Knowledge graph - Scientific news
On February 16, 2026, enterprise Graph AI provider Graphwise announced the commercial availability of Graphwise GraphRAG, a low--code AI workflow engine designed to transition experimental generative AI prototypes into production--grade business architectures. The platform tackles critical operational limitations inherent in conventional vector--only Retrieval--Augmented Generation (RAG), where arbitrary document chunking degrades complex relationships, prompting Large Language Models (LLMs) to suffer from semantic drift and hallucinations. By utilizing an enterprise ontology and knowledge graph as a structured semantic backbone, the platform grounds LLMs directly in verified business data, entity relationships, and domain logic. According to benchmark testing on the multi--hop MuSiQue dataset, augmenting schemaless GraphRAG implementations with an ontology--driven graph reduced LLM response errors by more than 50%. Designed to accelerate time--to--value, the engine incorporates visual pipeline builders, out--of--the--box workflow templates for corporate policy analysis and technical support, and built--in explainability panels providing auditable data provenance for regulated industries.