Skip to content

The Intent Stack

A reference model for governing AI agent behavior within organizations

Seven layers. From intent discovery to execution. At every interface where authority is delegated or coordination is required.

BPMN standardizes process. DMN standardizes decisions. CMMN standardizes case management. No standard exists for governing the alignment between agent behavior and organizational intent at runtime. The gap is not in policy guidance — frameworks like NIST AI RMF address that. The gap is not in model-level alignment — Constitutional AI addresses that at training time. The gap is in runtime organizational governance: the infrastructure that discovers what an organization actually intends, formalizes that intent in a form agents can operate against, monitors alignment in real time, and adjusts governance as the relationship matures.

The Intent Stack — 7-Layer Reference Model showing layers from Intent Discovery (L7) through Execution (L1), with Constitutional AI substrate beneath, intent flowing downward, evidence flowing upward, and the three-tier Knowledge Architecture spanning layers.
The Intent Stack — 7-Layer Reference Model, v5.0

Who this is for

  • Standards bodies — OMG, ISO, IEEE members evaluating governance architecture for AI agent deployment
  • BPM & governance practitioners — professionals extending process, decision, and case management governance to AI agent contexts
  • AI governance researchers — studying runtime governance infrastructure for organizational AI deployment
  • Enterprise architects — organizations deploying AI agents at scale who need standardized governance architecture
Read the Specification Engage with the Work

Working specification, v5.0 — March 2026. Subject to revision through operational evidence. Licensed under CC BY 4.0.