Challenge
Relm Insurance, an innovative insurance company, was grappling with massive volumes of unstructured data scattered across hundreds of thousands of documents. They needed a robust AI solution capable of not only indexing and organizing this data but also providing real-time access and generating meaningful insights. Additionally, their underwriting team required a tool to streamline policy writing, which often involves subjective analysis such as assessing CEO risk or a company's exposure to global events.
Freestone.AI, already working on an AI solution for Relm, sought Intersog's expertise in building advanced AI-powered platforms and quickly scaling development teams with top talent. They needed a trusted partner to customize a platform addressing Relm's specific needs. Intersog's proficiency in AI, data management, and cloud-based solutions made us the ideal choice.
Solution
Together with Freestone.AI, we developed RelmAI, an Azure-based platform designed to transform how Relm's underwriting team retrieves and interacts with data. At its core Relm AI is an AI-powered search engine featuring a generative AI chat interface, providing structured outputs for complex queries.
Transformation of Unstructured Data into Structured Insights
The process begins with data extraction from SharePoint, where documents are fed into an indexer within AI Search. This indexer processes the data to create a structured index based on defined skillsets, efficiently turning unstructured data into structured formats. For example, it compiles financial tables from information buried across multiple files, enabling a comprehensive view of client data.
AI-Driven Enrichment
Beyond indexing, RelmAI enriches the data by adding valuable context. Once a document is indexed, a custom functionality is triggered using Azure Functions. This involves generating brief company descriptions via ChatGPT, classifying and categorizing data, tagging risks, and extracting relevant financial information. This enrichment adds layers of insights, such as contextual information about industry trends like the effects of COVID-19 on the insurance sector. The enriched data is stored in a PostgreSQL database, organized around Relm's clients for easy access.
Intuitive AI Interface for Underwriting
The platform provides an LLM (Large Language Model) frontend to Relm's data, allowing the underwriting team to interact conversationally. They can ask nuanced, subjective questions and receive comprehensive answers, helping them understand and write policies better and faster.
Enhanced Search with Source Verification
Users can search by company or conduct global searches across all data, with the AI delivering the most relevant documents in real time. Recognizing that AI models can sometimes produce inaccurate results (known as "hallucinations"), RelmAI provides source documents alongside AI-generated responses. This encourages users to verify information, ensuring trust and reliability in the insights provided.
Integration of Off-the-Shelf AI Technologies
By incorporating tools like ChatGPT into a custom solution, RelmAI ensures scalability and the ability to grow as new versions and extensions are added to these technologies. This approach leverages existing AI advancements while tailoring them to meet Relm's specific needs.
Security and Compliance
A multi-layered security architecture was implemented using Microsoft Entra ID for authentication, encryption, and secure networking. This ensures all data is protected and accessible only to authorized users, meeting strict security and compliance standards.
Results
RelmAI has fundamentally transformed Relm's approach to data management and underwriting, delivering significant improvements:
Operational Efficiency
Automating the indexing, categorization, and enrichment of documents has dramatically reduced the time users spend searching for information. The team now has access to up-to-date, enriched data in real time.
Improved Analyst Productivity
Analysts can now generate financial tables and reports by querying data previously buried across multiple files. This saves countless hours and improves accuracy, allowing for a broader understanding of client data.
User Satisfaction and Adoption
The intuitive, user-friendly interface requires minimal training, leading to high adoption rates among employees. Providing source documents alongside AI responses has built trust in the system, enhancing user satisfaction.
Scalability and Future-Proofing
Integrating off-the-shelf AI technologies into a custom solution positions RelmAI for future growth. The platform can easily incorporate new features and improvements as AI evolves, ensuring long-term value.
Security and Compliance
Meeting stringent security and compliance standards ensures that RelmAI can grow with the company while keeping sensitive information protected.
Through the collaboration between Freestone.AI and Intersog, Relm now has a state-of-the-art solution that addresses current challenges and positions them for future growth by fully leveraging AI's potential. RelmAI stands as a testament to how integrating advanced AI technologies into business operations can set new industry standards, enhancing efficiency, decision-making, and user satisfaction.
Post Views: 0