AI Agents: Control Plane Built, Agents Lag

Enterprise AI agent orchestration is increasingly concentrating on major model-provider platforms, with Anthropic's Claude leading significantly, chosen for its advanced model and reliable multi-step execution capabilities. However, the reality of deployed "agents" falls short of this ambition. A substantial majority, 71%, of enterprises report that a quarter or fewer of their deployed "agents" are true multi-step orchestrated workflows, with most still functioning as basic chatbot wrappers. This gap highlights that the orchestration layer, including platforms and control planes, is being developed ahead of the actual orchestrated agent portfolio.

Looking towards the end of 2026, enterprises anticipate a hybrid control plane, blending provider-native services with external orchestration tools, to mitigate the primary fear of vendor lock-in, cited by 35% of respondents. Investment is prioritizing agent workflow tooling and security, indicating a focus on building robust operational capabilities. Despite this strategic build-out, fiscal control remains a challenge, with over a quarter unable to prevent runaway agents from incurring significant costs in real-time. The current landscape shows a clear directional signal: enterprises are strategically positioning for advanced agent orchestration while their deployed agent capabilities are still maturing.