and converting deals based on human connection, a totally new pathway was being discussed. The KPMG paper discussed about how they want to automate certain steps in Insurance value chain, such as Claim Management with AI integration and equip their own workforce with better technology so that they can focus on larger picture and efficient decision making. But the fear of potential replacement of human representatives has definitely set in. Especially in time of Covid, when business was totally shut and there was no way to reach out and work to retain clients. In countries like India, majority of human representatives in Insurance domain are not even on permanent position but on contractual basis, with the conversion of clients being the sole propellant of raise in income.

Is the Ghost real?

3 years have passed since and there has been a lot of technology innovation in the meantime. Companies have definitely moved on to digital process of onboarding new customers. Data gathering is much more structured and streamlines than earlier. And since data gathering is streamlines, data analytics has been gaining traction. Now decisions are not made just on observations and intuitions but are backed by data, and they say data never lies. But when it comes to reaching out to clients and converting the business, are humans being replaced by bots? Definitely, there has been trend on integrating chatbots in customer facing websites. But what is their conversion rate? Is it better than earlier or worse? Are these chatbots more efficient than human representatives? Has AI really taken over the jobs of humans or is it taking baby steps to potentially eliminate its ultimate competition, its own creator? Or it is just another technological advancement which is aimed at making life easier?

Automation, including Artificial Intelligence (AI), is increasingly being adopted in the insurance industry to streamline and enhance various processes, including claims management. Self-adjudication refers to the ability of systems to automatically assess and make decisions on claims without human intervention. Automated claims handling involves end-to-end automation of the claims process, from submission to settlement, using advanced technologies like AI, machine learning, and robotic process automation (RPA).

For tasks such as document recognition, fraud detection, and processing claims data, we leverage Machine Learning and AI for faster and more accurate decision-making. This definitely reduces the reliance on traditional claims personnel for routine and repetitive tasks. Instead, the claims personnel can focus on more complex tasks that require critical thinking, empathy, and nuanced decision-making. However, there is always a consideration that the decisions made by AI systems need to be fair and unbiased, avoiding discriminatory practices.

Let’s meet the ghost.

What is AI? Any logical task which humans can do repetitively, machines can do it better, and this is artificial intelligence. AI is enabling computers to learn and make intelligent decisions based on data. For example, rather than categorizing emails in your mailbox yourself, you train the machine to categorize emails based on their content.

AI/ML implementation and usage in Insurance domain

Let us discuss briefly how an AI model is used.

Insurance companies gather vast amounts of data related to claims, including information about policyholders, historical claims, damage assessments, and various other relevant details. Raw data is often noisy and may contain errors. In the preprocessing step, data is cleaned, transformed, and organized to ensure its quality and relevance. Features or variables that are crucial for assessing claims are extracted from the data. These features could include information like the type of incident, location, date, and historical claim patterns.

We train AI models on historical data. The model learns patterns and relationships within the data to make predictions or classifications. There are different algorithms which can be employed for different problem sets for the model to learn from data. For claims assessment, algorithms like decision trees, random forests, support vector machines, or neural networks may be suitable. Once a model is trained, next step is to fine-tune and optimize for performance improvement. Evaluation metrics, such as accuracy, precision, recall, or F1 score, are used to assess how well the model is likely to perform on new, unseen data. The model is now ready to analyze new, unseen claims data and make predictions or assessments based on the patterns it has learned during training. Modern systems incorporate mechanisms for continuous learning. As new data becomes available, the model can be updated to adapt to changing patterns and improve its accuracy over time.

With the advent of GPT models or LLMs, there has been lot of speculation about AI taking over most of the jobs. However, there is always another side of the story. While these LLMs excel at tasks such as text completion, language translation, and question answering, they might not be the ideal choice for tasks that involve structured data analysis, such as claims assessment in the insurance industry. The strength of GPT models lies in their ability to generate coherent and contextually relevant text based on the input they receive. They are trained on a diverse range of internet text and can understand and generate human-like text across various topics. However, they lack specific knowledge about structured data, such as numerical or categorical data typically found in insurance claims. For tasks related to claims assessment, traditional machine learning models or specialized algorithms designed for structured data analysis are more suitable. These models can effectively process and analyze numerical and categorical features associated with insurance claims, making them more appropriate for tasks involving data-driven decision-making.

Having said that, there could be scenarios where GPT models play a role in the broader claims management process. LLMs can be leveraged for processing and understanding unstructured text data, such as claim descriptions or communication between insurers and claimants. Chatbot and virtual assistance experience can be enhanced by integrating LLMs to handle customer inquiries, provide information about the claims process, or generate responses to common questions.

When it comes to adoption of any technology, it is all about finding areas of improvement and automating mundane tasks.

Where are we heading?

It is imminent from our discussion so far that AI systems can analyze large datasets more efficiently than traditional manual methods and will definitely be at forefront of decision making in Insurance domain or any such domain, where decision making is subjected to huge amount of data processing. AI models can easily identify patterns and figure out relationships within data than their human counterpart. There is no denying that with this kind of efficiency, it is more suited to replace a substantial human workforce for repetitive and mundane tasks. But this is what machines have been doing through out human history. Technological advancements have consistently aimed at automating routine activities, thereby freeing up human labor for more complex and creative endeavors.

A century earlier, the process of weaving fabric was a labor-intensive task carried out by skilled weavers using hand looms. Each step of the process, from setting up the warp to manipulating the shuttle, required careful manual work. With inventions like Jacquard loom, the speed and efficiency of the weaving process significantly improved. While it did displace some traditional hand weavers, it also led to the creation of new job opportunities in machine operation, maintenance, and design. History is full of such examples. When Henry Ford introduced assembly lines for car manufacturing, it revolutionized the automobile industry by significantly increasing production efficiency, reducing the time and effort required for each task. Computers have been automating tasks for us humans right from their advent. Word processing, spreadsheet software, and email systems streamlined administrative processes, allowing workers to focus on more complex cognitive tasks. Now a days, Robots are employed in tasks such as welding, painting, and assembly, freeing human workers from repetitive and physically demanding activities. Even the field of agriculture has been adopting technology throughout the course of human history. Usage of tractors and harvesters have automated many farming tasks, which helped farmers to shift focus on better crop management and decision making.

Technology has always been a disruptive force and have challenged established order and processes. However, as new technologies are integrated and adapted, a new order emerges. This new order often brings about increased efficiency, convenience, and improvements in various aspects of life. From the invention of the wheel to the development of the internet, each technological leap has contributed to reducing entropy in the way we live and work. For example, advances in transportation technology, from trains to airplanes, have reduced the entropy of geographical barriers. What was once a cumbersome and time-consuming journey has become more efficient, connecting people and cultures across the globe. Vaccines, antibiotics, and medical technologies have contributed to longer and healthier lives, reducing uncertainty in the face of diseases. Sustainable energy solutions and eco-friendly practices work towards maintaining a balance in natural systems, reducing the disorder caused by pollution and resource depletion. AI systems are not going to be any different. It aims to empower humans for better decision making and relieve them from boring mundane tasks. We humans have finite energy and time, and we must better put them to use in constructive tasks.

Humans are social creatures, at least for the foreseeable future! Most of our decision making is based on the social connections we form and the society we live in. Yes, as a customer we like prompt responses to our query, but when it comes to closing a deal, we would like to know who is sitting behind the screen. Whom we are going to blame or hold accountable, if things go wrong. Trust plays a crucial role in decision-making. Knowing the individuals behind a product or service creates a sense of trust and reliability. People are more likely to engage in transactions when there’s a perceived connection with the individuals or entities involved. And this connection can be enhanced with better experience, and this is where the confluence of AI and humans lie. From recommending products to tailoring marketing messages, AI can help deliver a more personalized and satisfying customer experience. When it comes to adoption of AI based tech-stack in Insurance domain, story is going to be no different.

AI algorithms can analyze vast datasets to identify potential customers based on their demographics, online behavior, and historical data. Human agents can use these insights provided by AI to personalize communication, address specific customer needs, and build relationships during the sales process. AI models can assess risk more accurately by analyzing diverse data sources, including social media, telematics, and historical claims data. Underwriters can utilize their expertise to interpret complex risks, make nuanced decisions, and ensure that ethical considerations are taken into account. AI algorithms can recommend personalized policy options based on individual customer profiles and preferences. Human representatives can work with customers to explain policy details, address specific requirements, and provide guidance on coverage options tailored to the customer’s unique circumstances. Routine queries can be handled by AI powered chatbots by providing instant response and human customer service representatives can step in for more complex queries, complaints, or situations requiring empathy, understanding, and personalized assistance.

There is always going to be a symbiotic relationship between humans and technology. It is all about how cognizant we are in making use of technology. Till the time human emotions and social connections are the driving force of the society, we must not fear technological advancements, as they will be the tools to bring order to the chaos. The dystopian future of machines taking over humans can remain fiction!

--

--

Reeshabh Choudhary

Software Architect and Developer | Author : Objects, Data & AI.