Region: Global
Authors: Kevin Petrie and Merv Adrian

Traditional data teams, long accustomed to SQL tables, struggle to manage and govern unstructured data. But this new frontier—ranging from emails and documents to pictures and video or audio clips—is essential to the success of artificial intelligence. Organizations that apply AI models to these proprietary unstructured objects can enrich outputs, improve accuracy and gain competitive advantage. Data engineers and scientists must acquire new skills, tools and techniques to make these diverse inputs usable for GenAI in particular.  

This survey-based report assesses the state of unstructured data management. We explore requirements, obstacles, use cases, best practices and tool evaluation criteria for teams that support AI. Then we define how leading adopters integrate unstructured data, validate and govern it, then refine features and organize metadata to support model training and inference.

Data, AI and business executives should read this report to learn: