Gen AI is only as effective as the people that use it

The good news is that humans will remain a critical part of the technology’s deployment — whereby machine-generated cost efficiencies and informational accuracies are supported by human talent, expertise and experience

Unquestionably, the technological breakthrough of the past few years has been the emergence of generative artificial intelligence, or Gen AI for short. The technology has since accomplished a myriad of achievements that were unthinkable a few years ago, like passing the final MBA exam at Wharton School, or creating a painting that sells for over US$1 million.

Every industry appears to be embracing its capabilities, and the insurance industry is no exception. However, it’s important to acknowledge a number of points.

First, it’s an evolving area for all industries, and there is a strong likelihood that we’ve barely scratched the surface in terms of what it can accomplish for us.

Second, despite talk of Gen AI being able to run our world autonomously, decision-making will and should remain with humans, rather than reside with machines — at least for now. Insurance serves as a good case study to showcase the enormous benefits it brings to businesses; while at the same time, it demonstrates where use of the technology will stop, and human decision-making takes over.

Smart machines, smarter humans

An area benefitting from the technology is underwriting. Before the advent of Gen AI, underwriters could spend days or even weeks researching a specific risk, particularly within more complex lines like property and engineering, to arrive at a specific profile and price. Information pertaining to such lines tends to be very technical, and in the past, this would require a lot of human analysis to understand each risk.

Thanks to Gen AI, we are able to create bite-sized documents that summarise these risks within minutes. Such documents could be one or more pages, depending on how detailed the underwriter wants them to be. From this point onwards, humans take over the underwriting process, and make decisions based on the risk analysis carried out by Gen AI.

At QBE, we’re trialling this approach with a handful of risks. One of them is cyber, where we are leveraging Gen AI to analyse a rapidly expanding list of threats, and understand how these impact customers. It’s a fast-moving area. Hence we need to be on top of how likely these risks could materialise. We also need to establish the potential financial fallout from each of these as well. Gen AI is helping us with this.

More granular analysis provides greater accuracy

Another area is travel insurance. We are now able to provide customers with granular quotations that consider the risk profile of individual airports, airlines, countries, microclimates and more. Thanks to the ability of big data to analyse terabytes of information found in the public domain, we are able to consider the number of bags lost at a particular airport for example, or the likelihood of adverse weather-related events like flooding that could abruptly end a holiday at a specific resort.

QBE is also using Gen AI to process claims faster and with more accuracy. It has been particularly impactful with personal insurance products, like motor insurance or home insurance — high volume and low complexity offerings where we are able to save significant amounts of time processing these.

Notably, Gen AI has proved to be invaluable in fraud detection. The technology is able to spot anomalies that humans might miss, including emerging fraud trends, and alert these to our claims professionals.

Customer trust in data and technology

The quality and accuracy of Gen AI’s work is highly dependent on robust data. Already we have seen multiple examples of biases negatively influencing the findings and actions of Gen AI, with a wide range of companies and industries experiencing reputational and financial damage due to these. It is therefore imperative that AI governance policies are set, and that datasets are diverse, and rigorously tested.

There are also the issues of data privacy and cybersecurity. Keeping data safe and secure is a priority for all companies, irrespective of size or the degree of regulation they are subject to. Yet repeatedly throughout 2024, we saw well respected — and well resourced — organisations fall victim to cyberattacks. Instilling a robust cybersecurity framework, which among many instructions, outlines dos and don’ts, is essential in shielding businesses against these attacks.

A further risk is copyright infringement. When training Gen AI systems, we must ensure all content is sourced ethically, and that such exercises don’t break copyright laws. Conversely, while we are not infringing these laws, others may be. An emerging risk that is challenging to the insurance sector is AI-generated deep fakes. But it is also in this area that AI is deployed to identify these fakes so what is bad, may also be good.

Ultimately our industry must gain trust from customers and wider society, especially when it comes to the use of Gen AI — if customers don’t trust the technology, they won’t trust us. Therefore, wherever possible, we need to be transparent about where we have deployed the technology, how it is transforming both insurance and customer experiences, and the risks as well as benefits associated with its usage.

Some roles may disappear, yet others will emerge

Having the talent to work side-by-side with Gen AI is critical for businesses as well. And despite much being said about the potential of the technology to replace jobs, new roles will emerge. The World Economic Forum estimates that over the past five years, 85 million jobs have been displaced, yet 97 million new jobs have been created.

As the technology evolves, and with it our talent needs, it is also important to work with regulators and government bodies to keep them informed of the latest developments. In some markets like Singapore, Hong Kong and Japan for example, regulators are proactive in setting up frameworks to ensure Gen AI is deployed and developed across a number of sectors, while encouraging workers to upskill in this area.

However, we need to be mindful that some markets are less quick on the AI uptake and will follow suit at a later date. It is therefore important that in such places, we don’t rush ahead with new AI-powered solutions, only to be asked to roll back our advancements.

A bright future awaits

All things considered, the overall future of Gen AI is positive for the insurance sector. Combined with other relatively new tools and technologies, we should soon be able to extract more value from it. For instance, high resolution imagery of infrastructure assets like wind turbines captured by drones can help us assess the risks associated with such technologies to more accurately underwrite these, while also better assessing their damage during the claims process.

The insurance industry can also improve its current Gen AI models by leveraging synthetic data, which mimics real-world data when it is difficult, overly expensive, or simply impossible to access such data.

The above advancements are leading to greater personalisation, which in turn is delivering a richer customer experience. As Gen AI further advances, we will see greater input from both technology and people — combining machine-generated cost efficiencies and informational accuracies, with support from human talent, expertise and experience.

A bright future awaits.

To discuss any of the topics mentioned above, connect with me on LinkedIn. Stay up-to-date with our risk Insights and expertise by following QBE Asia

This article was first published on February 25, 2025.