Key Findings
44% of technology professionals at large enterprises run active discovery tooling to verify their inventory of agents, MCP servers, and LLMs.
74% of technology professionals at large enterprises are confident their testing would catch a production-impacting failure.
19% of technology professionals at large enterprises have a gate that automatically blocks every bad release.
76% of technology professionals at large enterprises believe they could disable a misbehaving agent in under 15 minutes.
33% of technology professionals at large enterprises have an instant kill switch in place for misbehaving agents.
87% of technology professionals at large enterprises experienced security incidents overall.
74% of technology professionals at large enterprises say they have a complete picture of true spend per agent.
60% of technology professionals at large enterprises overran their budget last quarter.
42% of technology professionals at large enterprises run prompt edits through the same pipeline as a code change.
53% of technology professionals at large enterprises route agent-related changes through any standard pipeline before production.
37% of technology professionals at large enterprises run less than half of their agent-related changes through a standard pipeline.
More than 40% of organizations decide case by case whether to trust an agent change.
58% of organizations that promote agent changes to production check every change against a fixed, repeatable standard.
58% of technology professionals at large enterprises report an increase in production incidents per 100 changes since deploying AI agents.
75% of technology professionals at large enterprises say their agents are secure end to end.
88% of the technology professionals who say their agents are secure end to end experienced security incidents.
77% of technology professionals at large enterprises are confident they have a complete inventory of every agent, MCP server, and LLM in their environment.
34% of technology professionals at large enterprises have a dedicated configuration system for AI behavior.