Is Artificial Intelligence relevant to insurance? – IBM Blog

I love the game of chess and was shocked when IBM’s Deep Blue chess-playing machine defeated the world chess champion in 1997. That ground-shaking event divided the world with excitement and trepidation about a future with thinking machines. In this first of two posts, I investigate the anatomy of artificial intelligence and its impact on insurance.

The fascination of artificial intelligence

History has shown that the human preoccupation with creating a machine which replicates human thinking had in fact been simmering for centuries. In the late 18th century, The Turk, a chess playing machine captured the attention of the world. It turned out to be a fraud, with a human player behind the machine. In 1847, George Boole first described a formal language for logic reasoning and in 1936, Alan Turing described the Turing machine.

Along with the explosion of information technology in the 1950’s, the term ‘artificial intelligence’ (AI) was coined by John McCarthy in 1956. After the success of Deep Blue, IBM again made the headlines with IBM Watson, an AI system capable of answering questions posed in natural language, when it won the quiz show Jeopardy against human champions. Continued advancement in AI development has resulted today in a definition of AI which has several categories and characteristics.

The early versions of AI were capable of predictive modelling (e.g., recommending similar Netflix shows based on your previous choices) or robotics (e.g., developing a distance map of objects around a vehicle to enable semi-autonomous driving). Soon after, AI’s capabilities extended to Speech and Natural Language processing, such as with IBM Watson, and for Image Recognition, which is now ubiquitously used for unlocking phones and other biometric security. The four categories of predictive modelling, robotics, speech and image recognition are collectively known as algorithm-based AI or Discriminative AI. It represents AI that can sift through data and divide them into classes (of attributes) by learning the boundaries. It is used to return a prediction or result based on conditional probability.

Emergence of Generative AI

More recently, a new category of AI has emerged to stir the imagination (and fear) of humankind. Generative AI is artificial intelligence that can create new content. It has taken the world by storm. ChatGPT acquired 100M users within two months of launch. Google, Microsoft, Snapchat, and Salesforce release rival products shortly after. Academia is in an uproar over originality of authorship, and governments have even started to outlaw its general use.

Whilst Discriminative AI sought to classify data by its attributes to recognize an object, Generative AI seeks to map the distribution of attributes on examples and manipulate those attributes to create new examples. This ability to manipulate attributes and create new examples has added a new dimension to AI—creativity.

Artificial intelligence applied to insurance

The insurance industry has always made extensive use of data and algorithms, such as in the calculation of insurance premiums. The insurance business model itself is predicated on the use of mathematical and statistical methods to process personal and non-personal data to underwrite risks and price insurance policies, to quantify losses, to pay customers’ claims, and to identify and prevent insurance fraud. The impact of AI, both Discriminative and Generative, has immediate and long-term effects on the business of insurance.

The deployment of AI can help insurers in multiple aspects, from underwriting, to claims, customer service and fraud prevention. Below are some typical use cases and demonstrates the primary impact on the automation of internal processes and on improved customer service.

Customer service and conversational AI

This is an area where insurers are most advanced in its early adoption of AI. Conversational AI, based on natural language processing, can interpret spoken and written human language and respond accordingly. It offers customers and the insurer’s system to interact in a human-like manner. Chatbots and voice assistants are already offering round the clock service whilst maintaining quality of service. We will continue to see more advanced and specialised conversational AI developed to handle more complex dialogue particularly in claims handling. Generative AI will make the conversations more expedient and relevant.

Claims automation

AI tools in the claims handling process can expedite the handling of claims and lead to faster settlement. AI’s Image recognition can automatically read, interpret, and process documents and images (e.g., extracting information from medical records, recognise vehicle types or evaluate damage). By collecting large amounts of historical data, Discriminative AI can be used to make plausibility assessments and ensure quality and uniformity in the adjusting process. Complimentarily, Generative AI will be able to help the adjustor summarise the data and generate a preliminary report.

Fraud detection

AI can be used to analyse large amounts of data from multiple sources to spot unusual patterns as indication of fraud. Pattern recognition on vehicle damage data can be used to detect cases of fraud. It can also detect manipulated images that would raise suspicion.

Pricing and underwriting

AI offers new possibilities in the pricing and product design of insurers. With the combination of data, new risk characteristics can be developed to provide more accurate insurance cover. With the willingness and consent to share one’s private data, products can be tailored more precisely for each customer.

For example, the increasing availability of medical data, in combination with medical progress now makes it possible to offer term life insurance for people with serious pre-existing conditions. Leveraging Generative AI’s ability, a unique and personal life insurance policy can be underwritten with contributions from personal medical data. Beyond medical data, other public data such as meteorological data, using AI’s ability to process large data is having effect on property insurance.

The road ahead

The use of AI in the insurance industry today is still nascent. AI is still an emerging technology and the road to implementation will have challenges. However, the use of AI in society is becoming prevalent. Insurers must adopt AI to stay relevant to their customers and draw down on the cost saving benefits of adopting AI in the near term. Ignoring AI is costly. Take a step towards adopting AI. List down your ideas for how AI can improve the way insurance is managed. In the next post, I will explore the limitations and challenges we face with AI, and how we can mitigate them as we implement and scale our use of artificial intelligence.

Discover more blogs about artificial intelligence

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *