Why Large Language Models Fail at Enterprise Data
Show notes
In this episode, I'm joined by Chris Harrison of EmergeGen AI.We talk about why large language models break down on enterprise data, and how his team builds a unified semantic data layer using ontologies and knowledge graphs, an approach he calls "super ontology."Chris shares how their model read and filled out a 300-page commercial lease proposal in seconds, why AI should put humans "on the loop" instead of in it, and how zero-copy architecture lets them work with banks and government without ever seeing or storing the data.We also get into why he believes small, domain-specific language models are the future, the energy and cost problems with giant models, and how quantum computing could change AI forever.