Action
A behavior of an AI system that directly changes something outside the system itself: a database write, a sent message, a payment, a deletion, a call to another system. Actions are the category of risk agentic AI introduces that generative governance was not built to assess. An action does not have to be a direct tool call a human can see. It can also be indirect: an output that is consumed by another machine, a webhook, an event pipeline, an automation rule, rather than by a person, so that a "draft" nobody read still ends up changing something real one step downstream. Apply the first question of the actions-versus-outputs test (Section 3G) to where the chain actually ends, not to the AI system's own last step.
Defined in 3 GAGE programs, which carry 5 distinct definitions of it. The wording above is taught in Certified Agentic AI Governance Professional (CAAGP).
How each discipline defines it
The same term does different work depending on who is using it. These are the definitions as each program teaches them, unedited.
A behavior of an AI system that directly changes something outside the system itself: a database write, a sent message, a payment, a deletion, a call to another system. Actions are the category of risk agentic AI introduces that generative governance was not built to assess. An action does not have to be a direct tool call a human can see. It can also be indirect: an output that is consumed by another machine, a webhook, an event pipeline, an automation rule, rather than by a person, so that a "draft" nobody read still ends up changing something real one step downstream. Apply the first question of the actions-versus-outputs test (Section 3G) to where the chain actually ends, not to the AI system's own last step.
A ROS 2 asynchronous, goal-oriented communication pattern for long-running tasks. A client sends a goal; the server accepts or rejects it, executes the task, publishes periodic feedback, and returns a final result. Supports goal cancellation. Defined by a .action interface file with three sections: goal, result, and feedback.
What somebody commits to do about an issue, with a single accountable owner, a due date and defined completion evidence. The eighth object. One issue may need several actions; each action addresses exactly one issue. Owned in depth by Module 16.
The component of a robotic system responsible for changing the physical world through actuators. The action must be physical (changing position, force, or physical state of something in the environment) to satisfy the robot definition. Digital outputs that do not drive physical change do not satisfy the action requirement.
A long-running goal-based interaction with three phases: a goal sent by a client, periodic feedback during execution, and a final result. Actions are preemptable. Appropriate for navigation goals, manipulation sequences, and any task that takes more than a fraction of a second and where the caller needs progress updates.
Where it is taught
The exact lessons this term appears in. The first module of every program is free with a free account.
- Why Agentic Breaks Generative Governance · Taking the Controls, Certified Agentic AI Governance Professional (CAAGP)
- The GRC information model · Integrated Governance, Risk and Compliance and Enterprise Context, Certified Enterprise Governance, Risk and Compliance Professional (CEGP)
- The definition of a robot: sensing, computation, and action in the physical world · What a Robot Actually Is, Foundations of Robotics and Physical AI
- What ROS 2 is and the coordination problem it solves · Your First Robot Software Stack with ROS 2, Foundations of Robotics and Physical AI
- Services for requests and actions for long tasks · Your First Robot Software Stack with ROS 2, Foundations of Robotics and Physical AI
Where it shows up at work
A definition is where Action starts. The lessons above teach it inside the competencies below, which are what job descriptions ask for and what the GAGE credential records. Every link here is drawn from the same careers map that builds a role's readiness check.
Governs AI agents that take actions: tool access, least privilege, interruptibility, cascading actions and accountability for what an agent did.
One lesson on this page teach it as a primary outcome.
Understands AI that acts in the physical world: perception stacks, vision language action models, sim to real transfer, sensor fusion, and where each of them fails.
One lesson on this page teach it as a primary outcome.
Roles that hire for it
Each role page shows the full competency set, what the seat decides, and a readiness check against your own evidence.
- Chief Information Security Officer, AI security focusneeds it at depth
- AI Security Architectneeds it at depth · 8 open postings this week
- Chief Technology Officer, AI engineering editionneeds working knowledge
- VP of AI Governance and Ethicsneeds working knowledge
- Agentic AI Engineer, orchestration editionneeds working knowledge
- Robotics AI Engineer, embodied systems editionneeds it at depth
How job descriptions say it
Postings rarely use the textbook word. These are the phrases the careers map reads as asking for the competencies above.
Where the frameworks place it: IMDA Model AI Governance Framework for Agentic AI; NIST AI RMF.
One interview question about it
An agent can send emails and update records. What may it touch, what needs a human, and how do you prove afterwards what it did?
Asked of candidates for agentic ai controls and authorization boundaries. Answering it in your own words is exactly how a GAGE topic is passed, and the credential records the pass where an employer can check it.
Terms it appears with
Not an alphabetical neighbourhood: these are the terms taught in the same lessons, ranked by how often they appear together.