ChatGPT has captured the imagination of the public, bringing conversations about Generative AI (GenAI) to the fore – but how will it impact the insurance industry?
ChatGPT brought GenAI into the limelight this year, stirring up conversations about how AI might impact the world as we know it.
The chatbot developed by OpenAI was estimated to have reached 100 million monthly active users in January, just two months after its launch. By comparison, it took TikTok nine months after its global launch to reach that number, and Instagram, around two years.
Generative AI comprised less than 1% of total US venture capital (VC) funding of $238.3bn in 2022, according to PitchBook and the National Venture Capital Association (NVCA). But an estimated 450 startups and several new funds for generative AI indicate the potential for growth.
Viju Shamanna, vice president, AI Lab at Ushur, said, “Part of the excitement surrounding ChatGPT 3.5 and Large Language Models (LLMs) over the last few months can be attributed to the fact that they are now easily accessible to the public through a simple conversational interface.”
In addition to essays and poems, it generates emails, blog posts, and code snippets and can handle long-form conversations through prompting.
Tom Chamberlain, VP customer & consulting at hyperexponential (hx), added, “Everyone has read stories about how scarily good ChatGPT is at writing content – ask it an abstract question and more times than not it gives a comprehensible answer.”
News or noise?
With all of the buzz around the chatbot, the question remains: is ChatGPT, or other technologies like it, going to change anything in the insurance industry?
Chamberlain said that although it is not actively used in insurance today, he expects that it is possible we could see it being more involved by the end of the year.
We are already seeing this play out. According to the Financial Times, insurer Zurich has begun testing how it can use ChatGPT in areas such as claims and modelling. The company is reportedly feeding the most recent six years of claims data in an attempt to identify the specific cause of loss across a whole section of claims, with the goal of improving its underwriting.
The use of AI in insurance is not new, however. Ushur’s Shamanna said that over the last decade, AI has enabled various downstream tasks such as image classification, object detection, text processing and classification, information retrieval, conversational AI, document processing and recommendation engines.
Jayne Lansdell, head of IT strategy and governance at Markerstudy Group, added, “AI is already extensively deployed across the insurance industry, protecting customers and underwriters from fraudulent activity by spotting patterns in data and activities.”
More recently, Lansdell said, the use of GenAI, either on-line or via voice channels, has extended into the provision of high frequency/low value customer services interactions. This enables customers to get 24/7 services without the need to queue, be passed around or redirected.
Ushur’s Shamanna added that the insurance industry often uses AI and ML to complete tasks intelligently without the conversational response GenAI provides. One of the most common examples being triaging risk, which saves a huge amount of time for underwriters.
“The AI can categorise quotes, for example using a traffic light system; Green equals yes, Amber means it needs looking at and Red means no. The more you feed into the algorithm confirming or correcting decisions, the better the algorithm becomes. But it doesn’t talk back to you.”
GenAI, however, differs in its transformative power to “augment” some of these tasks and its conversational abilities. Ushur’s Shamanna said, “For example, in a conversational setting, a generative engine such as a Large language model (LLM) can generate more contextual and personalised responses for end users. In a cognitive question-answering system, Generative AI can formulate specific answers in a natural, human-friendly way.”
Another way in which GenAI is attractive, according to Marketstudy Group’s Lansdell, is its ease of application, given that much of the heavy lifting of the technology underpinning the capability has been provided by global and well-known tech brands.
“It can sit on voice channels and online journeys, fronting what we would call ‘live’ customer services which is delivered either through web chat or contact centres. It is the ease and the quality of engagement which makes GenAl a winner.”
If an organisation is training the GenAI well, Lansdell added, it can be like dealing with one of their expert customer service agents without the need to wait in line. The key point of sensitivity is knowing when to escalate the conversation to an agent to resolve complex or sensitive issues.
Potential use cases
The majority of the best applications for GenAI are currently in conversational and assistive applications, such as chatbots, writing assistants and content automation.
According to Lansdell, GenAI or bots are helpful for online journeys where customers want an ‘in the moment’ answer to a question. The customer would be able to ‘say in their own words’ what they want, and the bot will ensure the call gets to the right team to resolve.
“In the customer service arena, it certainly can have benefits, whether that is policyholders getting support around the product they’ve bought, enquiring about cover or making changes or checking payments.”
hx’s Chamberlain explained that hx tested whether ChatGPT could pass a technical whiteboard test. “It wasn’t perfect, but it did a pretty good job, which raises a potential use case – quickly reviewing and documenting code,” he said.
In a similar vein, the technology could scan legal documentation like contracts and simplify them to, for example, quickly pull out the most critical legal parts and explain them. Or it could compare a contract from the previous years to identify any changes.
“One of the more exciting use cases I’ve heard for ChatGPT,” Chamberlain continued, “is using it to help teach and upskill a workforce in complicated actuarial or insurance topics, for example explaining what a generalised linear model is.”
Proceed with caution
When it comes to the use of GenAI in customer service, it remains to be seen what the impact on the customer will be.
On one hand, according to Ushur’s Shamanna, GenAI as well as Large Language Models (LLMs) can play a huge role in creating a digital-first, responsive, and empathetic customer experience. “The future of customer experience will be built on cognitive empathy and scaled through the lens of data and technology. Customers expect contextual and intelligent conversations.”
More and more, Shamanna continued, customers want to engage via seamless journeys in their channel of choice and expect these interactions to deliver their desired outcomes quickly. “True end-to-end automation can be realised by automating front-end conversations along with back office knowledge work automation.”
On the other hand, however, questions have been raised over whether using GenAI in customer service operations could erode trust and frustrate customers.
hx’s Chamberlain said that chatbots aren’t new but still fail to solve a key frustration from customers, “they want to speak to a human, not a robot.” However, if ChatGPT could be integrated to make sure the chatbots sounded like a human, perhaps customers would feel more comfortable and trust the responses given.
“Unless ChatGPT was able to also recreate human-like voices and answers on a phone, it’s unlikely to completely replace customer service jobs, though. A human voice is always preferred and trusted more than a robotic recording.”
Conversely, Markerstudy Group’s Lansdell argues that the use of GenAI in customer service is not missing the human contact, since it is the human contact of many millions of customer services experts and consumer interactions that have informed how the chatbot behaves.
Moreover, she continued, AI is constantly being reviewed and assessed to ensure that the experience does not devalue the customer’s journey and the fine tuning of when to escalate to a ‘live service’ agent is really important to get right. “This will be particular to the service, product and person involved and will need to be specifically developed for each provider and could be the difference between retaining and losing a valued customer.”
Another potential shortfall of GenAI as it currently stands, is it may be lacking in privacy and security measures. hx’s Chamberlain said that for all of the use cases in which the technology could provide value, they require something ChatGPT isn’t yet able to deliver – total privacy.
“ChatGPT is so open and so today you must be careful what you put into it. GenAI solutions that can promise total privacy and security could be a game changer for the industry… It’s incredible watching how people are using ChatGPT, and there’s no doubt it can be used to help the world of insurers, but only if used responsibly and in areas that are relevant.”
Keep up with all the latest FinTech news here.
Copyright © 2023 FinTech Global