How AI could change insurance

Expert risk article | November 2023
Will AI be an evolution or a revolution in the insurance space? Here we discuss its likely impacts on the industry and hear from 鶹ýapp risk and technology experts about the company’s own plans to explore its potential responsibly and creatively.
  • The insurance industry is believed to be among those with the highest potential value from AI technologies, which could add a potential $1.1trn annually.
  •  AI technologies will allow insurers to draw on larger datasets to improve their processes, from automating customer support to honing risk modeling and predictions.
  •  Data-driven assets and AI could transform claims management, including prevention, notification, settlements, and fraud detection.
  •  Along with the opportunities offered by GenAI come risks around data protection, confidentiality, the threat of cyber-attack, ethical concerns, and liability exposures. 

The numbers are staggering. ChatGPT amassed 100 million users in its first two months [1], only for that record to be smashed a few months later by rival Threads, with 100 million users in just five days [2].

It took TikTok roughly nine months to reach that number of users and Instagram two and a half years, but the uptake of ChatGPT after it launched in November 2022 was unprecedented. The speed of its adoption prompted a wave of speculation in the media and business world about the disruptive potential of such technology.

With digitalization increasing rapidly in many areas of life, the amount of data that can be leveraged is proliferating. As a data-driven industry, insurance is not new to artificial intelligence (AI) or using data analytics across its value chain to improve products, interactions, prevention, claims and processes. According to several studies, the industry is among those with the highest value potential from AI. 

Estimates point to the market potential of generative AI (GenAI) reaching $15bn by 2025 and $32bn by 2027 in the insurance and finance industries alone. McKinsey foresees AI technologies could add up to $1.1trn [3] in potential annual value for the global insurance industry.

Put simply, there are two types of AI application. The first uses AI’s ability to identify patterns in data and draw conclusions or make predictions from them. The second, GenAI, uses large language models (LLMs) trained on vast existing datasets to generate new content that mimics human creativity, whether it is images, text, coding, music, art, or interactive simulations.

“AI enables insurers to enhance their value proposition by better predicting and therefore preventing risks,” says Michael Bruch, Global Head of Risk Consulting Advisory Services, 鶹ýapp Commercial. “AI depends on having good-quality data. We are constantly evolving and expanding our data quality across the business at 鶹ýapp – not only at 鶹ýapp Commercial, but across all 鶹ýapp entities – to train the models we use. This can help us assess and model extreme weather events, for example, gathering data on secondary perils such as floods. Or, by constantly improving the granularity of our location data, we can help corporate customers better identify climate-risk exposures.”

AI can take on the role of “cool problem solver,” adds Bruch. “By enabling us to adopt a more predictive, preventive, and proactive approach, AI can shift our perspective from looking behind us in the rear-view mirror, and paying out claims, to evolving into an organization with a sharper view of the road ahead, supporting clients in preventing and mitigating risk, and avoiding losses. The power of AI-generated insights can help businesses and societies become more resilient.”

As more data and risk intelligence becomes available, particularly for weather or cyber exposures, AI could enable insurers to provide coverage for emerging risks and technologies that are not yet insurable today, including parametric solutions. 

AI and data-driven assets are being implemented at 鶹ýapp to prevent loss and improve the claims journey.

Prevention: The Weather Alert tool predicts the location, timing, and impact of extreme weather events, with 96% of customers saying they take steps to prevent damage as a result of an alert. The tool has 1.9 million registered customers.

Notification: Another AI tool detects motor accidents in real time using an app and tag on windshields, or via sensors in connected cars or smartphones. With 40,000 active users, it provides simple and digital first notification of loss, allowing 鶹ýapp to initiate contact with the client.

Claims assessments and settlements: AI assets enable immediate coverage checks, more efficient claims assessments and solutions, such as automated preassessments in motor insurance and an extensive repair shop network, and faster assessments of low claims for efficient settlements.

Detecting fraud: AI-based tools efficiently utilize available data to check for fraudulent activity, including deceptive language, inconsistencies, or unusual behaviors; manipulated images or video evidence; predicting the likelihood of fraud based on historical data; and social network analysis to uncover fraudsters among claimants, insured persons and witnesses.

The potential of GenAI across the insurance value chain is an exciting prospect. Open AI’s ChatGPT hit the headlines first, but it has now been joined by Microsoft’s Co-Pilot and Google’s Bard as the most prominent tools in the market.

“The new technology means insurers will be able to draw on larger datasets to improve their processes, from automating customer support and manual tasks to honing risk modeling and predictions. It will also facilitate data entry, data cleansing and the classification of data,” says Meenesh Mistry, Business Model Transformation Executive at 鶹ýapp Commercial.

A wider range of risk data will enable underwriters to offer more targeted, bespoke insurance solutions and smarter pricing. GenAI will also automate underwriting tasks, including data extraction and wording comparison.

“In corporate insurance, customers provide detailed risk information and questionnaires that can exceed 100 pages,” explains Ulrich Kadow, Head of Global Product Management and Underwriting Transformation at 鶹ýapp Commercial. “Using AI, key information could be extracted and clearly displayed to the underwriter and risk consultant, leaving more time for the ultimate risk assessment. Similarly, broker wordings can be compared against an 鶹ýapp master product wording, with GenAI highlighting key differences. By streamlining tasks like these, underwriters can focus on more complex challenges.” 

Larger datasets will also provide more scope for GenAI to pick up anomalies or unusual patterns of behavior that could indicate fraud, says Kadow. Loss triggers could be spotted, and future claims could be better predicted, including claims surges. GenAI’s ability to analyze images and videos could be another useful tool in the claims process. 

Investments by 鶹ýapp in IT totaled $5.7bn globally in 2022, a significant portion of which was dedicated to AI technologies. At 鶹ýapp Commercial, a GenAI exploration program has been set up in partnership with 鶹ýapp Consulting and Microsoft to investigate a number of use cases globally, including an internal application that offers all the functionalities of public LLMs but is safe and secure to use. These use cases are being tracked across the 鶹ýapp Group, to identify synergy potential across different entities and share learnings.

Example use cases at 鶹ýapp Commercial include training intelligent chatbots to answer queries about risk appetite and underwriting, 24/7, in multiple languages on multiple channels, and using GenAI to summarize key exposures and generate content using cited sources and databases for enhanced risk assessment.

鶹ýapp plans to roll out GenAI-enabled business software in due course and all employees are being encouraged to experiment and learn how to apply this new technology at work. The company actively supports its employees to become fit for AI. Its Fit4IT and security awareness initiatives aim to boost employees’ digital skillset, and more than 7,000 employees have participated in training offered by an internal Data Analytics Academy.

The timely processing and settlement of 鶹ýapp Commercial marine claims in North America has been boosted by the introduction of Neptune, an AI-enabled tool with a management ‘cockpit’. Based on real-time information, such as workload across department, teams, and adjusters, it automates the assignment of claims cases to adjusters in the first notification of loss (FNOL) stage and provides real-time KPI to claims managers. 

Previously, the claims team had to manually assign each new claim to the appropriate claims adjuster based on static business rules and incomplete information. Now they can do it with the click of a few buttons and have a daily updated overview of workloads.

Although they are the focus of much excitement, GenAI solutions are in the early stages of development, with risks and limitations insurers must address before the technologies can be safely rolled out.

“We see LLMs working very well for general language tasks, but they are not yet designed for the specific technical language we use in insurance, such as legal analysis of policy wordings or claims handling,” says Mistry. “GenAI models need to be further trained and customized for the language, data, and processes of our industry. This is the focus of our use cases.”

When it comes to affecting employment in insurance, Bruch sees the impact of AI as “an evolution rather than a revolution”, creating new jobs and changing existing job profiles gradually over a number of years. “AI is not a replacement for human emotional intelligence, which is so important in the workplace. But I see new roles being created for professionals who can develop, implement, and manage AI systems and evolve existing ones. AI will largely affect job profiles as a supporting, complementary technology, providing vast opportunities for digital upskilling.”

Along with the opportunities offered by GenAI come significant risks such as data protection, confidentiality, ethical concerns, and liability exposures. AI and GenAI solutions are only as effective as the data they are trained on, and historical data can reflect inequalities GenAI could perpetuate by creating biased content, possibly leading to discrimination. 

GenAI can also create convincing-looking information that is factually incorrect, a process called ‘hallucination’. This is already being used by cyber criminals to commit new types of fraud, phishing emails, or generate ‘deep fakes’ – digitally altered hoax videos using footage of real people (see panel).

“Any use of AI at 鶹ýapp must be compliant, safe, responsible and within clear guardrails,” says Christopher Rau, Head of Digital Law and Data Protection at 鶹ýapp Commercial. “We are committed to maintaining our five principles for . As a global company, we have to consider a variety of regulations around the world, including AI regulations as they evolve in different territories, such as the recent EU AI Act.”

AI has been a key strategic topic at 鶹ýapp for several years, and new trends are continuously monitored and explored by the company’s global team of data scientists and AI experts. Data is a core element of the business. Technology and data analytics are increasingly at the heart of what insurers do.

Threat actors are already using AI-powered language models like ChatGPT to write code. GenAI can help less technically proficient threat actors write their own code or create new strains and variations of existing ransomware, potentially increasing the number of attacks they can execute. 

“We can expect an increased utilization of AI by malicious actors in the future, necessitating even more stronger cyber security measures,” explains Rishi Baviskar, Global Head of Cyber Risk Consulting, 鶹ýapp Commercial. 

“AI can be used to carry out more automated attacks, as well as develop new techniques to steal or poison data. When you think about the potential to combine AI with the proliferation of the Internet of Things (IoT) and the speed of 5G, for example, we may have serious issues on the horizon.”

Voice simulation software has been a recent addition to the cyber criminal’s arsenal.

In 2019, the CEO of a British energy provider transferred €220,000 to a scammer after they received a call from what sounded like the head of the unit’s parent company, asking them to wire money to a supplier. The voice was generated using AI [4]. In August 2023 researchers documented instances of deepfake video technology designed and sold for phishing scams. The going rate? Just €250 for a full video. 

It’s not all bad news though. AI will help threat actors, but it is also a powerful tool for detection. 鶹ýapp has a partnership with cyber insurance provider Coalition to offer customers cyber coverage with AI-powered security tools that help them spot, prevent and respond to cyber risks.

“We might see more AI-enabled cyber incidents in the future, but investment in detection backed by AI should also catch more incidents early,” says Baviskar.

[1] Reuters, ChatGPT sets record for fastest-growing user base - analyst note, February 2, 2023
[2] Reuters, Exclusive: ChatGPT traffic slips again for third month in a row, September 7, 2023
[3] McKinsey, Insurer of the future: Are Asian insurers keeping up with AI advances? May 3, 2023
[4] Bloomberg, The Next Wave of Scams Will Be Deepfake 
Video Calls From Your Boss, August 25, 2023

Images: AdobeStock

Keep up to date on all news and insights from 鶹ýapp Commercial