FIRST PUBLIC USE
March 20, 2026
- What it is
- The practice of writing down what an agent is for, and what it may not do, before it is capable of doing anything.
- What it is not
- Not prompt engineering and not system instructions. Those tell a model how to behave. This tells the organization what it has agreed to.
- The evidence it produces
- A purpose statement and a prohibition list that were authored before deployment and can be shown to predate it.
- Who is accountable
- The business owner who defined the purpose, jointly with whoever approved the prohibitions.
- The examination question it answers
- What was this agent supposed to do, according to a document written before it did anything?
CANONICAL EXAMPLE
Before a financial services firm deploys a Copilot Studio agent to handle client onboarding, its Intent Architecture document specifies: the data sources the agent may query, the decisions it may make autonomously, the decisions that require human approval, the team accountable for its behavior, and the condition under which it must be taken offline. That document is the Intent Architecture for that agent.
USAGE GUIDANCE FOR CONTENT
Use Intent Architecture as the solution frame that follows the problem frame of Intent Gap. It appears naturally in posts and newsletter editions that move from 'here is what is broken' to 'here is what a governed organization does differently.' It is not a product name or a vendor framework — it is an organizational practice. Do not use it interchangeably with Zero Trust for AI or any Microsoft product.