Solving the Insurance Intake Challenge with LLMs and GPT-4
When ChatGPT was first introduced, it attracted a lot of attention and interest. As the technology evolves, there are still a lot of questions on how to best make use of it, where to start and how to do this in a safe and effective way.
We sat down with Tom Wilde, CEO of Indico Data, and Dorota (Dori) Zimnoch, Global Industry Strategist, WW FSI at Microsoft, to discuss how to leverage cutting-edge technology to accelerate intake processes, enabling data-driven decision-making at scale and drive better outcomes through better intake. Find out how Indico Data helps insurers to streamline underwriting, claims, analysis, regulatory compliance and more.
Dori, where do you think insurers can take steps to innovate in their digital transformation?
The insurance sector has made a lot of progress in digital transformation over the past few years, and events like covid probably accelerated this. At Microsoft we’re seeing many clients moving to cloud. It’s great to see that, but we also realize there is still a lot to be done. The insurance industry would not exist without data, it’s the basis for risk analysis. And now we’re at a point where we’re being empowered by technology to finally start transforming this data into insights and insights into actions. Insurers can greatly benefit from new technologies such as generative AI, and we need to make sure that the foundations of these new technologies are embraced.
There is a lot of interest for digital intake – what is your perspective on this?
Digital intake is a very hot topic right now. With the increasing influx of data, it is important that you ask yourself 1. Is the data trustworthy, 2. What’s the quality of the data and 3. How can the data be used? It’s a very challenging topic – and every conversation we have with insurance companies starts with these questions.
What some of our customers are doing to solve the problem of digital intake, is that they focus on the biggest and largest pain points, and solve those first, one at a time, but in a relatively speedy manner. Some customers managed to get a solution to their problems implemented within three to four weeks. It’s also very energizing to see how business and IT come together and start talking the same language. It’s no longer business passing their problems to IT and waiting for them to come up with a solution, but rather they both sit in sparring sessions together and brainstorm: ‘’Okay how do we do that?’’ This is really a game changer for the insurance industry and something I observe daily.
Tom, can you tell us a bit more about Indico Data’s integration with Microsoft’s Azure OpenAI Service?
Of course! We announced the integration two months ago and we’re very excited about it. By extending our Large Language Model (LLM) capabilities with Azure OpenAI’s latest in generative AI technology, we’re inching closer to Indico Data’s overall strategy of providing the leading solution for the transformation of unstructured data into actionable insights. By drastically improving digital intake insurers can achieve more favourable outcomes for their customers and their business. One of our first use cases focused on enabling insurance carriers to increase claims intake capacity, underwriting and improve processing efficiency.
Dori, what should insurance carriers, brokers and the whole industry think about, when starting to apply generative AI?
I think it’s important to link to the business problems, that can be solved with these new technologies. For example, look at your customer journey and start with some small things you can improve. I’ve seen some great examples where processes run much smoother and the customer experience is greatly improved, by applying this technology to support call centre agents, in, for example, handling client questions and preparing and sending out documents immediately after a customer call.
Tom, anything you’d like to add?
The way I’d like to think about GPT and similar new large language models is that it’s very much like a new programming language. It’s a programming language with data as opposed to code. And this has a really transformative effect: everyone in the organization can now become a software programmer without knowing anything about code. By using ChatGPT, you’re essentially programming with prompts. Prompting ChatGPT4 is becoming a new skill set – and also raises questions about managing prompts with the organization. There is a whole new set of mandates to deploy this technology successfully around safety, security and scalability.
There is a lot of talk about AI displacing labour versus enabling labour. Dori, what is your perspective on this?
I think it is helpful to think about this technology as a co-pilot: something that helps you to do and achieve more with less. It’s like an assistant to your side, allowing you to do extra tasks and extra checks, faster analysis of data, summarization, translation, creation of content. But it’s crucial to still have a human being at the end of the line, to agree with what’s been proposed and to have the last say in taking action.
Tom: “AI won’t replace underwriters, but underwriters using AI will replace underwriters who don’t use AI”
Dorota: Totally agree, Tom. We cannot stop this revolution, the technology will move forward regardless. It’s our decision how we respond to that: sit and wait – and miss the train? Or start moving now and start to experiment: start testing the prompts and start testing solutions. I hope this will inspire insurers to get started and test more!
What are some of the things you need to think about when you start deploying this technology across the organization?
Tom: When thinking about deploying this technology across the organization, it’s helpful to think about some of the challenges this will create. For example: compliance and governance. There is also an urgent need for companies to accelerate their cloud strategies. You will need a cloud strategy to use this kind of technologies: GPT4 is a massive model that can’t be lifted up and installed inside a firewall. It can be tailored and fine-tuned for specific customers, though. It is a cloud-based solution, so another motivation for having a robust cloud strategy.
Any final thoughts you’d like to share?
Tom: We spoke about what organizations can do to adopt these technologies. I think it’s also important to think about what you can do as an individual to upskill yourself, familiarize yourself with this new technology and make it a part of your day-to-day life.
Dorota: Totally agree! There are tons of materials out there, free of charge, that can help you educate yourself and help to familiarize yourself with this new technology. Then it’s time to action! Work with your team on developing those prompting skills and provide feedback in a structured way to continue to improve. Human-in-the-loop remains a very important piece of this, to make sure that decisions are robust and accountable. As powerful as GPT4 is, it’s still not perfect in terms of its ability to recall all the relevant information, so the human-in-the-loop will continue to be an important part of the equation.
About Indico Data
Legacy intake solutions have forced organizations to choose between speed and accuracy,
weighing risk versus revenue. With Indico Data’s Intelligent Intake Solution, companies don’t have to compromise. The Indico Data Intelligent Intake solution gives document-intensive industries the best of both worlds, enabling data-driven decision-making at speed and at enterprise scale. Indico Data’s Intelligent Intake solution drives better outcomes through better intake.
Indico Data was founded in 2014 and is headquartered in Boston, USA. Their leadership team brings years of experience and deep expertise in AI and ML-powered solutions.