Vitech, a global provider of cloud-centered benefit and investment administration software, faced challenges with unstructured and isolated data across various internal platforms. The lack of a centralized and easily navigable knowledge system affected productivity and data consistency. To address this, Vitech collaborated with AWS to build VitechIQ, an AI-powered chatbot that accesses an internal repository of documentation.
Built using generative AI with Amazon Bedrock, VitechIQ provides users direct access to crucial information, removing the need to sift through multiple platforms. With the support of AWS Knowledge Bases, a fully managed capability for Amazon Bedrock, Vitech can implement entire Retrieval Augment Generation (RAG) workflows without having to manually build integrations or manage data flows.
To develop VitechIQ, technical components were carefully selected based on respective evaluation criteria. For instance, Amazon Bedrock was chosen to host Large Language Models (LLMs), taking advantage of its serverless experience and seamless integration with AWS tools. For data storage, Amazon Simple Storage Service (Amazon S3) was used, storing mostly .pdf documents. Amazon Aurora PostgreSQL was used to store vector embeddings due to its excellent performance and compatibility with Vitech’s existing systems.
VitechIQ uses different LLM models to convert document data into numerical representations and perform semantic searches to generate user-specific results. The application uses AWS tools to create a secure environment, and its logs are monitored with Amazon CloudWatch.
Since its inception, VitechIQ has centralized Vitech’s knowledge hub, leading to enhanced efficiency with a reduction of over 50% in queries to product teams. The interactive AI chatbot is used by approximately 50 users daily, equating to around 2,000 queries monthly. Being adaptable, VitechIQ has extended its scope to include Vitech’s extensive API documentation, assisting development teams with their knowledge retrieval.
Moving forward, Vitech aims to enhance VitechIQ by using LangChain Expression Language (LCEL), modernizing the Streamlit application to host on Docker, and automating the document upload process. They are also exploring opportunities to create similar capabilities for external customers, maintaining their commitment to using AI for operational efficiency. This initiative showcases how AWS and Vitech work together to drive technological innovation and upscaling, setting a new standard in the corporate world.