Insurers’ Journey to Better Decisions Begins with Foundational AI/ML Thinking

By Jennifer Herz, Gautham Nagabhushana, Jean-Louis Arsenault, Juan Rosa Medina Touzard and Bharanidharan Sridharan

The handwriting is now on the wall. After lagging industries like life sciences and banking, insurance is fully embracing artificial intelligence/machine learning and exploring the possibilities[i]. Insurers need to take the first step to recognize the power of data, by rationalizing and modernizing the data residing within their ecosystem and beginning to experiment with AI/ML to achieve this. AI/ML unlocks the insights that lie within your data to help you target the right customers with more compelling products at greater efficiency and better margins, fueling growth and innovation. There is no limit to the potential gains. Just replicating the success stories from other industries can result in massive cost savings, but recognize getting here takes vision, patience, and fortitude.

With the proliferation of new data sources, AI/ML is quickly becoming table stakes[ii] to evaluate risks, grow products and further distribution opportunities in the insurance industry. For life and annuities, AI/ML is helping the industry get ahead of new exposures like COVID-19 and better predict key areas of risk (smoker propensity models, BMI models, predictive models focused on cardiovascular disease, for example). Within group insurance, AI/ML is being deployed to support the evaluation of experience ratings, allowing for fine-tuning to drive competitiveness. For property and casualty, AI/ML are driving new services, helping to identify new distribution pathways, accelerate underwriting and claims process management, and helping retain customers while continuing to help mitigate risk. The numbers are impressive. We recently helped a commercial insurer client identify roughly $65 million in underwriting submissions for a three-month period that it claimed had not received proper approvals. Do we have your attention yet?

We’re not saying AI/ML is not scary. Frankly, it is. It has the potential to impact how insurance is constructed, risk is evaluated, and claims are paid — it will impact people, process and technology and that needs to be factored into any approach to build and mature these capabilities. This a highly complex set of technologies[iii] that are not well understood by many who do not have “PhD” after their name. Nonetheless, the desire to execute AI/ML projects in insurance is very high. Implementing AI/ML, like any technology, is a journey. Before you can even get out of the starting gate with a pilot, you have to have a set of “high-quality” data you can use to start to test and learn. Your ecosystem needs clean and actionable data that is complete, that is stored and interlinked properly. But, how do you know what forms the foundation for this journey?

If you’re like most insurers, leveraging AI/ML for intelligent decision-making means you will have to shore up and modernize your data foundation[iv], including data governance and master data management. Putting in place a foundation of solid data and governance processes will enable you draw actionable insights from a wide variety of traditional and non-traditional data sources, including your own internal databases, wearables data, auto telematics, building sensors and third-party data. This will entail reviewing standards, consistency, interfacing, external data, and compliance. Once this data foundation is in place, the AI/ML journey can begin.

That brings us to the question: How should you begin your AI/ML journey?

What keeps you up at night?

For any business, there are unknowns that, if revealed, could help kick off better decisions. In insurance, if you have your sights on growing revenue, you likely want to know how you can be more effective with your products and service offerings. Consequently, your questions might be things like, “Who am I targeting?” and “How much will that customer pay?” “How would that customer like to engage with me?”

Having access to quality data enabled by AI/ML allows carriers to design insurance products tailored to an individual’s needs, improving customer experience in the process. Customers are looking for personalized recommendations that help anticipate needs based on what we know about them. Customers may want more choice and options, as well as guidance for what coverage they need at different times for different potential risks. For example, predictive or prescriptive recommendations and pricing based on extracted customer data would yield a personalized policy.

Alternatively, you might want to focus internally on improvements to your underwriting, actuarial, claims, and product distribution. That would lead to questions such as, “What’s the right way to distribute this product, through traditional channels or direct to consumer?” “Who are the good risks and where are they located, what data helps us identify profitable segments or micro-segments?” For example, smoker propensity modeling, geo/hazard modeling, etc.

On the other hand, your burning questions might revolve around improving your operations. Hence, you may want to know, “How can I use data and models to optimize operations?” “Where can I automate more service tasks and service responses to increase my efficiency?” “Which processes can go straight-through?” “Where can we cut response times and increase customer satisfaction?” From digital assistants that perform mundane customer service tasks to image processing of aerial drone photos and videos, IoT, Deep Learning analysis and extraction of text, carriers are quickly adding the use of AI/ML in reducing costs and fraud, improving their balance sheets.

These are high impact, high value questions insurers are trying to answer, which is why to unearth any insights and apply AI/ML, it is imperative to focus on the quality and create governance around your data. However, in order to uncover the insights you are looking for, you will need to address internal data quality issues, identify external data sources that can help and put governance and MDM in place to ensure quality.

Our recommended approach is:

• Start small and build.
• Identify a few high value use cases.
• Source targeted subsets of internal and third-party data to support the use case.
• Assess the quality and strength of each data source as you build your models.

Ignore AI/ML at your peril

Insurance by its nature is a conservative business that does not easily embrace change. But you can no longer afford to ignore the fact that AI/ML will be key to your competitiveness going forward. You can’t unlock competitive advantage without getting a handle on your data, which needs to be current, accurate, and actionable.

Building AI/ML capabilities is a rewarding journey. The average ROI on AI investments is currently 1.3%[v] but with a significant potential to reach higher returns in the medium to long term. However, recent experiences by leading carriers shows that outcomes outweigh the risks.

When it comes to starting out with AI/ML, both technology and business executives need to consider the following:

Set up of AI/ML

• It is imperative to think strategically when you identify the key business use cases and overall business objectives of your AI/ML initiative. Set priorities but be agile in your approach.
• The data architecture is key. Getting that right requires understanding the data sources that need to be leveraged, determining the size and format of the data and how to streamline it.
• Master Data Management (MDM) and Governance are other critical pieces of the puzzle. Develop an MDM and Governance process stewarding the data to ensure it can be leveraged in a repeatable, scalable, and sustainable fashion.

How to get AI/ML started

• Your first AI/ML project should not target ground-breaking capabilities, regardless of the pressure to show quick results. AI/ML capabilities are built as layers. For instance, when Google built its mapping application, it first built a robust mapping software with longitudinal and latitudinal data of big cities to then add navigational capabilities and place-of-interest layers. All good AI/ML offerings should be built in small chunks as layers.
• Assess the quality of each data source — not all data is created equal! Do a “quick and dirty” iteration — with available or quickly collectable data — and go through several stages of refining data and algorithms. AI/ML platforms and solutions can rapidly analyze huge volumes of different types of data that cannot be analyzed by simple analytical models. They can guide you to which set of data sources is the most useful to be pursued, collected and analyzed. Cognizant’s evolutionary AITM [vi] addresses this typical challenge and accelerates your AI programs. For many carriers, starting small, with structured focused pilots are a great way to get started to create that test and learn culture, explore and test the quality of the test, and gauge where the organization is in terms of embracing a data centric culture.

This is a big undertaking, but don’t be afraid to jump into the AI/ML pool. Go in with sound data and questions you would like to have answered — as well as realistic expectations — and know that you’re taking steps to safeguard your company’s future.

Stay tuned for our next post in this series: The journey to becoming a data-centric organization: make your organization data and AI/ML-ready

[i] https://risk.lexisnexis.com/insights-resources/research/state-of-ai-ml-in-the-insurance-industry
[ii] https://www.mckinsey.com/industries/financial-services/our-insights/insurance-2030-the-impact-of-ai-on-the-future-of-insurance[iii]
https://www2.deloitte.com/content/dam/Deloitte/de/Documents/Innovation/Artificial-Intelligence-in-Insurance-Whitepaper-deloitte-digital.pdf
[iv] https://www.cognizant.com/whitepapers/modernizing-insurance-data-to-drive-intelligent-decisions-codex6348.pdf
[v] https://www.cognizant.com/whitepapers/ai-from-data-to-roi-codex5984.pdf
[vi] https://www.cognizant.com/us/en/ai/evolutionary-ai

About the Authors

Jennifer Herz is the Partner, North America Insurance Consulting, Underwriting Center of Excellence Lead with Gautham Nagabhushana, Practice Area Business Partner/Leader, North America Insurance Intelligence. Jean-Louis Arsenault is the Director, AI & Analytics/Digital Transformation, Juan Rosa Medina Touzard is Manager, Insurance Consulting and Bharanidharan Sridharan is the Consulting Manager, Life and Annuity New Business and Underwriting Transformation.