Tom Chamberlain, VP Customer & Consulting, hx, takes a look at ChatGPT usage across the insurance sector;
Some companies may be apprehensive when it comes to adopting new AI technologies like Generative AI, which is dominating the headlines. However, insurance giant Zurich seems unafraid, recently announcing its plan to use ChatGPT to ‘improve customer outcomes’. The industry behemoth is already feeding large swathes of claims data in a bid to identify specific causes of loss. But, are we all jumping too fast on a technology that might not be as ‘intelligent’ as we’re all led to believe? The pros are vast but the cons shouldn’t be ignored.
Faster claims makes perfect sense
It’s no surprise this is where Zurich is using generative AI. Processing claims can be a painfully tedious job for insurers involving a large amount of paperwork and data. The lengthy process can create poor claims experiences, which, according to Accenture, could cost insurers billions of dollars and is the key reason why customers choose to switch insurers.
Generative AI has the ability to automate this process so it takes mere seconds by analysing claim forms in real time, identifying discrepancies and pinpointing information gaps that may be present.
Upskill your workforce and customers at speed
One of the more exciting use cases for Generative AI is using it to help teach and upskill a workforce. Take, for example, all of the complicated actuarial or insurance words, acronyms and core topics. Does an underwriter know what a generalised linear model is? Does an actuary have time to explain it?
This isn’t only important within an insurance business, it could also make a huge difference to sales and renewals with customers. Generative AI like ChatGPT could be used to help quickly identify changes in a contract, or explain why your cover increased or decreased. It can make it easier for pricing changes to be more transparent, which in turn helps customers understand and accept those changes.
Do not blindly believe a single source of truth
The internet can provide vast amounts of information at the click of a button, but very few would ask the internet to explain a complex problem and only click and believe the first link they click on. However, generative AI does not share sources and backlinks, it regurgitates data according to an algorithm based on probabilities of what humans sound like. Often, this does not come out as fact.
Questions also arise around biassed programming. ChatGPT is credited for its ability to decline inappropriate requests, but it remains biassed according to Bloomberg. Not only this, but it is also subject to errors in its underlying training data. There are limitations on how much it knows as it relies on what data it has been trained on and when it was last trained. ChatGPT provides no validation or context for its generated outputs, and today cannot be trusted to be factually correct.
Is it worth losing your IP?
A data breach for insurers could result in not only their personal information being stolen, but also their customer’s private information. It can lead to fraud, financial losses, legal liability and significant reputational damage for the insurer and their business.
Now consider being an insurer who is funnelling this data into a generative AI tool that never promised privacy – you’re opening your business and customer data to attack a second time and relying on the vendor to keep it safe. This isn’t a “worst case” scenario, ChatGPT recently confirmed a data breach.
Generative AI and chatbots such as ChatGPT do offer significant benefits to insurers such as faster claims processing and upskilling the workforce. However, insurers should be wary of blindly trusting third-party technology with incredibly important data, especially if that data is part of your business’ IP.
Be the first to comment