DataRobot surveyed 413 AI practitioners and leaders across highly regulated industries to map the full agentic AI lifecycle, including where organizations are investing, where they are stalling, and what it will take to close the gap. This report surfaces five critical unmet needs standing between today’s pilots and a production-ready agent workforce.

The barriers are not theoretical. Across security, infrastructure, cost control, vendor lock in, and agent governance, AI leaders are hitting the same walls, and the stakes are rising as agents move closer to customers, revenue, and compliance boundaries.

Key insights from the report:

• 94% of organizations experience Day 2 Ops issues after deploying agentic AI.
• Only 11% of AI leaders are “Very Satisfied” with their hyperscaler’s agentic AI tooling, yet 72% are still deploying on one.
• 63% require on-premises deployment, but most vendor stacks were not built for it.
• 72% have already exceeded their expected operating budgets; 71% say running agents costs more than building them.