Kathmandu. Insurance companies are increasing the use of artificial intelligence (AI). However, according to experts, this technology should be limited to processing claims and detecting problems, but it should not determine whether claims should be paid or not.
In particular, there is a big difference between using AI to gather information, detect discrepancies or process simple queries, and using AI to determine whether to pay a claim, EIP CEO Ross Sinclair said in a statement. According to Sinclair, insurance companies should use their own rules engines to automate the claims process. So that they maintain control over claims decisions.
A survey by S&P Global Ratings found that 81 percent of insurance and reinsurance companies are prioritizing automation. While 71 percent are using AI to improve their internal support functions. Half are focusing on customer service and 39 percent are using AI for risk management (including fraud detection and pricing).
S&P analyst Andreas Lindberg said in a September 21 report, “The use of AI to improve customer experience, underwriting, risk management, and claims processing is growing at a rapid pace. Operating benefits are more visible than financial benefits.” ’
A survey of 121 insurance and reinsurance companies found that 83 percent are still in the early or mid-stages of AI adoption. However, only 8 percent have fully implemented AI projects across their entire organization.
Insurance companies expect AI to improve their efficiency by 7 percent and increase revenue by 5 percent by 2028. In 2025, 16 percent of companies expect cost savings of more than 3 percent through AI. This number is expected to increase to 80 percent by 2028.
While S&P has improved claims handling, underwriting, and customer service, AI still hasn’t had a significant impact on insurance companies’ profits or credit valuations.
More than three-quarters of those surveyed identified data privacy, security, and regulatory compliance as the main barriers to widespread adoption of AI. About 62 percent keep a record of AI models to monitor their use and associated risks.
