In future to succeed in utilizing and bringing value out of data analytics in insurance, data scientists and actuaries must collaborate to understand the analytics models. At present only 3% of the insurance companies in the US are doing the same.

• Definition / Scope
• Market Overview
• Top Market Opportunities
• Market Trends
• Industry Challenges
• Technology Trends
• Pricing Trends
• Regulatory Trends
• Other Key Market Trends
• Market Size and Forecast
• Market Outlook
• Distribution Chain Analysis
• Competitive Factors
• Key Market Players
• Strategic Conclusion
• References

Definition / Scope

Big Data has become a catchall phrase used colloquially to represent the capture and analysis of large databases to glean useful results. A singular, comprehensive definition of Big Data is elusive.

A common definition of Big Data is the analysis of data demonstrating such volume, velocity, and variation, that standard data-processing equipment is not sufficient for the task. Given the speed of technical advancement in data-processing hardware and software, the concept of a large database is a moving target; however, much faster increases in production and collection of data suggest this definition will be valid for years to come.

Therefore, it is common for Big Data analysis to occur on a server (perhaps in a cloud), rather than on a desktop or laptop computer.

• With data analytics in Insurance, the scope of this report covers use of data analytics across various areas in P&C (Personal and Commercial) Insurance as well as L&H (Life & Health) insurance businesses respectively.
• The term ‘Big Data’ is being used as a common phrase at present times where the technology captures and analyzes large databases to gain useful results. The analysis of data involves such volume, velocity and variation for which a standard data-processing equipment is not enough. Thus, big data analytics is generally conducted on a server (cloud) rather than on hardware (computer). 
• Most common data analytics software packages use Microsoft Excel or spreadsheet format. The hard drive storage must be huge as the amount of data will be high. Second factor to consider is velocity or speed at which data is created and analyzed. For instance data used to monitor audio and video recordings can be used to alert authorities in case of some danger approaching. The third factor is variation which considers the formatted and organized data where massive data in form of texts, numbers, words, images and sounds are used to draw inferences. 
• Uses of big data in Insurance: 
  • Underwriting and pricing: The role of big data in insurance operations is to match the premiums and losses through pricing and underwriting more accurately. For instance, the use of big data in pricing is use of telematics data. The plans based on telematics are often known as PHYD or pay-how-you-drive plans. The benefits of PHYD is high where it is fully transparent to customers. Through the feedback of the insights from data analytics the behavior of the customers can be monitored and improved.
  • Satisfaction of customers: Another use of big data in insurance is to improve customer satisfaction. Two major approaches in customer satisfaction is first, to resolve service related problems and second, to provide personalized offers as per customers’ preferences. For instance, mitigating customer service issues and addressing customer queries in real time. In addition, advance data analytics is also able to detect customer satisfaction from data produced during telephone conversation where software detects if there is some problem and automatically transfer the call to an authority to solve the problem
  • Increasing coverage: One example of big data use in insurance is computerized catastrophe models for pricing and underwriting extreme weather conditions such as windstorm or earthquake. Data driven catastrophe models estimate loss outcomes and probabilities more accurately. In absence of such models, private insurance for such exposures will be expensive and cannot be made available too.
  • Mitigating fraud: In the US the annual estimates for costs of insurance fraud totals around $40 billion. In 2018, the number of fraud detection used by insurers in the US was also around 70% made possible by big data methods. The big data analytics has capabilities for fraud detection and mitigation and two major approaches used by the businesses are text analytics and social media analytics.
  • Operational efficiency: Technologies are being used to make businesses run more efficiently. Big data is used to improve insurer efficiency in ways that benefit consumers. With the help of data analytics insurance agents are able to complete the majority of personal line insurance application fields by requesting only one or two items from prospective customer. By restructuring the application process, the search costs of the insurers are being lower than before.

Market Overview

• In the US both P&C and L&H insurers seem to be leveraging advance analytics which has led to numerous benefits including reduction in expenses and rise in sales. In the US, 78% of the insurers are using advance analytics to improve pricing and product development, 67% of the insurers are working on customer relationship management, 64% for profitability and expense management and finally 60% for technological innovation respectively.
• With data analytics, 50% of the insurers encountered reduction in underwriting, 43% encountered rise in sales and 47% encountered increase in profitability respectively. 
• Further, since 2017, the data collection across all areas have increased significantly. There is astounding rise in data collected from social media (164%) and web scrapping (169%). Besides other sources for data collection has also increased with 49% rise in internal customer data, 33% rise in customer surveys and 150% rise in clickstream data respectively. However, since past 5 years the most of the increase in data leveraged by the insurers in the US can be accounted by wearables (534%) rise. 
• The US insurers also expect to make a difference in their businesses with the use of data analytics. The highest impact is expected to be made in identifying fraudulent claims followed by understanding risk drivers and automating large classes of standardized underwriting among others. 
• In future to succeed in utilizing and bringing value out of data analytics in insurance, data scientists and actuaries must collaborate to understand the analytics models. At present only 3% of the insurance companies in the US are doing the same. 

Top Market Opportunities

Improve customer insight: Data analytics is insurance is creating new possibilities for the insurers and changing the idea of how insurer deals with intermediaries and policyholders. The analytics enables personalization and customer centric service by giving insight into 360 degree view of the customer.

As customer retention is one of the major contributor of profits to the insurance businesses, insurers in the US seem to be spending 7 times more to attract new customers and retain the existing ones. For instance, predictive analytics can help identify whether a customer is in verge of exiting the company and can suggest corrective actions to mitigate such risks. Corrective actions if taken appropriately can improve customer satisfaction and retain valuable policyholders.

In addition, the analytics can also determine if the customer is switching to the competitor and offer next best action to retain the customer. Thus, these targeted offers through analytics can be delivered through different communication channels which altogether improve customer engagement and increase profitability for businesses.

Innovative business models: Data analytics also support insurers to balance their service through digital and physical channels by gathering insights from both channels and improve their services respectively. For instance, in digital channel gamification is becoming a very important tool where policyholders can easily play game like scenarios to determine right policy fit and also real-time feedback to engage customers in the channel.

The digital channels sometimes create leads and potential customers can be matched with right channel to crack the deal. Besides gamification, insurers can also use social networks to strengthen value-added information through direct, digital or intermediary channels.

For instance, during claims intake, companies can leverage information in unstructured areas of applications and supplement it with social media posts. Another important data is geospatial data which can help insurers identify whether policyholders are being legitimate about their cause.

For instance, for the ongoing analysis of fraudulent claims and their impact on the business, companies can also use solutions to analyze reports and create visualizations of data patterns. Thus, by leveraging these new business models, insurers have the opportunity to add value to their businesses. 

Manage risk and finance: Through the help of data analytics risk models, carriers handle the volume and variety of data available at present while enabling response planning. The insurers can identify potential losses at policy level and calculate the immediate effect of a new policy on portfolio while it is being quoted.

With the means of powerful data warehouse, the insurance companies can analyze trillions of data rows and rapidly deliver results that underwriters can use to price policies by street address, proximity to fire stations or other granular factors such as city, county or ZIP code among others.

Another implication of predicative analytics is that, the solutions can help companies anticipate losses before disasters strike or as events unfold allowing risk mitigation and evaluate response plans. Other insurers are also integrating risk management and financial performance management which could definitely become useful to the insurers. 

Market Trends

One of the biggest obstacles for the insurance companies to become more data driven is getting access to large quantity to reliable data which can be used to get insights. The most usable form of data is actually structured in a table with information regarding policy, a vehicle, submission and claim.

However, most of the information that the insurers get access to are unstructured i.e. data in form of images, text documents and sensor data. Thus, structured information which is captured in the IT system only represents a portion of information weighted by decision makers.

This information gap further hinders the insurers’ capability to utilize machine learning algorithms and AI to enable decision-making.

Industry Challenges

Lack of skilled workforce: In order to be successful, insurance companies need to retain the talented pool of employees and recruit the same. At present, there is high demand of data analytics & software development professionals and lack of analysis and interpretation skills among them are creating source challenges for the insurance companies.

In the US, in 2017, Data Scientist was ranked as one of the top jobs with average salary over $116,000. In addition to that, there is shortfall of 1 million STEM (science, technology, engineering, and mathematics) professionals that is likely to continue for 10 years.

The scenario is so clear that over the last 5 years there have been substantial increase in H-1B visa filings for data scientists and US employers are also searching for analytical skills in foreign labor markets.

Further, IBM has pledged to make investments in workforce hiring and spending 1 billion dollars in training over the next four years. However, the shift after the training is also likely to create significant challenge especially in managing intellectual property. 

Rising customer expectations: As insurers look forward to leverage data analytics they also need to understand the risk across customer segments and both positive and negative implications to consumers. The trend has also shifted to customized product and services for the consumers where consumers are willing to share private information under condition that they receive some value in form of service.

Some of the technologies such as connected devices, internet, mobile and social media have become the mainstream at present and the delivery and consumption of these services will be different than that of traditional channels.

Around 87% of the millennials in the US are using smartphones to transact for business which suggests that it is a huge challenge for the insurance industry to leverage data analytics and other digital technologies to engage and enhance customer experience of the new generation. 

Technology Trends

Across the insurance industry, the data analytics is being driven majorly by two important technological developments that include: 

• Rapidly evolving IoT and the network of connected device
• Advancing capabilities that enable firms to analyze high-velocity data

Some of the other major data analytics technology applications in the insurance industry are as follows: 

• Telematics and connected devices: These are the devices that share information online through cloud and have implications for pretty much every corner of the insurance industry.
  • Vehicle Telematics: These are embedded with sensors that send real-time driving data to insurers which in turn can help in timely interventions and lead to more accurate pricing and risk profiling. By 2020, it is estimated that 25% of the auto insurance premium revenues i.e. unlocking $30 million value and this will be driven by telematics in the US insurance. Drivers that are speeding or taking other kinds of risks behind the wheels will get warnings in form of alerts in real time. Especially for the commercial vehicles and rental cars, the information can be sent directly to fleet operators or vehicle managers. In case, the sensor detects an accident, it can initiate claim or emergency contacts too.
• Wearable devices: The technology has most application in health insurance, where insurers operating in this segment have begun to experiment with the wearable devices by providing financial incentives for the policyholders who are opting a healthy lifestyle. Till present, the scope of wearables have been limited to use of pedometers which track daily walking distances. Besides, in workplace, the wearable device can also give immediate alert for unsafe action, track fatigue, body temperature among others. Additionally, it also prevents a worker from engaging in any risky behavior. Thus, wearable helps manage risks in near future by taking corrective action, this in turn prevents a claim or more accurately predict the possibility of claim overtime.
• Smartphones: These are the connected deceives we carry in our daily lives and is providing wealth of information to insurers. Ranging from social media updates to GPS tracking, insurers can get insight into customer’s current activity and send updates from whether and safety warnings to targeted marketing materials.

Pricing Trends

• Since more than 10 years, Americans buy insurance product as commodities. The commercials of the insurance companies have always been portrayed in terms of the price; price being the biggest differentiator and factor providing competitive edge to the insurance companies.
• In addition to the price centric behavior of the companies in the market, the customers who purchase auto or medical insurance in particular, have the choice to switch insurance carriers with blink of an eye. As competition is increasing, big data and associated technologies are providing companies opportunities to reshape into modern insurance landscape. With the help of data analytics, insurance companies are able to get access to broader form of data such as historical, continuous and real-time and create a more accurate risk profile. Thus, this enables insurance companies to offer more competitive prices that ensure profit by covering perceived risk and working within customer’s budgets. 
• Accurate or competitive pricing: Especially in auto insurance companies are identifying risks accurately and building a proper risk profile to ensure profitability. As bigger data is available in insurance industry at present, this is allowing insurers to price policies with a better understanding of vehicle safety. For instance, all-encompassing information about weight, horsepower, bumper height, crash ratings and safety features create more improved predictive analytics models which helps determine fair price. The underwriting aspect of insurance is also being transformed with data analytics. In the US, the technologies such as Hadoop has allowed insurers such as Allstate to get access to deeper customer information and note patterns which is helping the company generate competitive premiums to attract new customers. Other new players in the market such as Metromile has also established a new business model in auto insurance called auto insurance by mile i.e. price is determined on the basis of miles covered. The entry of Metromile has not only disrupted the industry but also created pressure for other traditional players to advance on their own models. 
• Usage based Insurance: This business model links premium rates with driving behaviors. Some of the companies such as StatFarm, The Hartford and Progressive are already utilizing the UBI method for underwriting. Technologies such as IoT and sensors are providing new ways to collect and analyze more data. For example, Progressive with its UBI program called Snapshot was able to collect more than 10 billion miles of driving data which became an ideal go to for low-mileage drivers. In the US, by 2020, 36% of the insurers are expected to use UBI and this is headed to transform the pricing models in insurance business. 

Regulatory Trends

• In the US, some of the insurance data related governing acts include: US Fair Credit Reporting Act (FCRA), Electronic Communications Privacy Act (ECPA), and Health Insurance Portability and Accountability Act (HIPAA) among others. These are further supported by other secondary level regulations to protect the interest of consumers. The main act that governs the overall insurance scenario in the country is National Association of Insurance Commissioners (NAIC) which is presently working on a Cybersecurity Model Law. 
• At present, the regulatory environment for data in insurance is very complex and further continues to increase costs for insurance companies in case they comply with these regulations. For instance NAIC Model Law requires companies to follow standards for data security, investigation and notification of a breach of data security and as this is applicable to all insurers, the costs of establishing such standards is expected to cost over $52 million per year for an insurer with 50,000 employees. 
• Future developments in data analytics may lead insurance companies to comply with a broader set of regulations. As data and other technologies crosses industries cross industry regulation of data is applicable to all companies. For instance, one trend that is gaining traction in the market is autonomous vehicles bringing disruption to the private auto market by raising questions on liability when autonomous technology fails. The Department of Transportation (DOT), National Transportation Safety Board (NTSB), National Highway Traffic Safety Administration (NHTSA), the courts, or other governing bodies have determined that auto liability lies with the manufacturer. 

Other Key Market Trends

Prevent reduce Fraud and waste: In past, generally there was significant evidence of fraudulent claims, however at present, data analytics is being used to recognize who is most likely to commit insurance fraud before such incidence takes place.

For instance, an agent can simply monitor data of the insured in real-time from outside sources such as social media platforms to see whether they might be engaging in fraud. Discrepancies in claim filing is being known through such techniques. 

Support in pricing premiums: To gain competitive edge in the market, insurance companies are coming up with new and unique ways to derive actionable insights from data analytics to track individual policyholder behavior and price policies accordingly.

For instance, predictive modelling is being utilized to predict the probability of policyholder being involved in an accident or having their car stolen. Predictive analytics also gives insights on individual policyholder by providing information about their driving habits and behaviors and then comparing them against other policyholders in their database.

For example, in auto insurance a small box is kept inside the vehicles which then records all the driving information of the driver and later can be connected to a mobile application to derive insights out of the data recorded. If policyholders who drive responsibly will be incentivized by having to pay lower premium and bad drivers are disincentivized which motivates everybody to drive responsibly and thus reduce risk of accidents. 

Self-servicing policies: This is next major innovation in the insurance industry. Companies along with brokers that offer customer portal for the policyholders to manage policies will not only load off some work but also make customers happier.

Some of the life insurers such as MassMutual, and P&C companies like Geico and Allstate have also adopted customer portals and brokers and policyholders are demanding for it. With the help of data analytics the agents can automatically make smart recommendations to customers’ right at the moment they are buying a new policy or making changes to existing one. 

Market Size and Forecast

• The insurance industry market size in the US, in 2018 was estimated to be around $1.33 trillion of both L&H (Life & Health) and P&C (Personal & Commercial) lines of insurance combined.
• The data analytics industry market size in the US in 2018 was estimated to be around $100 billion. 
• In terms of insurance analytics, North America remains the largest market region in the world and US being the leader within North America region. 9 out of 10 of the major players in the insurance analytics market belong to the US market.
• In the US, in 2018 a total of $2.5 billion was spent on insurance analytics whereas the global investment remained at $4 billion. So far, this suggests that US is aggressive in leveraging analytics in insurance.

Market Outlook

• The global insurance analytics market was valued around $6.53 billion in 2018 and is expected to reach $16 billion by 2023 registering a CAGR of 12.5% during the forecast (2019-2023) period. 
• Within the US, the insurers are tapping onto use of data analytics across various areas with most in underwriting and claims. By 2020, insurers are likely to increase smart building data collection to 52%, telematics data (70%) and claim and underwriting information data by 61% and 52% respectively. With focus on these factors, insurers will be able to provide more personalized, faster and curated experience for the customers.
  • By 2020, the insurers will also be able to manage risks well. Through data analytics technology, insurers will be able to reduce time spent by employees on tasks (49%), identify-high risk cases (45%) and build better risk models for decision making (45%)
  • By 2020, insurers will also be able to make improvements in claims management. With 82% reduction in fraud and identification of claims up to 80% mark will be enabled through data analytics use.
  • At present, the telematics application is limited to auto insurance only, however by 2020, the telematics is expected to have implications in homeowner insurance with 43% of companies expected to utilize the technology in new line of insurance product.
  • Finally, a larger number of insurers in the market (48%) are looking forward to cloud solutions and a significant number are investing in Hadoop (37%) for distributed storage and processing of large data sets.

Distribution Chain Analysis

Advanced analytics is applied throughout the value chain of insurance businesses: 

• Pre-sale: Most of the data analytics is used in identifying new rating factors based on the wealth of customer information and also in risk segmentation during underwriting process. Somewhat analytics is also used to determine competitive price for policies across product lines and understand the customers’ propensity to buy policies.
• Underwriting: Data analytics is utilized to match the insurance policy with particular customer according to their needs. In addition, it is also used for fraud and non-disclosure protection. Whereas, the predictive analytics particularly helps in understanding the policyholders’ propensity to complete the purchase and take necessary actions to keep them engaged in case they are intending to leave.
• In-force management: Multivariate experience analysis technique is most used at this stage. In line with that, analytics is used to cross-sell and up-sell across multiple channels. Each channel having its own strategy based on the behavior and need of customer on that particular channel. An area that needs to be worked out in this stage is to retain customers for long term.
• Claims: Finally, claim settlement is the final stage in the insurance value chain where data analytics application has been limited to fraud prevention and identifying fraud claims. One area in which data analytics needs to be utilized is claims triage or understanding customer needs in claims processing which could lead to efficient settlement.

Competitive Factors

• In the US, insurance industry is one of the legacy businesses and is competitive at the same time representing an estimated $507 billion or 2.7% of the US GDP. As customers are demanding more personalized experiences from the insurers, number of leading insurance businesses are exploring machine learning and data analytics to improve operations and to serve customers better. 
• State Farm, one of the leading insurance companies in the US are exploring computer vision to detect distracted drivers, the online program for the same was launched in 2016 called Drive Safe and save. The application consists of 2D dashboard camera images and measures the behavior of the driver based on several factors such as texting, talking on the phone, operating radio among others. Further, the image classification is done on two main photo regions: the head region and the bottom-right quarter where the drivers hand normally appears. The program is headed to collect and interpret data on drivers which is likely to play a significant role in company’s ability to customize insurance options and provide discounts on policies. 
• Liberty Mutual announced plans to develop automotive apps with AI technology and products aimed at improving driver safety in 2017. Its own dedicated Lab called Solaria will be working on the innovation along with open API developer portal which integrates company’s proprietary knowledge and public data to inform how these technologies will be developed. The company is also testing an application to help drivers involved in car accident to assess damage to their car in real-time using a smartphone camera. The app is further expected to use data driven model whereby it is able to scan images of thousands of cars and also provide damage specific repair cost estimates. 
  • In May 2016, Liberty also announced plans to launch $150 million VC initiative called Liberty Mutual Strategic Ventures (LMSV). The early stage venture fund also focuses on innovative technology and services specially designed for insurance industry. The VC firm has already invested in two major companies called Snapsheet and Przbyl that are in fact involved in AI and machine learning to support the data analysis capabilities of the company.
• Allstate, a leading insurer in the US recently partnered with EIS (Early Information Science) Agency that helps businesses to improve their performances through data analysis. EIS developed a virtual assistant called ABIe (The Allstate Business Insurance Expert) to support Allstate agents in seeking information on commercial insurance product lines of the company. Before ABIe, the agents of Allstate only sold personal line products. In addition, ABIe has also helped those agents to access information they needed to successfully communicate with potential clients. As a result ABIe processes about 25,000 inquiries per month. 
• Progressive Insurance: At present, the company is leveraging machine learning algorithms for predictive analytics based on data collected from client drivers. The telematics app by Progressive called Snapshot has successfully collected over 14 billion miles of driving data and the company incentivizes with an average discount of $130 after six months of use. The company is also looking to collect more data and reduce their time to insight. As a result, with data analytics, the company marked an increase in commercial lines from 0 to 9% from 2014 to 2015 and also in personal line 2 to 6% respectively. The revenue also increased to $23.4 billion. The AI platform utilized by Progressive Insurance was developed by a company called H20.ai. 

Key Market Players

Top 10 companies providing AI (data analytics) technology solutions in Insurance in US market landscape are as follows: 

• Insurify: Boston based company has developed a virtual agent called Evia that constantly verifies customer data to compare real-time quotes from carriers in various zip codes and provides coverage recommendations accordingly.
• CCC: Chicago based CCC has a AI based collision estimate called Smart Estimate that allows the insured to make claims accurately and timely.
• Lemonade: The New York based company provides paperless policy for homeowners insurance. The company uses AI to create policies and handle claims via mobile and desktop channels.
• ZestFiannce: The AI software of the company helps insurance underwriters identify risks based on traditional and non-traditional data around customers. The data analytics system helps insurers boost their underwriting profits and create better risk profiles of customers.
• Clearcover: The Company uses AI technology to insure users and quickly process claims. A simple questionnaire is to be filled after which the company uses AI technology to choose a plan/policy that better fits the need of the customers.
• Flyreel: Denver based company offers data analytics solution for underwriting property insurance. Its AI platform gives a 360 degree view of the property to the underwriters and helps manage claims and risks accordingly.
• Inshur: This New York based company is the mobile-first way of purchasing insurance for TLC insurance. The professional drivers can search and choose among 100’s of plans. In addition, they can switch policies and file claims through the mobile app
• Cape Analytics: California based company uses deep learning and data science and collaborates it with geospatial imagery. It provides property value and risk information to insurers that provide insurance to home and businesses.
• Galaxy AI: It has developed its own dedicated AI platform called Galacticar that helps in providing automated claims. The efficiency of claims through the platform is high as it prioritizes claims that are high in severity.
• Insurmi: The Arizona based company utilizes AI technology to provide conversational and automated customer service. The platform provides assistance in choosing best plans for an individual based on their profiles and requirements. Its dashboard called Chat Tracker also keeps record of customers’ progress in real-time

Strategic Conclusion

The insurance industry has been utilizing many sophisticated analytical technique at past however, the emergence of data analytics in the insurance business has enabled the industry to cater to customer demands. Further, the consumers are benefited from these insights which have further encouraged them to share private information to seek more customized products and services.

Thus, with the data analytics revolution insurance companies operating in traditional ways are being continually challenged and their competitive edge would most likely depend on their ability to retain and recruit talented people. In near future, the insurance company is likely to grow on the foundation of data analytics and other advance technologies.

References

1. https://www.naic.org/insurance_summit/documents/insurance_summit_2018_CIPR_01.pdf
2. https://www.willistowerswatson.com/en-US/Insights/2019/05/infographic-life-insurers-embrace-predictive-analytics
3. https://www.ibm.com/downloads/cas/GJX0JABM
4. https://www.willistowerswatson.com/en-US/Insights/2019/02/decoding-the-hidden-value-of-unstructured-text-data
5. http://www.crocouncil.org/images/CROC_Risk_Implications_of_Data_and_Analytics_to_the_Insurance_Industry_20170106.pdf
6. https://www.theinstitutes.org/guide/harnessing-power-real-time-data-analytics-insurance
7. https://customerinsightleader.com/others/applications-analytics-insurance-pricing/
8. https://www.dataversity.net/five-ways-data-analytics-transforming-insurance-industry/#
9. https://www.treasury.gov/initiatives/fio/reports-and-notices/Documents/2018_FIO_Annual_Report.pdf
10. https://www.idc.com/getdoc.jsp?containerId=prUS44998419
11. https://news.crunchbase.com/news/a-record-2-5b-went-to-u-s-insurance-startup-deals-last-year-and-big-insurers-are-in-all-the-way/
12. https://www.insurancebusinessmag.com/asia/news/breaking-news/insurance-analytics-market-worth-16-billion-by-2023–report-99970.aspx
13. https://www.willistowerswatson.com/en-US/insights/2018/05/advanced-analytics-and-the-future-insurers-boldly-explore-new-frontiers
14. https://www.soa.org/globalassets/assets/files/e-business/pd/events/2018/predictive-analytics-hong-kong/2018-pa-hk-03.pdf
15. https://emerj.com/ai-sector-overviews/machine-learning-at-insurance-companies/