Roots Automation, a creator of artificial intelligence-powered digital coworkers, announced the release of InsurGPT, a generative AI model designed specifically for the insurance market.
The company’s digital coworkers are AI-powered, digitized employees that can think, read, and intuit like people. They are pre-trained to understand and interact with documents, systems, and processes commonly found in insurance.
InsurGPT further expands the natural language capabilities of digital coworkers by utilizing proprietary large language models to read and extract data across structured documents, such as ACORD and medical claim forms, and unstructured documents, such as insurance applications, loss runs, quotes, and first notices of loss.
“We have fundamentally imbued our Digital Coworkers with deep insurance experience and knowledge through InsurGPT,” said Chaz Perera, co-founder and CEO of Roots Automation, in a company press release. “It simply reads insurance documents like a human would and produces accurate results, fast and at scale.”
The tool also utilizes insurance specific data and documents alongside systems and process knowledge with the aim of improving the accuracy, speed, and validity of data extraction and inferencing, while reducing false-positives that non-industry specific generative AI models can produce. It is natively embedded in all of Roots Automation digital coworkers, and each interaction between a human and a digital coworker improves the InsurGPT model.
“Critical to the development of InsurGPT was the fundamental belief that while companies like OpenAI and Google – and the entire open-source community – have created something great, they are not interested in going deep into the insurance sector, nor have they considered the security, data privacy or heavily regulated nature of the insurance space,” said John Cottongim, co-founder and chief technology officer of Roots Automation, in the release. “We have.”
Roots Automation is based in New York and was founded in 2018.
Source: Roots Automation