When to use a standard agent or a workflow agent
You influence how an agent processes information and which information is taken into account by adding agent skills to your request. This provides additional knowledge and context to an agent. When you use a standard agent, you can associate a skill. When you use a workflow agent, you can associate skills to a step in the workflow.
|
Key feature |
Use a standard agent |
Use a workflow agent |
|---|---|---|
|
Control |
When you want the AI to decide how to execute the instructions. This is ideal when you want the agent to be creative or adaptive. |
When you want to explicitly define the execution path or when you need deterministic, predictable behavior. |
|
Resilience |
When the skill-based executions carried out by agents are completed in a single run. This means that if there is a failure, the entire operation must be run again. |
When you want to have checkpoints that allow execution to be paused and restarted. This means you can resume the workflow from the last successful step after a failure. This is particularly important when the cost of re-running the entire process is high (for example, when generating large artifacts, or calling external systems). |
|
Impact |
When you can run the same operation multiple times with the same input and the end result is the same or when the impact is low-risk. |
When steps results in an action, such as sending emails, charging payments, writing back to Sitecore at scale, that should not be repeated on retry. |
|
Complexity |
When only one agent is required for focused, single-domain tasks. |
When you have multi-step business processes that coordinate multiple agents, human approvals, or external system integrations. |