A Primer for Artificial Intelligence in Insurance

Over the last decade, Artificial Intelligence or AI has become synonymous with new technology. Broadly it is a term used by many to mean many things.  In its simplest form, AI is the ability of technology to perform tasks associated with intelligent beings. Inborn characteristics of humans, such as the ability to reason, think and configure, discover meaning and insights, and learn from past experiences – are complex tasks that new-age technologies like AI can undertake and solve with excellent proficiency.

To understand AI is to go deeper into intelligence and what meaning we derive from it. Psychologists characterize human intelligence as the inherent trait to combine diverse abilities- learning, reasoning, problem-solving, language, and perception. Most living beings can use these traits singularly, including animals and insects. But combining them into a single powerful feature lends a complexity that only humans can leverage. AI technology replicates this, constantly evolving to undertake simultaneous large and complex functionalities to make future decisions and take action.

Since AI is purported to make machines emulate human-like functionalities, the degree to which AI can replicate human capabilities is the criteria used to determine AI types.


AI can be classified;

  1. Based on Functionalities: This classifies AI based on its ability to think and likeness to the human mind. There are four types of AI-based systems in this category:
  • Reactive Machines – One of the oldest forms of AI, their capabilities are minimal and only based on responding to stimuli. They do not have memory functions, cannot remember and store information, or learn and give insights based on past knowledge. A famous example of this machine is IBM’s Deep Blue, which beat chess Grandmaster Gary Kasparov in 1997.
  • Limited Memory Machines – Besides having reactive capabilities, such machines can make decisions based on historical data. Most present-day systems that use deep learning are trained by large volumes of data stored in memory as a reference model to solve future problems. Chatbots and Virtual Assistants are typical examples of this.
  • Theory of Mind – Touted as the next level of AI systems currently being innovated, these systems have been proposed to understand better the entities they interact with, discerning their emotions, feelings, beliefs, and thought flows. While artificial emotional intelligence is just a theory today, developing machines to understand humans is the next step in AI evolution.
  • Self-awareness– Only hypothetical for now; these AI systems have evolved to think like the human brain and become self-aware. This is the ultimate objective of AI research, making a machine capable of experiencing emotions, needs, and desires.
  1. Based on Capabilities: This alternate system of classification separates AI into:
  • Artificial Narrow Intelligence (ANI) – All existing AI, including the most complicated and capable ones, fall under this category. ‘Narrow Intelligence’ refers to AI that can only perform one task autonomously using ‘human-like’ capabilities. They cannot do anything more than they are programmed to do, have a narrow range of competencies, and are more reactive or ‘limited memory’ AI.
  • Artificial General Intelligence (AGI) – The ability of AI to learn, perceive, understand, and function as a human being gives it enormous, independent competencies, thereby reducing training times. This makes AI almost like humans in our way of functioning.
  • Artificial Superintelligence (ASI) – This will mark the pinnacle of AI development when it manifests for real. ASI will be the most capable form of intelligence on earth, better than human beings in everything – more extraordinary memory, faster data processing, quicker analysis, and decision-making capabilities. The development of AGI and ASI will also lead to a scenario called ‘singularity,’ with massive intelligence and implications across industries.


Regarding applications in practice, AI can be classified based on where and how it is used.

  • Analytic AI – Powered by Machine Learning and Deep Learning techniques, this system scans tons of data for patterns and dependencies, producing recommendations that provide business insights to support data-driven decision-making. Sentiment Analysis and Supplier Risk Assessment are just two examples of this AI, performing functions like inventory optimization and demand forecasting in the manufacturing and FMCG industry.
  • Functional AI – Very similar to Analytical AI, this system scans large amounts of data to derive patterns and dependencies. However, it doesn’t give recommendations but also acts based on them. For example, it can pick up machine parts breakdowns when deployed on cloud software and trigger commands to shut the faulty part down, going beyond just raising alerts.
  • Interactive AI – As the name suggests, this system automates communication without compromising interactivity, like chatbots and personal assistants responding to prompts, providing relevant responses to questions, and understanding conversation contexts.
  • Text AI – This system recognizes any data in text form, speech-to-text conversion, machine translation, and content generation. AI-powered text databases go beyond conventional keyword search capabilities to find contextual content and correlate information. Thanks to Semantic Analysis and Natural Language Processing or NLP, the AI builds semantic maps, recognizing synonyms to understand user context.

For example, Charlee.ai’s A.I.-Based Predictive Analytics engine sifts through millions of past claims and unstructured data to accurately predict claims severity and litigation based on a continuous spectrum of pre-trained insights. It can also identify over 160 + insurance fraud schemes and meet deadlines outlined in the Fair Claims Practices Act to help insurance carriers stay compliant.  

Visual AI– Using visual aids to identify, classify, and sort objects or convert images and videos into insights is what this AI system. Helping insurers estimate damages to automobiles based on damaged car photos is the capability of this AI, which also covers augmented reality.


AI capabilities are developing rapidly, and their intervention is helping solve hitherto impossible scenarios and identify low and high-severity claims, like in the insurance industry, with quick intervention while bringing down costs and time wastage in mundane, repetitive, error-prone tasks. Businesses already use it across industries in various capacities, and leveraging it has multiple long-term benefits.

Written by: Dr. Charmaine Kenita

Technology and Business Writer


John Standish

Co-founder & CIO


What Products Are You Interested In?