Covid-19 will Increase AI Adoption in Insurance
Updated: Aug 7, 2020
Insurance has been hit hard by Covid-19 and economic hardship. With many insurers focused on cash conservation, leading insurers can emerge from the crisis even stronger if they make smart investments in AI. Insurers’ massive customer datasets and their famously manual processes create some ‘quick win’ AI opportunities.
Insurers must proactively adopt AI because the business outlook isn’t great. Lloyd’s of London estimates that 2020 underwriting losses will hit $107 billion. (Re)insurance firms are looking at billions in claims for business interruption and trade credit insurance losses (insurance for when buyers can’t pay sellers).
Insurance policy sales will be affected by falling economic activity. Production and movement of goods and commodities stalled. That means fewer companies need insurance for cargo, energy, commodities, shipping, etc.
Many insurers will emerge shaken from the Coronavirus era. They will then walk right into the worst recession since the Great Depression according to the IMF, which will reduce demand for some types of personal and commercial insurance.
Why AI Now?
At first glance, a post-pandemic economy might not seem the best time to invest in artificial intelligence. After all, shouldn’t firms preserve as much capital as possible?
In fact, this is an opportune time to develop AI capabilities. AI is good at automating logical, repetitive processes and generating insights from data. This allows insurers cut costs and discover new revenue streams. The maturity of AI software vendors means that AI tools need not be built from scratch. If insurers understand their business needs and develop an AI strategy to meet those needs, small investments in AI can achieve high ROI.
Large insurers have been investing in digital transformation for years. Converting data into digital format, embracing mobile & web-based customer interaction, and upgrading tech stacks has given these firms the infrastructure to adopt AI quickly. Digital transformation has laid the groundwork for AI transformation.
Insurance Business Priorities
Insurers should focus on three priorities in a post-pandemic economy: cost reduction, risk reduction and customer insights.
Covid-19 and the economic downturn will affect insurers in the revenue (underwriting income) and expense (claims) columns in the short term. The immediate business priorities for insurance companies are cost and risk reduction until the market picks up.
For cost reduction, insurers should prioritize process and claims automation. They will streamline workflows so that the same work can be done with fewer people. AI tools such as intelligent Robotic Process Automation (RPA) will be relevant here.
For risk reduction, insurers will invest in better fraud detection tools since fraud is expected to increase during hard economic times. According to the Association of British Insurers (ABI), the 2008 recession saw a 17% increase in fraudulent insurance claims compared to 2007.
Insurers will also reduce risks by improving underwriting standards. This means understanding risks better and insuring higher quality risks to prevent large unanticipated claims. Machine learning and Natural Language Processing (NLP) tools can search through past insurance policies and understand how to price new policies.
Insurers can then focus on preserving revenue streams and discovering new revenue sources. Investing in AI for customer insights will yield significant ROI because insurers have large amounts of customer data. Enterprise search software powered by machine vision can quickly search through internal databases and document repositories to give agents and customer service staff a 360-degree view of customers.
AI for Cost Reduction: Claims Automation with Intelligent RPA
Insurers can use intelligent RPA to automate claims processing to cut costs and pay claims more quickly. Claims handling is rife with challenges that intelligent RPA can solve, such as manual data input, multiple data sources (documents, emails, images, mobile apps) and time-intensive decision making.
Traditional RPA software automates manual and repetitive tasks without using AI. It simply records and replicates employee actions and mouse clicks to generate an invoice or report, for example. This only works if the process never changes. Most traditional RPA tools must be updated when invoice layouts or reporting requirements change, for instance.
Intelligent RPA systems from vendors such as UiPath and Automation Anywhere add machine learning, NLP and machine vision to RPA tools. Instead of just replicating human action, intelligent RPA finds the most efficient way to automate tasks while dealing with new data and changing requirements.
AI-enhanced RPA can reduce time and cost across three phases in the claims life cycle: data input, validation, and adjudication.
In the data input and validation phases, RPA bots automate manual claims entry of data from multiple sources. Machine vision and machine learning may enable bots to adapt to changing templates, document formats and validation rules. Machine learning also allows bots to learn and improve when fed new data or when someone flags a mistake.
The adjudication phase determines whether the claim should be paid or denied. Machine learning and NLP allow bots to analyze past claims decisions and read through insurance policy documents to determine how to handle the current claim.
Intelligent RPA is powerful because it can automate tasks that require multiple people across teams. Properly deployed, it can free up claims staff to focus their effort on high value and complex claims.
AI for Risk Reduction: Improved Underwriting and Fraud Detection
AI-enhanced underwriting provides three benefits: better risk management through data insights, automating tasks to better use underwriters’ time, and improved competitive advantage from writing higher quality business.
Most insurers only process 10–15 percent of the data they have.
Underwriting traditionally relies on experience, rules and judgement. Underwriters consider past policies, risk tables, client profiles, claims history, sector/market risks, and assess how a new policy impacts the overall risk of the portfolio.
Improving underwriting standards is critical in a post-pandemic economy. Insurers cannot afford to misjudge risks and incur large losses in this environment. The challenge is that underwriters are stretched thin by having to evaluate large numbers of insurance policies.
According to an Accenture study, most insurers only process 10–15 percent of the data they have. There are too many data sources — past policies, telematics (car insurance), wearables (health insurance), third-party databases (e.g. shipping data), and social media. Humans can’t process this information at scale. Even automated underwriting tools only help so much since they are not ‘smart’ enough to handle anything beyond simple cases.
AI can help underwriters assess policy risk faster and more accurately. Machine learning and NLP tools can ‘read’ and understand past insurance policies and help underwriters construct a robust policy for a new risk. Predictive analytics tools can study past claims, losses and other risk metrics for similar insurance policies to predict the likelihood and magnitude of future losses.
AI tools also help underwriters monitor risks over time, which is useful when insurance policies are renewed. Cape Analytics, an AI startup, uses geospatial imagery to photograph real estate from above. Machine vision then tracks how properties change over time. Underwriters can use these insights when renewing insurance policies for these properties.
AI will not replace human underwriters. AI tools simply provide higher quality data insights. It is up to underwriters to use these insights to build a profitable book of business. AI-driven analysis also frees up underwriters to focus on complex and high value business.
The FBI estimates non-health insurance fraud in the US at over $40 billion per year, which can cost families between $400–700 per year in extra premiums.
Machine learning can analyze historical claims and customer histories to detect potential fraud. Machine vision and image recognition can study pictures and video of potential damage to cars or houses and flag suspicious cases.
Ant Financial, the fintech arm of Alibaba group, built a mobile app for car insurance called Dingsunbao. Powered by machine vision, the app uses the phone’s camera to detect damage to cars and instantly pays out claims. In theory, the app may also be able to flag suspicious claims where cars are damaged on purpose.
Fraud detection is essentially pattern recognition — identifying features that match past cases of fraud. Machine learning is ideal for this. Models trained on large datasets of past fraud can detect suspicious patterns in new claims faster than humans can. Furthermore, the models get better as they are fed more real-world data.
AI for Customer Insight: Enterprise Search Software
Insurance companies sit on more customer data than most other industries. If you’re an individual customer, insurers know about your life stage, family, medical & travel history, home & car ownership, and more. If you’re a corporate customer, they know a lot about your business and your employees. They also know about your insurance history.
In a perfect world, insurers would use this data to develop a 360-degree view of any customer. Customer service would be seamless. Up-selling and cross-selling opportunities would easily be identified.
In reality, customer data is stored across disconnected systems and in different formats (e.g documents, email, images, PDFs). Datasets are often too large to search through in a timely manner and are also restricted by access rights.
AI-enabled enterprise search software enables employees to search through the company’s digital systems, including document databases, CRM systems, emails, websites, documentation, call center logs, among others. Staff can search using questions or keywords and pull up the relevant media that answers the question.
Enterprise search systems use machine vision to recognize images and text on screen. Optical character recognition (OCR) is used to digitize text in scanned PDFs to word documents or spreadsheets. NLP is used to interpret text.
The rise of remote work makes enterprise search systems even more compelling. Staff can search for information without spending time chasing down colleagues.
Claims adjusters investigating a claim can search for similar claims (e.g. Silver 2016 Toyota Corolla in the past 2 years). They can also pull up supporting documentation and past cases of fraud.
Underwriters can search for past customers and policies to price a new policy. Salespeople can pull up information on a customer to identify the products and services they most likely need.
Customer service agents can access customer information during a call and increase the response time for customer queries. Good customer service is critical in retaining customers during economic downturns when companies compete for a shrinking pool of customer dollars.
A 360-degree customer view is a strong competitive advantage in a post-pandemic economy because it helps companies retain good customers and attract new customers. More importantly, these customer insights enable insurers to create and market the right product mix that will propel them to growth and greater market share when the economy recovers.
Takeaways for Insurance Executives
Insurance companies’ first instinct after Covid-19 will be cost cutting and risk reduction. While some executives will tighten their purse, others have the chance to make small strategic AI investments that have the potential for quick ROI. More importantly, AI adoption now can give proactive insurers a head start when the economy turns around.
How should insurance executives and business leaders approach AI in these challenging times? How should they select AI projects? Should they build AI tools internally or buy from vendors?
The answers will vary according to the company’s unique situation. In general, companies should look at low-cost and high-ROI projects in the short term. Executives must also have a deep understanding of their business needs, pain points, and AI use cases. Then, they should select the AI tools that address these needs. Some companies go out and buy AI products and try to figure out their business needs later — this usually ends badly.
Building AI tools internally or buying vendor software both have their pros and cons. While building internally takes time, engineers and data scientists, the product is tailored to your needs. Buying from vendors is faster and cheaper, but the product may not be fully compatible with your processes and data.
A hybrid approach where insurers work with vendors to develop customized tools is another approach. This speeds up implementation while ensuring good product fit.
Finally, implementing AI projects requires both tech and business teams to work together. Business leaders and functional experts are crucial in ensuring that AI solutions fit current business needs and can be tailored to respond to changing requirements.
Contact us at firstname.lastname@example.org to discuss how your company can identify & implement high-value AI solutions. Feel free to contact our management team directly as well.
Managing Director, Financial Services
About the Author
Raj Shroff is an AI Consultant with a background in Insurance. Prior to joining Blu, he was with AXA in Hong Kong.