Which situation best justifies an agent over a fixed workflow?
Why D is the answer
The path to the answer depends on what is found along the way
Agents earn their overhead when the steps cannot be known in advance.
In detail
- An agent earns its cost when the path depends on what is found.
- Agents cost more to run and test than workflows.
Use case
An illustrative scenario, not a real incident.
A support agent's next step depends on what it finds in the account, so the path cannot be fixed in advance.
See it in different ways
flowchart TD
A{"Are the steps known in advance?"} -->|Yes| B["Fixed workflow"]
A -->|No| C{"Can a wrong step be undone?"}
C -->|Yes| D["Agent"]
C -->|No| E["Agent with human approval"]stateDiagram-v2 [*] --> StepsKnown StepsKnown --> Workflow: yes StepsKnown --> Agent: no Agent --> Supervisor: distinct specialists Supervisor --> [*] Workflow --> [*]
flowchart TD
Q{"Can you list the steps in advance?"}
Q -->|"Yes"| A0["Fixed workflow"]
Q -->|"No"| A1["Agent"]Key point: Choose agents when the path is unknown.
Why the other options are wrong
- A. The steps never change, but the team expects that they may one day
- A possible future change does not justify an agent today.
- B. The task is a single classification with a large number of labels
- A single classification needs one call.
- C. Latency must be as low as possible on every request
- Agents add latency.