What's an insurance copilot?

February 29, 2024 6:28 AM

A person using a laptop with a screen graphic.

Last February, Microsoft rolled out Copilot, an artificial intelligence assistant that can write poems and code. But what exactly is a copilot and how are insurers using this type of generative AI?

Digital Insurance reached out to insurance professionals to ask how they would define a copilot.

The responses have been lightly edited for clarity.

Sandee Suhrada, principal, Deloitte

In my mind, or at least from Deloitte's perspective, it's a personification of humans with machines. That's the philosophy we kind of follow. If I want to go technical, it's an AI-driven system assisting humans. The whole game is about enhancing productivity, performance, efficiency and intelligence.

For example, in underwriting, a copilot can help analyze internal and external data–customer data, demographic information, any past claims history and lifestyle factors–to create a specific profile for the customer. Based on the profile and the risk analysis, the copilot can help make specific recommendations that align with the risk tolerance of the customer, their financial situation, or their specific insurance needs.

Indranil Bandyopadhyay, principal analyst at Forrester

Indranil-Bandyopadhyay of Forrester — Indranil Bandyopadhyay

It is sort of an engagement assistant or more of an augmentation to any insurance activity. If you're an agent or you're a broker and you need information, be it from a customer perspective or from a product perspective, you can talk or write to the copilot in human language and get a response back.

For example, think about life insurance claims, which can be quite complex, because there are various documents and information coming from all over. As an underwriter you have to read, understand and summarize notes–you can have an assistant quickly do that summary, bringing back relevant information.

Sivan Iram, co-founder and CEO of Capitola

Copilot, for us, is an AI-based system that is able to take vast amounts of data and generate insights based off of it. We use it under the hood for our own internal purposes.

At Capitola, we think of ourselves as a new generation wholesale brokerage. The vision is to bring efficiency to commercial insurance distribution. As an intermediary, there's a lot of back and forth, there's lots of information coming in from agents to Capitola. What we do with copilot is we employ it in various spaces, one good example would be risk appetite matching.

Alok Bhargava, principal, EY Americas insurance data & analytics leader at Ernst & Young

The term copilot gained popularity over the past 10-12 months following the rise in popularity and use cases of AI to enhance technology transformation and streamline operations within companies.

However, the interpretation and application of 'copilot' can vary greatly depending on the specific context in which the term is used.

I would characterize a copilot as an initial experiment that incorporates a new capability into an existing ecosystem. This is formally done in a smaller scope to assess the feasibility, effectiveness, and potential effects of this new capability, before deciding to allocate resources for a larger-scale project.

INSTANDA CTO Kevin Gaut

A digital insurance copilot is an advanced software assistant that uses AI and machine learning algorithms to support, streamline, and enhance the tasks and activities of human users; much like a copilot assists the pilot in operating an aircraft. Unlike traditional AI assistants that respond to direct commands and perform individual tasks, copilots offer a more fluid, ongoing form of assistance by understanding context, anticipating need, and learning from user interaction.

In insurance, this digital assistance helps to streamline processes, make informed decisions, understand the data, and improve customer experience. One example is utilizing chatbots to answer specific questions and provide customers with an instant, personalized response. Other examples could be using AI to offer personalized policy recommendations based on customer data analysis or leveraging machine learning to predict future claims based on historical data.

Michael Nadel, senior director insurance at Simon-Kucher & Partners

Ultimately, it's really intended to be a system that goes alongside a human-oriented journey. And makes that task or whatever it is they're doing, done in a more simple way. AI can sort of augment either some of the lower value tasks, streamline some of the manual type of lift that's required for whatever it is and enable the human to be more efficient in what it is that they're delivering, without ultimately replacing the task all together.

Julian Balasingam, chief growth officer at Montoux

At Montoux we think of a copilot as an AI-powered assistant designed to help users with various tasks, often providing knowledge, support and automation in different contexts.

We expect that copilots will expand in the breadth of their task automation capabilities, enabling end-to-end workflow automation (likely through multiple copilots or AI agents working together). Furthermore, the capabilities of the underlying large language models are improving at a significant pace, and as such we expect that copilot capabilities will extend past just task-automation and support, to more complex analytical and decision-making support. We see a world in the near future whereby AI copilots can provide insurers with real-time answers to highly complex questions, the sorts of questions that would have traditionally required a team of actuaries or analysts to work on for some time. This additional efficiency, agility and insight will provide the insurers who are at the front of the AI adoption curve with a significant advantage in terms of ability to respond to market movements and the needs of their customers.

Patrick Davis, SVP Data and Analytics at Majesco

From my perspective, a copilot is a new type of user interface for software that assists a user in day-to-day tasks.

In other words, a copilot is a generative AI, natural language-based, interface to software that people are already using with a mouse, keyboard and touch. It has certain capabilities that are more run of the mill, like, 'Hey, can you summarize this data and navigate to and do this other task.' But it's got other capabilities that are really compelling that leverage large language models and generative AI in a much more significant way, summarizing and processing large amounts of structured and unstructured data. Anything that has to do with human language it's particularly good at, and it's particularly good at complicated items and making them simpler. But really, it's a different user interface. It's meant to assist a user and make their day-to-day lives easier, like a copilot in an airplane.

Tiago Henriques, head of research at Coalition

Copilots are chatbots that leverage large language models (LLMs) and generative AI to support users in various tasks and decision-making. Copilots allow users to ask questions and then provide them with guidance.

By using AI to automate time-consuming processes, like answering support questions, ingesting massive amounts of data, or speeding up tasks, human teams can focus on answering questions of facing the problems that AI doesn't have good patterns for or providing services to customers with specific needs, such as those experiencing a cyber incident.

Coalition's internal underwriting copilot helps underwriters by leveraging automation to scan company data for available information like industry codes or the potential number of employees at the company. This saves underwriters time spent searching for these public data points that inform quoting. They can also help with things like automated analysis of a company's privacy policies and terms and conditions, which can impact the types of policies that Coalition can offer to them.