MissionFlow
A MissionFlow is a mission, its boundaries, and the execution path an agent discovers at runtime to accomplish it.
Previously, this kind of process was usually described as an “AI workflow” or “agentic workflow”. The term MissionFlow makes the difference explicit. A workflow engine can branch, retry and run recovery steps, but its designers enumerate those paths in advance. A MissionFlow defines what must be achieved and within which limits; the agent plans the steps toward that goal and adapts the path at runtime.
Workflow vs. MissionFlow
| Workflow | MissionFlow |
|---|---|
| Designers define the steps, branches and recovery paths | Designers define the goal, the boundaries and the success criteria |
| Runtime selects among paths modelled in advance | The agent plans and re-plans the path at runtime |
| Retries and fallbacks are configured | Alternatives are chosen by the agent, within its permissions |
| Success = the defined process completed | Success = the mission was achieved, within policy |
Bounded autonomy
The autonomy in a MissionFlow is bounded. Every alternative path the agent chooses — another protocol, another tool, a refreshed credential — is still subject to the same authorization and policy constraints as the first. An adaptation that would need a permission the mission does not have is not a valid route; it calls for an approval or a stop.
Example
Mission: “Make sure this file reaches the destination securely before 17:00 and verify that it was processed.”
The agent tries SFTP. It fails. The agent switches to an API the policy already approves for this transfer, refreshes expired credentials it is authorized to renew, delivers the file, verifies the checksum and confirms processing. The mission is the same every time; the path may differ on every run, but never outside the boundaries.
Why it matters
Because the path changes from run to run, checking whether predefined steps executed is not enough. Whether a MissionFlow succeeded, how, at what cost and within which policies has to be read from the path the agent actually took. Observing MissionFlows is the job of AI Flow & Performance Monitoring (AFPM).