Sequential Workflow Pattern
Create a Select AI agent team that runs agent-task pairs in a configured order and passes results between dependent tasks.
The sequential workflow pattern is the current Select AI Agent framework workflow. Set the team process attribute to sequential, and list agent-task pairs in the agents array in the order in which the team must run them. The configured order becomes the workflow plan.
When to Use a Sequential Workflow
Use a sequential workflow when every request follows the same stages and you know the required handoffs before the team runs.
A sequential workflow is suitable for:
- Workflows that use a fixed series of processing or approval stages.
- Requests where one task output becomes input for the next task.
- Teams that need predictable, repeatable, and auditable run order.
- Pipelines that divide work among specialized agents without run-time routing.
- Use cases where all configured agent-task pairs must participate in the workflow.
About Sequential Workflow Agent Teams
A sequential team lists one or more agent-task pairs in the team agents attribute. The framework starts with the first pair, waits for that task to finish, and then runs the next pair. Each task uses the tools configured on that task. Do not list task tools in the team definition.
To pass data between tasks, configure the downstream task with an input attribute that names the upstream task. The downstream task can then use the upstream result with the original user request. The workflow finishes after the configured sequence completes.
The framework does not select, skip, or reorder agent-task pairs in this pattern.
Sequential Workflow Architecture
In a sequential workflow, the agents array defines the run plan. The first agent-task pair receives the request, each dependent task receives the configured prior-task result, and the final configured task produces the team response.
| Phase | Input | Processing | Output |
|---|---|---|---|
| Request submission | User natural-language request submitted to the team | The team starts the first configured agent-task pair. | The sequential workflow starts. |
| First task | User request and task configuration | The agent runs the task and uses the tools configured on that task. | First task result or human-input request. |
| Task handoff | Completed task result and configured task dependency | The framework makes the prior task output available to the dependent downstream task. | The next configured task starts. |
| Remaining tasks | User request, prior task output, and task configuration | Each remaining agent-task pair runs in the configured order after the previous task finishes. | Intermediate result, failure, or human-input request. |
| Workflow completion | Result of the final configured task | The team finishes after the defined sequence completes. | Final team response. |
Runtime Behavior in Sequential Workflows
Learn how sequential teams pause for user input, move from one task to the next, and pass results between configured tasks.
Human Input Handling: When the current task requests human input, the entire team pauses. Continue the same conversation with the requested information or confirmation. After the user responds, the current task resumes and the workflow continues to the next configured task.
Task Progression and Failure Handling: A downstream task does not start until the current task completes. Review team, task, and tool history to identify the failed step, returned error, or workflow that waits for human input.
Context Passing Model: Use explicit task dependencies to pass the result of an upstream task to a downstream task. Keep each task instruction clear about the expected input and result format so that the next agent can process the handoff reliably.
Use Cases for the Sequential Workflow Pattern
Apply sequential Select AI teams when requests move through the same known stages and each stage has a defined responsibility.
- Retrieve operational logs with one agent and analyze them with another agent.
- Gather account or transaction data, validate it, and then generate a report.
- Perform a business-rule check before an agent completes an update or external action.
- Generate content with one task and send it through a notification task.
- Run a standard enrichment, analysis, and summarization pipeline for every request.
Benefits of Sequential Workflow Agent Teams
Use sequential workflows to implement stable multi-agent pipelines with explicit run order and data dependencies.
- Predictable run order: The configured agent-task order determines the workflow path.
- Explicit dependencies: A downstream task can consume the result of a named upstream task.
- Focused testing: You can test each stage and handoff at a known position in the workflow.
- Clear traceability: Team and task history views show the ordered run path.
- Controlled specialization: Each agent focuses on a bounded stage and uses task-specific tools.