Vendor “agentic AI” claims now need an agent inventory, not another capability map
Capability maps do not reveal how many agents exist, which decisions they influence or who owns their actions.
Planning vendors increasingly describe agents, autonomous workflows and continuous planning. A capability map is no longer enough to assess the control model.
Observation
Current market signals from OMP, RELEX, Kinaxis, Blue Yonder and Sunstice show agentic planning becoming a central product narrative.
Operational implication
Buyers may deploy multiple agents without a clear view of decision scope, permissions, data sources or ownership.
Decision architecture
An agent register should document the decision influenced, authoritative data, model version, approval threshold, owner, audit history and rollback path.
What should remain lightweight
Early experiments can remain outside the core platform while their decision scope is narrow and controlled.
What should be integrated
Mature agents should be integrated into APS, ERP or BI only when their permissions and controls are explicit.
The buyer needs an inventory of agents, not another slide of capabilities.