How GenAI will change the risk equation for insurers

When a large container ship veered off course and hit the Francis Scott Key Bridge in Baltimore in late March, the entire US East Coast felt the emotional shock of the collapse — and the impact on business. The disaster raised red flags about the ports’ vulnerability and aging infrastructure. It is also set to have a significant impact on the industry, with record shipping losses costing insurers more than 4 billion dollars.

Misallocation of insurers’ unique powers
Insurance is a difficult business, and underwriters have perhaps the most complex task of all. In third-party risk assessment and underwriting, underwriters set premiums for insurers and guarantee payment to these clients when a catastrophe strikes.

Unfortunately, insurers spend most of their time, up to 70%on activities not essential to the underwriting work, such as disputes of incomplete applications, data collection and preparation, and administrative and sales tasks. The the result is significant with lost productivity, longer application processing times and limited number of applications processed, poor customer experience, higher chance of error, pricing issues, suboptimal churn rates and lost business to competitors. The field is ready for correction and begins with generative AI.

How generative AI can help
Generative AI (GenAI) offers the insurance industry a whole new range of capabilities. Its natural language processing capability allows it to understand documents like a human. Large language models (LLMs) can read huge volumes of information and can be tuned to accurately answer questions.

Consider, for example, how GenAI could be applied to the underwriting of workers’ compensation insurance in the US, a liability insurance policy for businesses to pay injured workers for their losses. All 50 states have their own underwriting guidelines for this highly regulated form of insurance. A multi-state entity like McDonalds has to insure its workers in the event of an accident under 50 sets of rules. One state will have a 350-page guidance document, and another will have 1,000 pages. Insurers must follow all the rules of all states.

With traditional AI tools, insurers can automate some data processing tasks, such as organizing and managing these guidance documents. They can perform basic searches and sort information. However, traditional AI capability is limited to manipulating structured data and simple text processing without deep understanding. This left most of the understanding and application of these rules up to insurers.

This is where GenAI helps because it deeply understands unstructured data and the context of each state’s guidelines. It can sort through all those sets of rules, data and documentation at lightning speed and, when well trained, can produce highly accurate assessment results. It can also track any changes in state-specific guidelines and ensure insurers are following the latest guidelines.

A key tool, but not a substitute for human judgment
There is no doubt that GenAI is increasing the capacity of insurers. However, it is not equipped to handle the more nuanced aspects of underwriting that require judgment and ethical thinking – that is still the job of human underwriters.

Also, GenAI is not good navigation. From confidence-shaking inaccuracies to amplifying biases and introducing new cybersecurity concerns, the risks are real and, in some cases, still unknown. So what’s the best move? Keep the human touch in the mix. Relying solely on AI for insurance decisions – be it new business, underwriting or claims – might be tempting, but risky.

Care should be taken to ensure that a GenAI underwriting tool is properly trained on specialized data and leverages reinforcement learning from human feedback (RLHF), augmented generation recovery (CLOTH), and other techniques, to dramatically reduce inaccuracies, hallucinations, and reflections of human biases. It should give the same answer every time for identical circumstances.

By gradually introducing GenAI, starting with small, controlled projects, insurance companies can stay ahead of potential pitfalls, refining their approach as they learn. This thoughtful, human-led strategy paves the way for a safer, AI-enhanced future in industry.

A win for the insurer and the insured
With GenAI, both the insurance company and the insured benefit from a realistic pricing decision, but based on the broadest and most relevant data possible. With the best GenAI solutions, insurers will have access to reliable and explainable results. By comparing similar case volumes, activity patterns and similar factors across insurance scenarios, underwriters can make the most accurate and ethical predictive analytics.

In a few years, most insurance companies are expected to have GenAI support. Initial insurance quotes, even for commercial buildings, will be accessible online with the click of a few buttons. When a commercial customer submits all the necessary documents, within seconds those documents will be processed and GenAI will produce a quote with eloquence and the ability to answer complex questions.

Underwriters are the heart of any insurance company and will remain there, but in the age of AI, they will have an even greater impact on the industry, advancing it faster than ever before.

See more:AI is a top priority, but preparation lags behind for insurersUsing artificial intelligence to streamline workers’ compensation awards