The insurance sector, like a large portion of the financial services sector overall, is becoming more interested in implementing AI. The technology has the ability to completely transform the insurance industry by improving accuracy and efficiency in a number of procedures, including underwriting and claims. However, industry leaders are taking a gradual approach as a result of worries about data protection and integration challenges.
Delaying the implementation of AI could result in ongoing inefficiencies, rising expenses, and a decline in competitiveness in a world that is becoming more and more digital. The abundance of data available is arguably the biggest problem facing the insurance industry today, but top research and thought leadership have repeatedly shown that inaccurate and low-quality data can significantly reduce trust.
According to an MIT Sloan research, “the cost of bad data is estimated to be 15% to 25% of revenue for most companies.” This tendency is particularly evident in the insurance sector, where carriers are subject to stringent compliance rules and regulations and underwriters want accurate data for risk evaluations.
How can insurance executives successfully harness the potential of AI technologies while juggling acceptance hurdles? They may minimize risks and optimize advantages by implementing AI gradually and utilizing tried-and-true tactics from other industries, like retail banking.
Ermir Qeli, Head of Data Science & AI at Swiss Re, recently spoke with Emerj Senior Editor Matthew DeMello to obtain a better understanding of the most beneficial strategies for AI adoption in insurance. They talked about how artificial intelligence is changing the insurance market and what insurance executives may take away from the banking industry’s data solutions experience.
Two important takeaways from their discussion will be the main focus of the analysis that follows:
Applying generative artificial intelligence (GenAI) to the methodical organization of unstructured data: concentrating GenAI efforts to address data gathering bottlenecks in the reinsurance and insurance sectors, resulting in more efficient datasets.
Simplifying data flows to precisely evaluate risk and stop fraud: By analyzing these enhanced data sets using machine learning and new generative capabilities, more time can be spent processing claims and less time processing submissions, which will eventually provide the customer with greater value.
Below, you can hear the entire episode:
Brief Recognition: Over his ten years at Swiss Re, Qeli has worked in a variety of roles to solve data science-related business problems and integrate analytics into the core of the organization. In June, he was appointed Head of Data Science & AI. He had a wealth of expertise in retail banking and handled a number of roles centered on data analytics, visualization, and software engineering before joining Swiss Re. After earning a degree in computer science in Tirana, Albania, he went on to the University of Marburg in Germany to earn a PhD in the same field.
Applying GenAI To Systematically Arrange Unstructured Data
The first part of the conversation discusses the ways in which the banking and insurance sectors use data differently. Insurance used to rely mostly on unstructured data, which caused inefficiencies in their systems, as Ermir Qeli points out. The insurance industry frequently requires assistance with complicated IT environments that make data analysis difficult, in contrast to banking, which has witnessed notable advancements in data organization and analysis.
According to Qeli, unstructured data makes data analysis more difficult and is difficult to integrate across different platforms. A disjointed approach to data management results from these challenges. In order to solve this, Qeli proposes that the insurance sector may benefit greatly from AI technology, especially GenAI, which is especially good at processing unstructured data.
He points out that GenAI has made it more easier to access cutting-edge technologies and conduct data analysis. For insurance businesses starting their AI journey, large language models (LLMs), the foundation of GenAI, are a great place to start because they are very good at processing unstructured data.
By employing natural language processing to extract pertinent information from documents with a lot of text, classify data into meaningful chunks, and spot patterns that would be challenging or time-consuming for humans to find by hand, GenAI automates data analytics. With the help of these automations, insurers can process enormous volumes of data more rapidly, with fewer human errors, and with insights that help them make better decisions.
Applications include, for example:
NLP, or natural language processing, is used to analyze and interpret policies.
Sentiment analysis for processing claims and client comments.
Predictive analytics for tailored pricing and client segmentation.
Using image recognition to evaluate claims for property damage.
Nonetheless, Qeli highlights that strong data foundations are essential for utilizing AI at scale and giving the business a competitive moat—a distinct edge gained by using AI technology. Assuring data accessibility, consistency, and quality throughout the company is a key component of strong data foundations.
Insurance leaders must invest in data governance frameworks, data cleaning procedures, and data integration solutions that standardize and harmonize data from many sources in order to achieve successful data hygiene in the insurance organization, according to Qeli.
Even the most sophisticated AI systems cannot provide precise and useful insights without these underpinnings. Building these foundations, according to Qeli, is not just a technical but also an organizational problem that calls for a change in mindset toward data-driven decision-making and ongoing enhancements to data management procedures.
Simplifying Data Flows For Accurate Risk Assessment And Fraud Prevention
Qeli talks about the steps involved in adopting AI, especially for traditional insurance companies that are weighed down by legacy systems and unstructured data. He draws heavily on his experience in retail banking. He suggests a practical strategy that begins with tiny, doable undertakings that show corporate executives their worth. This strategy aids in obtaining support and funding for larger projects.
The use of AI to sort and handle massive amounts of information, including insurance claims and submissions, is one striking example given by Qeli. Businesses can improve workflows and resource allocation by employing LLMs to condense and extract important information from papers. By accelerating claims processing and producing more accurate assessments, using LLMs for document processing not only increases operational efficiency but also improves customer service.
AI is being utilized not only to streamline data processing but also to stop insurance fraud. Qeli describes how abnormalities and patterns in claims data that could point to fraud can be found using machine learning techniques.
AI can, for instance, examine past claims data to spot odd trends, such several claims made from the same address in a brief period of time or claims that greatly depart from accepted standards. AI assists insurers in minimizing fraud-related losses and guaranteeing the prompt processing of valid claims by identifying these questionable claims for additional examination.
Additionally, Qeli promotes cutting-edge AI applications in the insurance sector, like parametric insurance. Parametric insurance uses preset triggers (such as weather indices) to automatically pay out claims, as contrast to traditional indemnity-based insurance, which necessitates a manual assessment of damages. The entire process is data-driven, and both pricing and settlement depend on high-quality data.
Additional application examples include:
Predictive modeling for accurate pricing and risk assessment.
RPA, or robotic process automation, is used to automate repetitive claims processing procedures.
Analysis of telematics data for usage-based insurance schemes.