Data has been at the heart of the insurance industry since its beginings. The explosion of data globally, along with exponential increases in computing power, is driving advances in artificial intelligence (AI). This in turn is creating unprecedented opportunities, especially for insurers where AI has already started to revolutionize the industry.

How insurers use this data to improve customer experience, reduce fraud or develop new business models will determine their success or failure for decades to come. Some insurers have already taken initial steps and are experimenting with AI use cases in the areas of pricing, customer interaction and fraud detection. Additionally, an increasing wave of InsurTech start-ups are entering the market with competitive offerings that differentiate them through a novel value proposition, attractive pricing or a superior user experience.

If navigating this new landscape isn’t difficult enough, the biggest challenge companies typically face are internal, including cultural and organizational barriers as well as talent shortages, according to Harvard Business Review (HBR) (1). A closer look at AI reveals estimates by Business Insider Intelligence (2) that it could save $390 billion in costs across insurers’ front, middle and back offices by 2030. This time horizon seems far off but insurance companies that have already implemented advanced analytical solutions are outperforming the competition today by 76% (3). Popular insurance industry AI applications range from risk management, fraud detection and prevention, personalized offerings and customer churn (attrition) reduction to climate change impact prediction. On-demand and usage-based insurance (UBI) offerings will be driven by AI’s ability to aggregate disparate data sources and ultimately deliver more flexibility to customers.

The implementation of AI could lead to a 31% increase in profitability rates by 2035 (4).

Key opportunities from AI in the insurance industry:

Product development

• On-demand insurance
• Microinsurance
• Usage-based insurance

Sales and marketing

• Personalized offerings
• Next best action
• Customer churn prediction
• Increased customer loyalty and satisfaction

Pricing and underwriting

• Modular / behavioural pricing
• Granular Risk Models
• Automated application scoring
• Risk segmentation

Claims

• Claim Process Automation
• Fraudulent Claim Detection
• Claim Appeal Processing

Agile AI-driven product development

“When developing a new product, insurers can take a page out of a large digital retailer/wholesaler’s playbook: Begin with something simple and add complexity as required. Developing a minimally viable product (MVP), which starts as bare bones and grows with iterations, may eventually prove to be a backbone of agility for product development.” (5)

The same type of agile development methodology used for software development and for Unit8’s projects can be applied to develop innovative insurance products. Advanced analytics and AI are expected to become the foundation of new insurance products, allowing for the hyper-personalization of new products and offerings. This iterative process allows features to be continuously added to a product or service. By viewing additional features with this modular mindset, a flexible building-block foundation is created that evolves with new inputs. Any data that future customers might provide from health and fitness trackers (e.g. smartwatches) could be leveraged for more customized insurance offerings.

This type of UBI deviates from the insurance models people are accustomed to. The advantage is that instead of relying on aggregated statistics and past trends, it models individual and current behaviours. This customized offering provides an attractive alternative, given the breadth of lifestyles, habits, cultures and personalities. Data privacy, and more generally peoples’ perception of data usage, is still evolving and will determine the pace at which these offerings will be made available.

Customer first sales and marketing

Insurers spend billions of dollars every year on marketing to acquire new customers. Depending on the study consulted and the industry, acquiring a new customer is from 5 to 25 times more expensive than retaining an existing one. That makes sense: there’s no need to spend time and resources going out to find a new client — it’s just necessary to keep the existing one happy (6).

Acquiring a new customer is from 5 to 25 times more expensive than retaining an existing one.

Nonetheless, sales and marketing are two of the biggest adopters of AI across the insurance value chain. The new product offerings mentioned above are allowing companies to offer customized options to potential as well as existing customers. This combined with customer intelligence allows the creation of distinct personas that can be targeted along any part of the customer’s buying journey. By creating such personas and better understanding customer life cycles and buying journeys, insurers can create a “next-best-action” tool. This gives sellers and marketers clear guidance on which offerings will resonate with which customers, thereby enhancing the efficiency of the consumer-focused sales journey.

Interactions with existing and potential customers are also changing with chatbots and new levels of natural language processing. These allow insurers to engage with customers 24 / 7, helping to offload more mundane communications in an automated way. Chatbots are now much more than just interactive conversational tools for customers.

Integrating chatbot functionality in diverse insurance offerings supplies an automated insurance terminal where customers can begin and end their entire insurance buying process. The analytics behind such solutions are combined with intelligent customer relationship management and allow this complex web of interactions to be visualized and streamlined at scale.

The conversation is not one-sided, though; customers expect proactive and relevant communication from their insurance providers. Indeed, “57 percent of insurance customers around the world, across all product types, prefer to hear from their providers at least semi-annually; only 47 percent receive that level of contact currently. As AI becomes more pervasive, an increasing part of lead identification [customer prospecting], generation, and capture will be automated by AI. This will slowly become reality for many companies that are looking to personalize at scale, so the key questions to winning market share will become: ‘How do [insurers] … understand and leverage AI tools to offer the most relevant product offerings?” “Where can [they] acquire data that [their] competitors don’t have access to, to help [them] understand what [their] customers really want?” (7).

From static to innovative pricing and underwriting

With customized insurance products based on hundreds of thousands of customer personas, the pricing and underwriting of these complex insurance policies become equally complicated. Yet traditional insurance pricing models are often still based on simple matrix systems that take a small number of variables into account, leading to high exposure risk and large loss ratios. This static approach to risk modelling will be replaced by a bottom-up approach in which internal data is combined with a growing pool of external data assets (economic, crime rate, credit, traffic and weather data, for example.) This data lake, together with a machine learning framework, will enable more dynamic and forward-looking risk models.

Transforming a traditional underwriting process is complicated with no “golden model” that can easily be implemented. Additionally, a defined list of offerings no longer exists, replaced by one with offerings determined by individual customer demands and situations. But four guiding principles can help undertake such a transformation:

• Break down silos: Digitally equipped teams with diverse skill sets will be deployed in flexible and fluid team structures, straying from typically rigid departmental structures. Data scientists, actuaries, IT specialists and others will collaborate closely on new processes or products.
• Lead by example: The senior leadership needs to set the tone and drive this digital transformation. The journey will include ups and downs, but continued support and good planning will allow those who undertake it to reap the benefits.
• Focus on quick wins: New offerings based on advanced underwriting models take time, but it is best to break down the longer and more complex projects into weekly or biweekly sprints. This improves the tracking and forecasting of project timelines while maintaining employee morale through the constant achievement of smaller objectives.
• Bring it all together: Automated data collection, streamlined decision processes, digital-first strategies, more agile product development — each of these elements improves the underwriting process, but systems thinking allows synchronizing and aligning them, producing remarkable results.

“But why is my premium more expensive?” Explaining to a customer why a particular model suggests one insurance plan over another or why one person’s premium is more expensive is a key topic that cannot be ignored. The myriad models ingest so much data that humans can no longer explain why one offers a particular suggestion or price. This “black box” approach should be addressed from the very beginning. Any model used, not just for pricing or underwriting and not just within the insurance industry, must include explainability features. They enable human interpretations of the model’s decisions and an understanding of the model’s decision process.

Any model used, not just for pricing or underwriting and not just within the insurance industry, must include explainability features.

A particularly interesting data source in the property and casualty insurance sector is geospatial data, previously known as geographic information systems (GIS). Historically, this type of data was only available to large companies and countries due to the high cost of sending satellites into space. Over the years, however, these costs have dropped considerably, as has the computing power to run analytics on a wide scale. Insurers now have the ability to perform a detailed analysis of a property using machine learning and computer vision solutions, that is using machines for recognizing and processing images and videos. “GIS can significantly aid in locating policyholders and assessing the impact. Accordingly, claims processing also gets expedited. Where previously it might take days or weeks to assess the insurance market’s exposure to a climate event, today, geospatial data visualisation tools using real-time flood warning data can quickly calculate the levels of exposure and export customer lists to help with mitigating damage from the upcoming event.” (8). The data can detect trends and changes over periods of time and combine this information with weather and climate data to more accurately calculate risk exposure or even the effects of climate change on particular regions, cities and individual structures.

AI could save $390 billion in costs across insurers’ front, middle and back offices by 2030.

The future of claims

Geospatial data can help underwriters better evaluate and model risk, but what about after a natural disaster? At that time, geospatial data can be used to estimate damages, predict payouts and reduce fraudulent claims. The most critical time for customers to evaluate their insurer is during the claims process; the customer experience during this process is a major trigger in determining whether they stay with an insurer or look for another. This doesn’t just apply to property and liability insurers, but to the industry as a whole.

The most critical time for customers to evaluate their insurer is during the claims process; the customer experience during this process is a major trigger in determining whether they stay with an insurer or look for another.

The claims process is complex as different claim types follow different processes and involve different departments. From the time a customer submits a claim until the time it is resolved, insurers have the opportunity to prove themselves. This starts with the options the customer is given to submit a claim or to have internet of things (IoT) devices and sensor solutions automatically trigger the filing of the claim. Automating submission tools not only makes it easier for the customer, it also standardizes the data received by the insurer, enabling simpler data aggregation. Handwritten forms can be submitted and processed with computer vision solutions just like pictures, and even customer–agent conversations can be recorded and the data aggregated into a single platform for analysis.

Data generated in this first step alone will create a library of events that can be used to check for fraud, to optimally route claims to the proper department or even to complete the claims process automatically based on the claim’s severity and complexity. The clock is ticking and customers expect the situation to be resolved quickly. They have submitted everything the insurer has requested and, as days, weeks and sometimes even months pass, they must be kept engaged. Insurers know the complex process claims go through, understand each step and what is being done; the customers do not. Keeping them informed about this process and continuing to engage them are key to ensure their satisfaction and avoid frustrated and costly calls with live agents.

Proactively engaging customers will be a major differentiator, but so will speed. Auto-mated claims management that includes computer vision and machine learning to analyse claims has already been mentioned. Based on anomalies, the system can flag the claim as potentially fraudulent for it to undergo further review. Here, again, historically, investigators would often need to physically inspect the damages, something that can now be automated based on weather, geospatial and claims data, or on many other types of information. Considering that fraud makes up 5-10% of insurers’ claim costs, intelligent automation tools can help reduce fraudulent payouts and help investigators focus on only the most likely fraudulent claims (9).

Fraud makes up 5-10% of insurers’ claim costs.

The same data used to analyse the type and validity of the claim is then used to estimate the payout. This again speeds up the time to resolution. Every automation and process improvement is executed to accelerate claim resolution and reduce payout time, ultimately focusing on improving the customer experience.

Insurers should start with an “everything is possible” mindset to unleash truly transformative ideas. Satisfaction surveys pertaining to claims consistently show that customers want a fast and intuitive process as well as transparency regarding the process and what happens next. Consequently, the digital redesign of a claims journey needs to go much deeper than superficial process improvements. The Spanish company Adeslas has completed the end-to-end digitization of its claims journey, applying such features as multichannel First Notification of Loss, automated claims segmentation and digital claims status tracking (10).

Data

Data is key to all of these topics, and every company across every industry needs to become a data company. Companies that embrace this change both from a process perspective and, even more importantly, from a cultural perspective will create a competitive edge. Insurance companies that successfully become data-driven will innovate, simplify and improve their customers’ experience while streamlining internal processes. To do this, they will need to rethink their approach to data; archaic designs from the past need to be replaced with agile methodologies that put data at the heart of decision-making.

This data analytics journey can be broken down into four parts, with questions to consider:

1. Data strategy: What is your company looking to do? This should be a visionary outlook but one grounded in where you want your company to be and, more importantly, what you want it to be doing in five years from now.
2. Use cases: What applications and processes will you focus on to deliver on your strategy? If better customer engagement, you could focus on an omnichannel customer platform. If more precise insurance payout estimates, the focus could be on analysing claims data with computer vision. Make sure to look broadly at your business and focus on what makes sense for you, not what the current buzzword is.
3. Data platforms: How is your data currently structured? Is it siloed or generally available for users across different business functions? Depending on the structure, working together with IT and business units to evaluate your data infrastructure, architecture and platforms will also be a deciding factor in regard to what is possible and where investments need to be made.
4. Data operations: How will these new applications and processes be maintained and managed? More dynamic solutions evolve over time, including adding new data sources or enabling new features. The data platforms mentioned above will help manage this evolution, but areas like data operations (DataOps) and AI operations (AIOps) are becoming more important to ensure the continuous success of your data-driven applications and processes.

Companies need to understand that this journey will take time. Key is to have an agile approach that allows quick wins but also quick failures — both of which will likely occur. It is important to keep the customer experience and customer perspectives in mind when developing new strategies, and to reinforce a proper data culture throughout the organization. Otherwise, these initiatives have limited chances of success.

At Unit8, our goal is to drive the adoption of AI in the insurance space and help companies close this knowledge gap. Key areas in which we help leading insurers solve major challenges using data, advanced analytics and AI include:

• FRAUD PROTECTION & COMPLIANCE — Fraud can be devastating to a business, costing insurers billions of dollars every year, with fraud schemes becoming increasingly complex and sophisticated. Staying one step ahead of fraudsters’ creativity as well as enabling a fast and reliable investigation process are crucial. Internal compliance and regulatory processes can use anomaly detection to spot deviations from baselines and norms. Regulatory bodies like the International Association of Insurance Supervisors (IAIS) and the Common Framework for the Supervision of Internationally Active Insurance Groups (ComFrame) set forth guidelines on monitoring and controlling such anomaly detection tools. This combined with tools to detect and prevent claims fraud builds a scalable risk framework across the business.
• CHURN PREVENTION — Changes in customer behaviour are causing a fundamental shift in the insurance distribution model. Consumers are embracing digital channels, and their experiences with leading tech companies have raised their expectations when buying insurance both online and offline. A seamless and consistent omnichannel experience is now the standard that all companies must strive to meet. This mix of touchpoints makes customer engagement much more complex. A monitoring system that identifies unusual customer behaviour allows proactive engagements by your teams to significantly reduce churn. And integrating a learning loop into this process can constantly improve it based on clients’ profiles.
• CLIMATE RISK SCORING — With the increase in natural disasters and global catastrophes, static models can no longer be used for proper pricing and underwriting. Unit8 uses AI and analytics at scale (billions of datapoints globally, including geospatial data) and has the expertise and tools needed to build models that allow new and innovative insurance offerings while properly calculating risk exposure. We have helped property and casualty insurers calculate the risk impact climate change and weather patterns have on their portfolio, allowing them to better price their underwriting models and also proactively engage their customers to take preventive measures and ensure the longevity of their infrastructure.
• FORECASTING — Forecasting is used in all areas of the business, from finance departments to claims departments. AI and machine learning allow for improved claims forecasting techniques that better predict payout values as well as the subrogation potential. AI enables the automation of payouts for the simple claims and guides experts on where to focus their energy to speed up the resolution of the more complex claims, ultimately leading to improved customer satisfaction. Similar approaches are used to forecast cash flows, revenues, investments and demand across all areas of the business.
• DATA — The need for accessible data and the shift away from siloed data are challenges for all companies. Companies that establish a robust and secure data strategy can scale the development of the above-mentioned new applications and processes. Unit8 partners with companies to build these platforms, while also helping them audit existing architectures and providing guidance on new innovations. We accelerate the adoption by working with both IT stakeholders and business users to evaluate, build and manage solutions that make sense for the company environment. These areas are only the starting point. Data and AI continue to fundamentally change the sector. Signs can already be seen of increasing automation and the application of analytics in claims management, underwriting and pricing. Trends such as UBI will amplify the importance of advanced analytics even further.

Footnotes

1. https://hbr.org/2019/07 building-the-ai-powered-organization
2. https://store.businessinsider.com/products the-ai-in-insurance-report
3. https://www.ibm.com/downloads/cas A6MPGEJW
4. https://www.accenture.com/fr-fr/_acnmeia/36dc7f76eab444cab6a7f44017cc3997.pdf
5. https://www2.deloitte.com/us/en/insights industry/financialservices/insuranceproduct- development-capabilitiesmodernization. html
6. https://hbr.org/2014/10/the-value-ofkeeping- the-right-customers#:~: text=Depending%20on%20which%20study% 20you,the%20one%20you%20have%20happy
7. https://www2.deloitte.com/content/dam/ Deloitte/ca/Documents/deloitte-analytics/ caen-age-of-with-insurance-aoda.pdf
8. https://www.geospatialworld.net/blogs/ importance-of-gis-in-insurance
9. https://www.iii.org/article/ background-on-insurance-fraud
10. https://www.mckinsey.com/industries/ financial-services/our-insights/ claims-in-the-digital-age

The Unit8 team is comprised of world-class experts in machine learning engineering and data science. Our team members have worked for such Silicon Valley companies as Google, Amazon, Palantir and Apple, as well as for leading financial services institutions; we bring their best practices to every project. We partner with some of the biggest companies in the world to solve challenges that directly affect their business. At Unit8, our mission is to drive the adoption of AI and data science in industries and empower organizations.