Graphwise Launches GraphRAG AI Solution
Graph AI provider Graphwise announced the immediate availability of its new GraphRAG solution. The company claims the product uses knowledge graphs and ontologies to provide better context for AI agents, reducing inaccurate answers by half in benchmark tests compared to vector-only retrieval-augmented generation systems.
- Standard Retrieval-Augmented Generation (RAG) systems can suffer from "garbage in, garbage out," where outputs are flawed if the retrieved data is irrelevant, from an unreliable source, or lacks proper context. - Vector-based search, common in basic RAG, struggles with ambiguous context and doesn't explicitly map relationships between different pieces of information, which can limit reasoning capabilities. - Knowledge graphs address this by creating a structured web of information that links concepts, people, places, and their relationships, allowing AI to understand context and connections more deeply. - Graphwise was formed in October 2024 through a merger of two established semantic technology companies: Bulgaria-based Ontotext, known for its GraphDB database, and Austria's Semantic Web Company, which developed the PoolParty semantic suite. - The company's GraphRAG is part of a unified "Graph AI Suite" platform that includes modeling tools and connectors, aiming to reduce the complexity of building AI solutions from multiple vendors. - Graphwise's President is Atanas Kiryakov, and its Vice President of Growth is Andreas Blumauer, who stated that their combined expertise aims to improve the reliability of corporate AI models. - One Graphwise customer, Avalara, a provider of tax software, implemented a GraphRAG system after finding their previous vector-based model lacked the necessary accuracy for critical tax applications. - Industry analysis from Gartner has placed knowledge graphs at the center of its 2024 Impact Radar, alongside Generative AI, highlighting its growing importance in the field.
