Caylent says 59.5% run agents in production

Caylent says 59.5% of 200 surveyed senior enterprise leaders already run AI agents autonomously in production, while 83% rank guardrails at least as important as model intelligence for accelerating adoption. The commissioned Censuswide survey covered U.S. and Canadian organizations with at least 1,000 employees, so it is a directional vendor sample rather than a census of enterprise deployment. The signal supports the control-plane thesis in Google’s agentic enterprise stack: builders selling production agents increasingly need approval, rollback, audit, and least-privilege controls in the product—not stapled on after procurement asks.