The AI Road to Business Value for Insurers

Tech enabling innovation in insurance is not a new thing. However, the insurance industry is lagging behind when it comes to tapping into AI's (Artificial Intelligence) potential, compared to other industries like retail and manufacturing. Rapidly changing customer expectations are pushing traditional insurers to react to the new demands and explore possibilities of the digital era.

Tech innovations are swiftly changing the competitor landscape

Insurance value chain has always been "process" driven, primarily due to a bulky risk assessment landscape, stringent regulatory framework, and a lack of transparency between players, all leading to poor customer satisfaction. Due to the recent explosion of "risk relevant" data, #InsurTech #startups are able to find sweet spots in the insurance value chain. Digital technology advancements are helping start-ups challenge traditional business models and promise an agile value chain, leading to improved customer satisfaction. Some incumbents claim #AI as an enabler for disruption while others are still struggling to gain real value from AI advancements—the debate goes on. But we can't ignore the fact that AI is already a part of our normal lives in so many subtle ways, be it using your finger or face to unlock devices, or automatically receiving personalized news/social media feeds without any conscious effort. Changing customer behavior and new digital expectations is intensifying competition between #InsurTech startups and established insurance players. Risk knowledge companies like #SwissRe are leading the way by using new technologies to anticipate and mitigate risks, creating smarter solutions. A perfect combination probably lies in marrying the power of technology with the knowledge of risk insights, keeping the customer at the center.

Data is the new currency that fuels AI

Millions of terabytes of data are created every day from both traditional business interactions and new data sources, like connected devices and sensors. Meaningful processing of data is crucial to unlocking its business value, and AI algorithms have been progressively maturing in supporting day-to-day business decisions. We all have more data today than we did yesterday, but turning data into knowledge and then knowledge into action needs combining the technological view with a business view. We need to develop an understanding of what's feasible technically and how it can generate business value. Successful AI applications are still limited to well-defined tasks/problems; therefore, we should always start with defining a use case that represents the business view of an AI enabled application. A good use case must serve one or more business functionality, have a high volume of good quality data, and generate value for the customer (internal or end customer).

Supervised learning models, which rely on human labeled data, are highly successful in automating repetitive tasks, leaving human experts with more time to focus on high value-adding tasks. Replicating known answers by using AI applications is yielding both top and bottom line improvements. An AI Machine Learning algorithm learns the transformation rules to create a desired output based on a given input. This contrasts with traditional programing, where the transformation rules are specifically coded. The key question is: are we prepared to depend on a machine for reading underwriting case files and extracting key information like face amounts, client decision, and other factors that drive priority of our actions?

Let's do a myth or reality check: AI technologies can enable touch-less underwriting and claims handling – is it a myth or reality?

We've seen substantial progress in leveraging clinical and lifestyle data for developing underwriting risk scoring models, and the best is yet to come when a few winners will emerge and rule the industry. Claims automation has also witnessed increased AI adoption for fraud detection, triage/prioritization modeling, and complexity estimation, but can we really rely on AI for touch-less claims processing? This is a big question of mindset and not just technical capabilities. Recent advancements in AI, especially in image and speech processing, are opening new opportunities for the insurance industry. Low-touch handling of underwriting and claims process is no longer only a possibility but becoming a reality. Lack of access to deep domain knowledge and functional experts is driving InsurTech startups to look at alternative data-driven solutions, and it's just a matter of time before touch-less underwriting and claims handling will become the reality for the industry, and not a myth as it still is for some traditional players.

Here's an example, in a typical claim scenario: adjusters must perform several actions, like collecting loss details, validating policy coverage, and ensuring the authenticity of loss. Even in the most sophisticated claims process, the adjuster has to make several decisions before approving the payment. Most of the key decisions are based on matching existing policy information with new loss information (i.e., decisions based on data). AI has tremendous potential to not only normalize and structure the data, but also make decisions based on pre-defined conditions. How about a death claim beneficiary taking a picture of the person’s death certificate on a smartphone and submitting it to the insurer, who can then automatically validate the authenticity of the death certificate, pass it through policy terms and conditions, and process the payment? It sounds very simplistic, yet it is possible. In a more complex scenario, the chat-bots equipped with speech analysis, NLP capabilities, image analyzers, fraud detection models, etc., would ensure all needed information is collected and all aspects of claim are authenticated before payment is released.

How can you get started on the long road of AI driven operational excellence?

AI is already creating a tangible impact on insurance, and customers are demanding more— the time to act is now, in order to stay ahead of competition. The first step is generally the most important step, so how can we become future-ready today?

1.   Tap the fuel (data) – Determine what data we already have in hand, identify the use case that surrounds the given data, augment the data with external sources(both public and private partners), and ensure a consistent availability of needed data to perform the given use case.

2.   Start small and fail fast – Focus on the initial but immediate business value, don't embark on the most complex use case that maximizes business value first. Agility is the key to success, and every little success matters more than failure. Keep a check on time and cost, and accept failures.

3.   Learn and grow – AI projects are challenging not just because of the tech novelty, but more because of the cross-functional collaboration needed to succeed. There is no single formula that works for all companies or use cases, but what always works is the growth mindset fueled with necessary tools and capabilities in an environment which inculcates data culture across teams/departments and value chain.

Insures are moving from risk mitigation to risk prevention, and risk prevention measures based on internal (available) risk data combined with external consumer provided data (like wearables, EHRs) are much more powerful in increasing customer safety and enhancing customer experience. AI technologies are proving to be an effective aggregator of data from traditionally disparate sources (images, speech, sentiment, bio-metric, etc.), and a capable provider of heavy-duty data processing in deriving meaningful information to create a significant impact on business performance.