KEY TAKEAWAYS:

--Technology has made an unprecedented amount of data on natural disasters available to anyone who needs it.

--MGAs and technology companies will start building better disaster-related insurance products.

--They still need more and better tools so their efforts can scale.

----------

We’re four months into 2023, and already California has been hit with multiple flash floods, skiers were left snowless in the French mountains and earthquakes shattered cities in Turkey, Pakistan and the U.K. Natural disasters used to happen once every few years -- some were considered “once in a lifetime.” Today, they have become a part of our daily news agenda. 

Technological advancements have finally democratized access to vast amounts of data on these natural disasters. We’ve never been better prepared to understand what could happen tomorrow. 

Bridging the data gap with new tech 

Tech has a deflationary and democratizing effect. In the past, hurricane data and modeling was beholden to a supercomputer that certain specific people, governments and organizations were given access to. Today, three people in a startup can access this data through the cloud and generate high-performance complex calculations at scale. 

However, you can give a person paints and brushes, but they do not automatically become Picasso. Data is only as valuable as the way you use it. On one hand, the more relevant data you access, the better your risk modeling should become. On the other, if you still rely on legacy technology and processes like Excel spreadsheets, this cornucopia of data will instead create unyielding data pools that don’t produce improved insights. Pricing tomorrow’s risk cannot be done with yesterday’s tools. Only with modern tools and technology can businesses generate actionable insights that feed into complex issues like climate change. 

Data transparency and consent essential 

The sheer amount of natural disasters the world is enduring means that, whether for a traditional insurance behemoth or a scrappy insurtech startup, data is available to better predict the risk of these disasters happening tomorrow. But data has a bad reputation -- for too long, businesses have hidden data mining details in the small print of privacy agreements most never bother to read. Do you know how and why a business is storing your data? 

Data should be used to create better, fairer policies in insurance and to explain when policies change. You can’t argue against an increase in car insurance if there’s data showing you driving recklessly, for example. Data can deliver clarity and causality; it can explain risks and costs, but only when explicit consent is given to access such data. 

See also: Improving Communication During Disasters

Modeling for tomorrow 

We’re seeing a huge amount of investment in products like parametric insurance and other related modern forms of coverage to do with climate-related risks. When you look at the challenges faced by some of the biggest insurance companies in the world, you see that a lot of them are exposed to natural perils, like earthquakes and hurricanes, which means the providers of technology in that space are incredibly motivated to make things better. If clients are facing losses that your models are not allowing for, there is a huge disconnect, which will push incumbents to innovate. 

Of course, there’s always more to be done. It’s the nature of the innovation game. As we see more advancement in data science and analytical technology, we’ll start to see MGAs and other technology companies emerge that can actually build dedicated products for businesses around parametric and disaster-related insurance. In fact, we’re already seeing new entrants like Descartes do exactly that. These are companies that you wouldn’t have been able to build just 10 years ago, but advancements in tech and an increase in skilled analytical professionals have made it possible. A 100-person-strong data science team just wasn’t possible in the past. 

The democratization of data means insurers and reinsurers have the data to make better decisions, but not necessarily the tools to do so at pace and scale. It's incredibly important that catastrophe insurance evolves to support businesses -- it can’t be so expensive no one can afford it. That’s not a useful form of protection, yet catastrophe insurance is going to be critical in certain areas for companies to survive. That means as an industry, we need to leverage all the advances in technology to make more of the world insurable.

KEY TAKEAWAYS:

--The 2022 deals show a growing percentage of Tier 4 insurers and MGAs purchasing core systems, indicating how important digital transformation has become to compete in the P&C market.

--Nearly all deals were deployed in the cloud versus on-premise, signaling a need for modern technology to support core deployments. 

----------

It would be difficult to dispute how critical core systems are in the insurance business – they are the enablers of significant in-and-out interactions and help drive efficiency and profitability for an insurance company. Today, many insurers and MGAs are accelerating their transformational journeys by upgrading their policy, billing and claims systems. And new research suggests that demand for new core implementations is not slowing down. 

A new research report from SMA examines core systems deals completed in 2022 from 18 of the top solution providers in the market today. The transactions span insurers and MGAs and support personal, commercial and combined personal/commercial lines business. What is clear from the research is that the core systems buying is consistent, which, along with continued deployments across all segments over the past few years, indicates a strong market.  

However, a few trends are emerging. The 2022 deals show a growing percentage of Tier 4 insurers and MGAs purchasing core systems, indicating how important digital transformation has become to compete in the P&C market, with more smaller insurers deploying new capabilities. In addition, nearly all deals were deployed in the cloud versus on-premise, signaling a need for modern technology to support core deployments.  

As digital transformation continues to drive innovation, we will continue to see investments in core systems. But it is important for buyers to be mindful of business objectives when choosing solutions – the wrong fit can obstruct goals, but the right solution can enable a digital insurance enterprise to reach new heights.   

For more information on core systems deals and buying trends in 2022, read SMA's recently published research report, "2022 P&C Core Systems Purchasing Trends: Insurance Market Dynamics Shift Foundational Technology Needs.” 

KEY TAKEAWAYS:

--Automation applied to an inefficient operation can simply magnify the inefficiency. 

--You will surely be asked which customers will be affected most by automation, so be prepared.

--Key changes will be cultural, so understand at the outset how much change will be needed -- and tolerated.

----------

Streamlining and automation often get talked about in the same breath, but there’s a big difference.

Streamlining is essentially simplifying an existing process. That is typically done by removing some unnecessary workflows from the larger effort.

By comparison, automation is “cost cutting by tightening the corners and not cutting them,” as Haresh Sippy, chief founder of Tema India, has put it. In the context of insurance pricing, automation typically means connecting disparate systems and data flows more seamlessly. Rather than just simplifying existing processes, these connections bring some overall structure and governance to the workflow, enable scheduling and triggering of activity and allow for reports to monitor progress. 

Automation shouldn’t just be a matter of saving time – important as that often is – it should bring new sources of value to pricing.

Automation done responsibly

Automation allows for doing more with less -- but automation applied to an inefficient operation can simply magnify the inefficiency. 

Traditional machine learning models, for example, have to effectively "fail" to learn. But will they learn fast enough for certain pricing applications? Automation has to be appropriate to the pricing circumstances for which it is intended.

When looking at how to apply automation responsibly, the six standards recommended by Microsoft are a good starting point: accountability; transparency; fairness; reliability and safety; privacy and security; and inclusiveness.

Improvement must be relative to something relevant

Insurers approach the pricing cycle of Analyze – Decide – Deploy in a multitude of ways, so no two automation projects are going to be the same. 

For example, companies working with traditional, generalized, linear models could make significant improvements (up to 40% resource savings in our experience) by automating the simplifying, grouping and curve-fitting factors that could lead to more competitive or segmented pricing. A next step could be the automated tuning of factor parameters and interactions, leading to applications that assist the interpretation of results.

The key is to identify where automation can improve your pricing process and deliver the most value.

Automation may do more than just replace what previously would have been done manually. Machines may reveal pricing insights that wouldn’t typically have been uncovered. Often, automation can serve to triage the value of making rating updates, as we have seen recently with some companies automating the tracking of potential inflation effects on their books of business.

See also: Insurers Turn to Automation

Which customers will be most affected?

In just about every pricing automation project we’ve worked on where companies are, for example, using technology to integrate and update data from multiple systems to adjust their pricing and are aiming to get new pricing to market quicker, the question arises: “Which customers are going to be most affected, and by how much?” 

In the fairly safe knowledge that the question is coming, automate the response, particularly as impact analysis can be extremely time-consuming if done manually.

Another reason for being ready for the question is increasing interest from regulators in understanding how machine learning and automation are driving pricing decisions. 

Key challenges are often cultural

Automation doesn’t necessarily always sit easily with established pricing practices. It pays to determine what those most involved are prepared to let go and the acceptable levels of scrutiny and review of automated processes at the outset. 

There is likely to be a need to introduce new working practices, because breaks or barriers in an automation-enhanced workflow can limit the benefits of automation. For example, a company that aspires to automated delivery of pricing updates can face real problems if the hand-off from pricing/product teams to IT/rate deployment teams is overly manual and complex.

Check out Majesco’s latest research report to better understand the strategic priorities and investments needed to adapt to today’s market challenges and focus areas for new products, value-added services, channels and digital expectations.

Read Now

Sponsored by ITL Partner: Majesco

Similar to how iPhone technology transformed the way we communicate, generative artificial intelligence (AI) is creating new efficiencies for the insurance industry. Most recently, Oliver Wyman has been working with leaders to augment and reinvent significant areas of their business. Here we share our latest research, a primer on generative AI, how insurers can get started, risk considerations, and opportunities and key actions for industry reinvention. 

Read More

Sponsored by ITL Partner: Oliver Wyman

With inflation showing staying power, how can your firm best harness risk, economic disruption and prepare for a potential downturn? Today’s challenging economic environment offers firms the opportunity to Reset4Value — and drive strategic repositioning, smart cost decisions, and fund their organization for a better tomorrow.

Read More

Sponsored by ITL Partner: Oliver Wyman

As the commercial insurance industry continues to evolve, insurers are under increasing pressure to streamline their operations, reduce costs and improve user experiences for agents AND insureds. The discipline of underwriting, involving assessing and pricing risk, is a critical function in this process, and the need for efficiency and thoroughness has never been greater.

One way insurers are addressing these challenges is by leveraging novel data, created by artificial intelligence (AI) to automate underwriting processes. By using machine learning algorithms to analyze vast amounts of data to predict the answers to underwriting questions, insurers can reduce the time and effort required to underwrite policies, while improving accuracy and consistency. It’s not too dissimilar from the way large language models (LLMs) like ChatGPT and Google’s Bard use vast amounts of unstructured web content to predict the next word in a conversational sequence.

Automation and data-driven decision-making also allow commercial underwriters to focus on more complex risks, while leaving routine and straightforward applications to be handled by machines. This not only increases efficiency but also allows underwriters to spend more time on high-value tasks that require their expertise and judgment. This is a MASSIVE win for underwriters who are forced into tactical, rather than strategic roles, manually processing applications rather than making value-added risk evaluations. In a low-complexity line of business like workers' compensation, a client of Planck was able to decrease processing time from hours to minutes and reduce submission errors by 29%, ultimately leading to a hit rate increase of 19%! 

Underwriting automation also improves risk management. Machine learning algorithms can identify patterns and trends that human underwriters may miss, allowing insurers to better predict and manage risk. This can lead to more accurate pricing, as well as fewer claims and losses for the insurer.

AI in underwriting also allows for more personalized risk assessments, which can lead to better pricing and coverage for clients. Insurers can use data from a variety of sources, including social media, satellite imagery and weather data, to gain a more complete picture of risk and tailor policies accordingly. This can help underwriters better understand the specific needs of each client, while also providing them with more targeted recommendations. Through a partnership with a top-three European carrier, Planck was able to identify that 75% of this carrier’s construction book of business was underinsured — leading to a 30% potential increase in revenue. 

See also: Insurers Boosting Their Use of AI

Underwriter augmentation of this sort is not a silver bullet, and it is not intended to replace underwriters entirely. Instead, if correctly deployed, AI serves to magnify underwriter abilities, enhance their effectiveness and ultimately make their jobs easier. 

Underwriting automation driven by AI-generated data can help make commercial underwriters more efficient, accurate and focused. By leveraging technology to streamline processes, insurers can improve customer service, reduce costs and better manage risk.

KEY TAKEAWAYS:

--The discussion has moved to how much analytical work and task automation are possible – and if it is feasible beyond simple repetitive tasks.

--Concerns have heightened about AI’s potential to replace humans and eliminate jobs.

--If it becomes impossible to tell if a person has created an image, email, letter or video, then the potential for fraud skyrockets, and determining liability in claims becomes more difficult.

----------

When I was asked in late 2022 and early 2023 about the implications of generative AI for insurance, my reply was always two-fold. First, I advised that it is absolutely a technology space to monitor closely, given the rapid advancement, broad application and unlimited potential. But secondly, I believed that the near-term use cases for P&C insurance were limited to more horizontal spaces and not so much to insurance-specific applications.

There is no question that even now there is value in automating and enhancing interactions. ChatGPT and similar tools are, at their root, designed for conversational AI – driving informed and automated chatbot-oriented interactions. Many types of communications can benefit from this technology – agency help desks, policyholder inquiries, claims status, internal conversations and many more. In addition, anywhere in the insurance enterprise where there is a need to summarize information, create digital material or extract data is now a possibility for AI to automate. In fact, this is more than a possibility – insurers are already deploying ChatGPT across many use cases.

Recently, I started cataloging all various interesting use cases of ChatGPT and, more broadly, generative AI across industries. However, I abandoned that effort as a hopeless task. There are many articles every day on how someone has used a generative AI tool to write code, pass exams, write papers, create art, images or videos, drive database queries, power conversations and more.

Within just a few months, I have seen the tone of these articles shift from the perspective that the AI output was amazing but not always accurate or high-quality to one where significant progress has been made. It was only a few months ago when it was sometimes easy to tell the difference between human- and AI-generated content; today, the task is far more difficult.

Now the dominant questions are not about whether the technologies are viable for real-world use cases. Rather, the discussion is about how much analytical work and task automation are possible – and if it is feasible beyond simple repetitive tasks (we already have RPA for that). The use cases are rapidly expanding into more complex, industry-specific areas.

This naturally heightens the concern about AI’s potential to replace humans and eliminate jobs. My fundamental view for many years has been that the AI family of technologies will augment humans and elevate the roles of industry professionals. Agents, underwriters, adjusters and others will focus on activities that require deep expertise, experience and empathy. I still believe that is true… but not as strongly as I did in the past.

See also: Google's $100B Mistake--and How to Avoid It

The other main question that arises is about the challenges of determining authorship. If it becomes impossible to tell if a person has created an image, email, letter or video, then the potential for fraud skyrockets, and determining liability in claims becomes more difficult.

The net of this blog is that generative AI in all its forms must be closely monitored by P&C insurers, and governments and the business world must develop the right regulatory/governing framework for AI. Experimentation with the technologies is mandatory. Now is not the time to sit on the sidelines and watch – things are moving too fast for that.

The recent launch of ChatGPT has underscored artificial intelligence’s newfound scale, speed and mass accessibility, as well as its long-term potential to complete complex tasks currently performed by humans. Suddenly, AI is in the headlines everywhere, drawing the attention of commentators, business executives and even policymakers. 

For the insurance industry, AI and machine learning (ML) technologies are by no means new. Across the sector, organizations have begun launching AI and machine learning programs over the past several years and are using the technologies to drive efficiencies in core parts of their business, including underwriting, risk management and claims. More recently, these efforts have intensified, driven in no small part by concerns about an uncertain economy and companies looking to “right size” their workforces.  

According to a survey of IT leaders we conducted in 2023, the insurance industry has been accelerating the use of AI over the past five years. But recently the rate of adoption has recently seen a remarkable surge, with a 30% increase in new AI/ML projects from 2021 to 2022. 62% of insurers say implementation of AI/ML has resulted in an overall reduction of their overall headcount, while other companies have focused on retraining employees whose jobs have been affected by AI. Most remarkably, 81% of insurers cited AI as their leading strategic IT priority, outpacing the use of cybersecurity at 63% and cloud at 58%.  

Over half of insurers say they are using AI/ML for product lifecycle management, as well as to drive innovation and data analysis. Intelligent search, document processing and customer engagement are other fast-growing areas. A sizable majority (65%) said they are leveraging AI and ML to improve speed and efficiency, while half say it is helping them predict business performance, and 46% are turning to these technologies to better manage risk.  

Challenges and Pushback

AI adoption across the industry hasn’t been without its challenges. First and foremost, insurers continue to face internal resistance to implementing projects. Over half (56%) of insurance IT decision makers said they have encountered pushback or scrutiny regarding the use of AI/ML in their organization. This could be the result of differing perspectives between business leaders and IT departments.  

Building trust in the results of AI/ML projects is another common issue. When asked if they feel the data that AI generates is reliable, more respondents (42%) said they only “slightly trust” the data than those (38%) who “completely trust” it. Moreover, less than half say there is sufficient governance in place to safeguard against any misuse of the technologies. 

Finally, more than two-thirds of respondents (67%) identified a shortage of skilled talent as the greatest challenge to greater AI/ML adoption. Other roadblocks include a lack of new use cases, algorithm/model failure and a lack of infrastructure necessary to support AI/ML. Despite these hurdles, 90% of insurers say they have grown their AI and machine learning workforces over the past 12 months.  

See also: OCR Plus AI Opens New Vistas

Room to Grow 

Even though over half of insurers say they’ve already realized substantial benefits from AI/ML, the survey also makes clear that there is substantial room to grow. The list of benefits to date is impressive:

• 81% cite risk reduction and an increased understanding of customers
• 79% have seen increased sales
• 77% have used AI to create more personalized marketing
• 75% say AI has increased productivity
• 73% have seen new revenue streams and operational cost reductions
• 69% cite improved customer satisfaction
• 67% have benefited from faster time to profitability and reduced the cost of product development
• 65% say AI has made them more innovative 

These numbers are remarkable, given the technologies’ relative infancy, and would indicate that the insurance industry has just begun to scratch the surface of what AI can do. As companies assess their existing projects and become more comfortable using artificial intelligence across more parts of the organization, AI will become an increasingly critical strategic differentiator and springboard for business success.

KEY TAKEAWAYS:

--Unless AI systems are trained on accurate, unbiased data by unbiased trainers, they can make faulty decisions on underwriting and claims that create legal and ethical risks.

--If great care is taken to train and review AI systems thoroughly, they can provide a host of benefits in customer service, claims and underwriting.

----------

Artificial Intelligence (AI) is transforming the insurance industry by enabling insurers to process claims more efficiently and accurately while improving customer experience. However, the adoption of AI in claims management and underwriting, while useful, poses significant risks and challenges that insurers must address to ensure successful implementation and avoid costly mistakes. 

Big names in the tech world, such as Elon Musk and Steve Wozniak, recently penned an open letter urging labs to pause the training of AI systems for at least six months or until developers and leaders agree on better rules-of-engagement -- ensuring AI is developed to improve human life and is not weaponized or used for harm. Zurich and Lemonade are some examples of Insurers that are investigating and subsequently creating controversy around ways to leverage this technology.

Here, I will explore the risks and challenges associated with AI in claims management and underwriting, as well as strategies to mitigate them.

Risks of AI in Claims Management

AI systems are only as valuable as the data provided by their creator. Any information beyond what an AI system is fed is then inferred to create what experts refer to as hallucinations. Hallucinations can be very convincing, even though they aren’t based on good data. What constitutes “good” data from “bad” is a mix of factors such as the accuracy of the data, the perspectives, opinions and biases of the human inputting the data and who is regulating or training these individuals.

These are just a few of the risks associated with using AI. But one of the most significant is that AI can be uncontrollable. With more and more competition to create the most knowledgeable AI systems, creators are getting to the point where they can’t understand or predict their behavior or accurately explain how a decision was made. This can be problematic if humans rely on AI’s assessments to determine specific actions, such as denying or paying an insurance claim. 

Following are some other ways these risks apply to claims management:

Bias

From healthcare to recruiting, all businesses are subject to bias. The insurance industry is no exception. The U.S. Department of Commerce reports that one of the most significant risks of AI in claims management is the potential for biases, discrimination, and inaccuracy. If AI systems are trained on biased or inaccurate data, the decisions made by the AI system will be the same. This can result in unfair treatment of certain groups of individuals or incorrect claims processing, leading to reputational damage and legal liabilities.

This problem has been seen more recently regarding the use of FICO scores in the underwriting of insurance policies. Insurers are obligated to ensure that bias is removed from underwriting and claims decisions and that the information being used is accurate.

Legal/Ethical Risks

AI systems also require access to large amounts of data, including sensitive personal information, which can be vulnerable to cyber attacks and data breaches. This is a risk because it can result in significant reputational damage, legal liabilities and loss of customer trust.

There are legal and ethical risks with AI, as well. For example, an AI system denying a claim based on biased or inaccurate data could result in legal action and reputational damage. Additionally, the use of AI in claims raises ethical questions about the role of humans in decision-making and the responsibility of insurers to ensure fair treatment of their customers.

Overcoming the Challenges of AI in Claims Management

The first step in overcoming the challenges of AI in claims management and ensuring successful implementation is to review the data for accuracy and quality. It is crucial to confirm where the data came from, how it was vetted and how sensitive data is protected. This includes regular monitoring and auditing of AI systems to identify and correct errors and biases.

With the amount of content that is created online, it’s easy to pull data that stem from questionable sources. These quality checks can allow insurers to identify and correct inconsistencies in the data. Data can also change over time, so your data must be current and relevant. Routine checks will enable you to do this.

Next, you must understand how the AI system was trained and by whom. Industry experts are better-equipped to provide the knowledge AI systems are fed versus those with just tech experience.

Collaboration between humans and AI systems is another key to ensuring fair and accurate claims processing. Insurers should establish clear guidelines for when and how human intervention should occur in claims processing and confirm that humans are adequately trained to understand and collaborate with AI systems.

An independent review board inside an organization should also be implemented to test the technical efficacy of the AI system and confirm it’s working as expected. This board can also determine the moral and ethical implications and whether the system should exist in the first place.

By implementing these strategies, insurers can overcome the challenges of AI in claims management and leverage the benefits of AI to improve the efficiency and accuracy of claims processing, enhance customer experience and drive better decision-making. 

See also: 'AI' or Just 'I'? Most Adaptable Will Win!

Benefits of AI in Claims Management

While AI poses risks and challenges, it also presents significant benefits in customer service, claims and underwriting.  

Customer Service

One of the primary benefits of AI is that it can process large amounts of data quickly and accurately, reducing the time and resources required for customer service. Here is how AI can enhance the overall customer experience:

• Use chatbots or answer bots to answer customers' questions quickly, reducing wait time
• Submit policy changes and other simple endorsements on behalf of the customer
• Notify the agent or customer of any outstanding items, such as payroll audits
• Note any significant rate increases at renewal and automatically re-shop the policy
• Create and forward policy documents, such as ID cards, declarations pages or COIs

Improved customer service can help to protect margins on commissions as a result.

Claims Management

AI can process claims end-to-end and eliminate customer or agent frustration throughout the entire claims process by:

• Managing First Notice of Loss (FNOL) and First Report of Injury (FROI) correspondence
• Classifying, indexing, extracting and relaying claims data into agency/carrier systems
• Screening for potential fraud and validating eligibility
• Calculating and setting reserves and paying claims
• Identifying, servicing and following up on time-sensitive activity with adjusters, such as legal demands within demands packages
• Forwarding explanations of benefits (EOBs) and policy information
• Identifying accounts for recovery or subrogation based on the value at stake

Conclusion

With the challenges involved with AI and the potentially significant impact it can have on the world, there should be checks and balances to mitigate moral, ethical and legal concerns. As the AI open letter so succinctly states, rather than a race to the top, all future research by AI labs should focus on how to make the systems accurate, trustworthy and safe.

With the challenges involved with AI and the potentially significant impact it can have on the world, there should be checks and balances to mitigate moral, ethical, and legal concerns. By taking a break to establish these standards, we can be better prepared to use AI in the future and continue to revolutionize the insurance industry.