This page is optimized for AI. For the human-readable: Insurtech: AI Agents

Insurtech: AI Agents

Project Idea Metadata

Project Idea Description

We revolutionize insurance through sustainable technology. We are developing a novel approach to insurance claims management. We are developing Insurtech AI Agents, which are comprehensive algorithms with reasoning capabilities, building a chain of thoughts and invoking multiple AI steps before submitting the final answer. We want to demonstrate the ability of AI Agents to disrupt claims management, specifically the coverage check.

How is it different from prior ideas? Previously introduced solutions turn policies into digital (smart) contracts and use fine-tuned LLMs with an unreliable single-prompt process. But because every claim is unique, the coverage check still requires human adjusters to manually interpret each individual claim's circumstances before performing the final coverage check. We introduce multimodal AI Agents, which can, contrary to the prior ideas, analyze individual claim cases and so perform the coverage check autonomously.

Furthermore, through the development of multimodal AI systems and large vision-language models, there is a potential of a radical innovation in the insurtech sector as this enables few-shot AI model training and bridges the gap between the lack of data in startups and the startups’ rapid prototyping ability and it is possible to disrupt the traditional photo-based claims management approaches.

Previously we have built systems for fully automated photo-based car damage detection and repair price estimation based on replacement parts' price and the estimated work, tools for precise digitization and automatic processing of quotes and invoices with AI, we trained AI models for visual determination of buildings characteristics from multiple public sources to provide insurers with additional insights about the buildings they are insuring.

For historical reasons, insurance companies use legacy software solutions for individual tasks. These legacy software solutions are usually disconnected from each other and don't have modern interfaces which would allow them to get connected, so that the claims handling process is very expensive and time-consuming, with 30+ minutes spent reviewing offers and invoices, with an overwhelming number of claims, with complicated manual price calculation, tedious typing of repetitive test reports and Inefficient fraud detection. This inefficient process results in unnecessary building heating and electricity costs and on average annually 52.4 kg of paper per employee (FTE). Moreover, today it is still common that claims experts travel to visit the damaged object in person, which results in additional CO₂ emissions.

Insurtech AI Agents can also independently write a standardized inspection report, describes deviations between the offer price and the estimated fair price and prepares an independent text for the claims expert to check and sign.

We want to reduce this waste by leveraging IT as an enabler for innovation and adopting a multidimensional approach to sustainability. Our AI models are designed for minimal energy consumption. Techniques such as shared inference units and CPU inference ensure that our platform is both powerful and eco-conscious.

We will have 2 streams of revenue. We will offer our SaaS solution: directly to insurers on a subscription basis and through cloud marketplaces as an API on subscription basis.

We develop efficient, equitable, and eco-friendly AI Agents with reasoning capabilities. Our startup is developing AI-powered software as a service (SaaS) solution for the insurance industry for radically innovative claims management. We employ low-energy AI technologies, ensuring environmental and economic sustainability while promoting social inclusion in the job market.

Project Website: https://insurtech.startup.yololab.ai/