Unlocking autonomous agents: Why smarter data science is the missing link

Agentic artificial intelligence (AI) is moving from theory to deployment. In Foundry’s new research for FeatureByte and Databricks, 57% of the participating enterprises reported deploying agentic AI in production, with an average of five to six new use cases planned over the next year. Despite this momentum, most AI agents remain semiautonomous at best, positioned midway on the agentic intelligence curve — beyond basic automation but short of independent reasoning, leaving the bulk of agentic value unrealized. Progress along that continuum is slowed by data bottlenecks, manual workflows, and limited talent capacity.