ChatGPT and Insurance? Yes! No! Maybe.
My father asked me recently, “What is this ChatGPT thing?” Like everyone else with access to mainstream media, he’s been bombarded with a wild array of ideas—AI is taking over! Google is finished! A robot fell in love with a NYT reporter! There has been a lot of excitement, and a TON of conjecture about what may be coming. We thought this would be a good time to revisit some basics around AI and insurance, take a look at how these new developments might enhance risk engineering and loss control, and explain how the Orbiseed AI differs.
Unprecedented Adoption
Artificial intelligence research lab OpenAI released its now-famous chatbot for public testing on November 30, 2022. What happened next was stunning—ChatGPT reached an estimated 100 million active users within two months of launch, making it the fastest-growing consumer application in history. By comparison, TikTok took nine months to reach 100 million monthly users, and Instagram about 2.5 years. Both Microsoft and Google responded very quickly with AI chatbots of their own, setting off waves of speculation that the internet as we’ve come to know it was about to turn on its head. And indeed, we are already seeing that internet search in general is on the verge of a paradigm shift—generative AI is being used to write school papers, troubleshoot code and even write music.
What IS ChatGPT??
We asked ChatGPT the most basic of questions, “What is ChatGPT?” and it responded with:
Put more simply, AI chatbots are built on large language models (LLMs), algorithms trained on massive scans of the internet. According to tech columnist James Vincent, “These AI tools are trained to predict which word follows the next in any given sentence. As such, they have no hard-coded database of ‘facts’ to draw on — just the ability to write plausible-sounding statements. Instead of serving up a list of links, chatbots can return fluent answers.” This means that ChatGPT can work like an assistant that can perform tasks or answer questions. As a fun example: “Write a paragraph describing the differences between automatic and facultative underwriting.”
What Does This Mean For Insurance?
Right now, innovation teams everywhere are being tasked by executive leadership to look for “AI solutions”. Here are some of the key ways that LLMs are about to overhaul the way we work:
1. Risk assessment: LLMs can be used to analyze a range of data sources, such as social media, news articles, and financial records, to help insurers better understand the risks associated with different customers and products. This can enable insurers to make better decisions about pricing, underwriting, and risk management.
2. Fraud detection: Insurers can identify patterns and anomalies across very large datasets, which can help to detect fraudulent claims and activities. This can reduce risk and save carriers and brokers millions of dollars each year.
3. Customer service: AI-powered chatbots and virtual assistants can be used to provide personalized support to customers, answer questions, and even help file claims. This will increase retention rates, and will be particularly applicable to customer-centric lines such as auto and health insurance.
4. Claims processing: LLMs can be trained to analyze and understand large amounts of data, such as medical records and accident reports, to assist in claims processing. This can help streamline the claims process, reduce errors, and improve customer satisfaction.
Not So Fast…
Here’s where we come back down to earth. While there WILL be a sea change in the way the insurance industry analyzes its massive (and growing exponentially) volume of data, for now, there are limitations that need to be overcome:
Insurance IS fundamentally a risk-averse industry that will be reluctant to adopt technologies that may deliver unpredictable outputs. For customer service applications, automated assistants will need to be 100% reliable.
LLMs are not designed to perform tasks that require reasoning and logical thinking, such as deduction and induction. Risk engineering is all about this, and it’s why we need human-in-the-loop (HITL) interfaces for validation.
Language prediction models struggle with tasks that require deep knowledge of specific subjects or data. For now, they are trained on, and derive their power from very broad data sets.
ChatGPT can only handle text data and is not capable of utilizing tabular data, diagrams, or images.
An LLM can only access written text up to the time it was trained. In the case of ChatGPT, the latest data is from 2021. Real-time data integration is not currently available.
What About Orbiseed?
Orbiseed’s AI is not a LLM. There are several ways in which our platform differs from these language predictor models and which make Orbiseed a more powerful tool for solving specific risk assessment tasks.
The emergence of applications like ChatGPT has brought the power of natural language (NL) AI to the fore, however, a single approach rarely solves the complex problems P&C enterprises face. Orbiseed’s supercharged IDP platform offers the power of the full range of AI approaches – including machine learning, deep learning & knowledge-based – to enable the most cost-effective and accurate solution in a single environment for a range of risk engineering applications.
With Orbiseed, P&C insurers and brokers can:
Apply AI to gain insights from complex property risk data and images.
6X extraction of the data needed to drive intelligent process automation.
Reduce risk with better-than-human accuracy.
Big Things Are Coming
On March 1, OpenAI announced the release of API access to ChatGPT, which means businesses can have custom tools built on the large language model, and chatbots are going to start popping up everywhere. As carriers apply the technology to their own data sets, rapid analysis or summarization of document content will become commonplace. Risk engineers will be able to easily query their historical book of business archive in natural language to ask questions like “What percentage of our properties have adequate sprinkler systems covering 100% of the building footprint?” Convergence with other rapidly developing technologies such as voice recognition and Natural Speech Technologies, will result in seamless automated customer support tools. And the inevitable integration with technologies that can handle both real-time and multi-modal data sources, LLMs will become central components of very broad, deep and powerful insurance industry AI solutions. We can’t wait!
AI is already helping proactive industry leaders to do more with their data. Need help understanding how to make use of these new technologies to enhance your risk assessment workflows?
