Enabling Artificial Intelligence in Life Insurance
The insurance sector can leapfrog into the present using artificial intelligence (AI). It is a key competitive differentiator in the future of intelligent insurance. However, AI is no panacea. It has a set of red flags that must be properly understood before implementation.
Artificial intelligence (AI) is a broad term that includes anything and everything from Boston Dynamic’s «Big Dog» to self-driving cars. At its core, an AI system is a piece of software that embeds (human) intelligence in machines. It allows systems to find solutions to problems on their own – be it discerning patterns (i.e. finding complex relations) or adapting to constraints. An AI-based system can generate better risk insights and thereby enhance decision-making capabilities. The exponential growth of various AI-related technologies like natural language processing, computer vision and deep learning has played a pivotal role in bolstering the case for AI.
Impact of AI on insurance industry
The insurance sector is beginning to see the value of artificial intelligence. The self-learning capability of AI systems allows insurers to make new product offerings across different geographies and customer segments. It offers insurers the ability to scale rapidly by increasing their prowess in areas like informal retrieval, robotic process automation, virtual assistants and data management. AI is used for efficiency improvement and process automation primarily in customer facing, underwriting, and claim management processes. However, due to increasing customer demands and cost pressures, AI will soon be leveraged in more complex tasks. The following areas will see the biggest impact of AI:
Algorithmic underwriting: This capability is generating significant attention. AI systems can be used to discern new patterns and relations that are beyond human capability. With the advent of new data sources and a general increase in data volume, it seems logical to employ machines in uncovering new relations and insights. Specific areas that AI will impact are:
- Facilitating the extraction of the relevant underwriting data.
- Building predictive models on risk assessment.
- Scaling to other data sources like home & industrial internet of things; using them for modelling purposes.
For the end customer, an AI-based underwriting approach adds more value in terms of the level of information provided. For instance, customers can benefit from getting instant decisions on coverage and better pricing. Haven Life, a unit of MassMutual, uses AI to streamline the new business underwriting process. With this AI system, it can complete the online application and approval process in about 25 minutes. This is a significant improvement for a process that can take days or even weeks.
Several conditions apply however. When it comes to underwriting, an AI system depends on the input data and underwriting rules. Haven Life, for example, relies on MassMutual’s data collected over years to build its models. For complex data or cases where the AI system has not been trained on sufficient data, it may not be able to give a qualifying answer. In such cases, the AI software should be programmed to hand over the control to the human.
Intake management: A trivial but relevant example of intake management is optical character recognition (OCR). Although helpful, this can pose a major hurdle to insurers that have yet to embrace digitization. AI automates this process by reading information from physical forms and then feeding it into the system. The technology has advanced to an extent where it can read letters and numbers for which humans may have a hard time deciphering. «Captricity», a startup based out of California, uses a combination of machine-learning and human verification for extracting information from handwritten forms. The firm says its technology can achieve an accuracy greater than 99%.
Personalized product offerings: The introduction of AI will assist in making more personalized product offerings. By working with granular customer data, it will be possible to ask customized and specific questions about a customer’s habits. Doing so will not only help improve existing products, but could open channels for further revenue.
New product offerings: Increasing AI adoption will result in the creation of new product offerings. An example of this is a company called «FitSense». It offers a health analytics platform collecting user’s data from different devices. The data is then used to build user profiles and generate actionable insights like health scores to quantify risks. By working with granular customer data, AI makes it possible to ask customized and specific questions about a customer’s habits.
Enhancing customer experience: Soon we may witness mass deployment of chat bots and virtual assistants. These would primarily help customers understand their insurance needs, answer queries and help customers choose an appropriate plan before a policy is purchased. An example of personalization could be modifying the number and nature of questions asked based on customer profiles and input when choosing a product.
Empower human personnel: Because of its innate capabilities like quick decision-making, AI can empower personnel by improving their efficiency. AI can also be used to increase target conversations, quote-to-bind process and accelerate the launch of new products.
Increasing operational efficiency: AI offers multiple ways to improve operational efficiency, including:
- Lower costs: AI can help organizations to resize and reduce their overhead costs. AI can be especially useful in performing routine tasks.
- Improved productivity: With AI taking care of routine and repetitive tasks, employees can focus on skilled tasks and improve their firm’s capabilities.
- Intangible benefits: AI-based systems offer intangible benefits like faster access to information, coherent responses to queries, better data management and information searches.
Challenges for insurers
With the adoption of AI, insurers will have to implement new processes. Efforts will have to be made to integrate different technology components and stakeholders across the value chain. Additionally, insurers will now have to build and operate a data repository that can be exposed directly to the AI systems. This data repository will be primarily used for training and testing. Here are some challenges that insurers face in deploying an AI-based system:
Building a strong base: For an AI system to answer business-oriented problems, it needs to be fed huge amounts of domain-specific information that covers a multitude of business scenarios. Building an efficient AI system is an ongoing process; insurers will have the added duty of ensuring that pertinent data is timely available for the AI systems to train on.
Exception handling: AI systems come with unique technological bugs which can require case-by-case exception handling. For example, natural language processing (NLP) which is used quite predominantly in AI-based systems still needs to work with diverse moods, accents, speed and tone. Furthermore, NLP has to infer the contextual meaning and understand the difference between homophones (like «by» and «buy»). The advancement of AI will bring new challenges, so methods must be developed in response. At the start, AI systems will require a lot of manual supervision. Employees will need to be trained to monitor and rectify AI transactions. Investments will be made on the recruiting and training of personnel.
Regulation and compliance: Since AI acts as a data repository and monitors transactions, insurers will need to ensure that adequate information security policies are in place. Timely auditing of security practices will be a norm. As most of the AI solutions will be cloud-based with data privacy policies, procedures will be of utmost concern.
Below are steps that should be kept in mind while initiating an AI program:
- Identify key business areas: Start with identifying critical areas affecting the business and the performance metrics to measure improvements. This could include the customer target-to-conversion ratio as well as any reduction in claim processing time.
- Identify AI applicable areas: Once the key business areas have been identified, map them to applicable AI areas such as predictive analytics, NLP and machine learning.
- Lay the groundwork: This includes building the data set, identifying the key skills for the chosen AI techniques, setting up metrics to monitor the AI performance and continuously improving AI capabilities.
- Initiate pilot programs: Start with a pilot using the created data set and tools that are either vendor specific or open source. Assess the result of the pilot using business performance indicators and AI indicators identified in the steps above. Use the results as a feedback to improve the AI techniques applied.
- Ensure scalability: Once a decision has been made on the AI system, it must be able to grow with the business using the appropriate technology stack and operational support.
- Encourage adoption: A parallel effort will be required to ease the transition to an AI-based system. A change management program will need to be implemented to facilitate employee adoption and ensure the use of AI into job functions.
Taking into account all of the opportunities created by AI for life insurers, we envision the following trends:
- Pilot programs: Given the inherent risks in the insurance sector, we see AI programs starting as pilots and proof of concepts. These will then be honed to an acceptable level of accuracy and scalability before being commercially deployed on a large scale.
- Increased partnerships: Going forward, we see more alliances between insurance and technology companies. The tech companies will assist insurers in understanding their business needs and then implement an appropriate AI-based system.
- Talent shift: Since it requires a more technical skill set for piloting AI-based systems, we see a change in recruiting and staffing requirements. There will be a preference for statistics and computer science graduates.
It is an exciting time to be in the life insurance sector. This sector is at the brink of a technological transformation, with a major component being AI. AI allows insurers to meet customer expectations, reduce costs and remain competitive. However, if systems are not implemented correctly, the advantages can be overshadowed. Opportunity for sustained growth lies ahead for those who embrace this transformation and implement an appropriate AI strategy.