Enterprise AI often produces convincing yet inaccurate answers, reducing customer trust and slowing enterprise adoption. Can a structured knowledge graph make AI answers more reliable, and if so, by how much? To find out, we funded an independent study conducted by the UK’s National Innovation Centre for Data (NICD) at Newcastle University.

The research found that AI agents using knowledge graphs instead of vector search alone:

• Were approximately 80% more truthful 
• Answered more than twice as many complex questions correctly
• Hallucinated less with more efficient token use

Join our 30-minute fireside chat with Neo4j Chief Scientist Jim Webber and the NICD research team to learn how this study was conducted and which findings surprised them most. Hear how lightweight graph structures dramatically improved retrieval and reasoning, and what the results reveal about deploying AI with greater confidence.