Roots Automation is teaching AI to read insurance’s messiest data
80% of insurance data is unstructured. Roots Automation built digital coworkers and InsurGPT to make sense of it.
InsureStory
Insurance runs on paperwork nobody can easily read. Broker submissions, legal demands, medical records: unstructured data makes up an estimated 80% of the industry’s information. Failing to process it well leads to inefficiency, fraud and underpricing, adding up to an estimated $100 billion in losses every year.
Roots Automation set out six years ago to fix that, helping insurers turn unstructured data into insight they can act on.
Digital coworkers designed to handle the workload of ten employees.
Its approach rests on two innovations. Digital coworkers are AI assistants that take on labour-intensive tasks, freeing human teams to focus on higher-value work. InsurGPT is the company’s own generative AI model built specifically for insurance, powering those coworkers to work with precision and speed on tasks like claims processing and premium calculations.
Both run on the Roots Autonomous Workforce Platform, which lets insurers deploy AI through simple text or voice commands. Crucially, it connects to existing systems, so insurers don’t have to rip out the technology they already rely on.
Roots is a good example of where insurtech is heading. Instead of replacing insurers, it targets one stubborn, expensive problem and gives existing teams better tools to solve it. As more of the industry’s knowledge sits in documents rather than databases, companies that can read that data well will have a real edge.