The insurance industry is shifting its approach to generative AI as it moves from early excitement to a more realistic phase focused on return on investment (ROI). While companies continue to invest heavily, the big question now is when and how they will see real financial benefits from these technologies.
Recent research from MIT shows that up to 95% of firms across various industries have yet to see clear ROI from their AI projects. However, some experts believe insurance could do better than this average—if companies stop searching for quick fixes and instead focus AI efforts on solving specific operational problems.
Anurag Shah, chief data officer at SIAA, an independent insurance agency alliance, points out that insurance could gain a lot from AI because its core strength lies in data. "Insurance is often slow to adopt new tech, but with AI and data, the impact could be huge," Shah said.
Early signs of AI benefits in insurance are mostly showing up in areas that boost efficiency rather than completely change the business. Tasks like faster quote generation, better renewal rates, shorter claims processing times, and improved employee productivity are where the most tangible improvements appear so far.
On the broker side, AI is helping with renewals, policy comparisons, client messaging, and handling documents. While these are smaller tasks, they can directly affect the bottom line. Shah explains that even a small boost in renewals can be measured clearly and show a good return. For insurance carriers, which operate on a larger scale, AI is being applied to underwriting support, claim prioritization, fraud detection, pricing, and automating customer service. Often, generative AI is used alongside existing machine learning tools to speed up decisions.
Patent data reveals that US property and casualty insurers like State Farm, USAA, and Allstate lead the pack in AI development, holding 77% of all AI patents tracked. Generative AI-related patents have jumped from 4% to 31% since 2023, especially in claims and customer service.
Brokers tend to adopt AI in smaller, faster projects aimed at improving daily operations and keeping clients. These projects cost less and are simpler to implement but tend to bring only incremental gains. Carriers, on the other hand, invest in bigger, riskier programs that touch major parts of their business like underwriting and pricing. These have potential for bigger payoffs but take longer to yield results.
Despite insurance’s rich data sets and skilled workforce, companies still face familiar hurdles like data quality issues, isolated systems, and resistance to change. Deloitte’s survey shows only about 20% of companies overall are achieving strong financial returns and fast impact from AI.
One challenge for insurers is understanding the true cost of AI projects. Shah notes that expenses can vary widely and be unclear, making it hard to predict ROI. This uncertainty reminds him of the early days of the internet, when no one could say for sure how building a website would pay off. Because of this, firms are starting to measure success at the level of specific AI use cases, such as how much faster claims are processed or how client retention improves, rather than looking at the entire AI investment.
Global claims company Davies has seen early benefits from applying generative AI to parts of its claims process. CEO Dan Saulter says it is too soon to measure the full impact but points to time savings for staff as a key indicator of value.
Looking at broader trends, Deloitte reports that most organizations expect to wait two to four years for AI investments to pay off. Only a small number see ROI inside one year. Often, AI gains come alongside other efforts like cleaning data, redesigning processes, and training workers, making it tough to isolate AI’s exact contribution.
Concerns about an AI “bubble” are growing as some big tech companies face stock drops amid heavy AI spending and disappointing earnings. However, Shah argues the market is simply adjusting expectations. He warns against expecting quick wins from flashy generative AI chatbots alone and stresses that improving insurance workflows requires many coordinated efforts.
Companies are becoming more careful with AI projects, focusing on those that prove their value and dropping others. AI investments keep growing, driven by competition and belief in future benefits.
Over the next couple of years, generative AI is expected to continue boosting productivity. Meanwhile, more advanced “agentic” AI systems that can manage complex tasks on their own remain a longer-term goal, with most firms expecting returns in three to five years.
Shah sums up what it will take for AI to succeed in insurance: good quality data, targeted use cases linked to business needs, and strong employee buy-in. He acknowledges that change can be hard and people worry about job losses, but overall AI should create more opportunities than it eliminates.
The insurance industry remains cautiously optimistic that AI will deliver meaningful results, even if the path is slower and more measured than early hype suggested.