Skip to main content
A Task processes its bound inputs without waiting for participant responses. Use it for grading, summaries, extraction, and classification, then use Conversation or Say / Message to discuss the result aloud.

​
Configure

  1. Select a Provider and Model available to your account.
  2. Write Instructions describing the analysis and required result.
  3. Bind every input the model needs, including a previous step’s Transcript when analyzing dialogue.
  4. Define required Output fields and review task limits and Maximum time.
  5. Enable Classification preset when the result should choose a named route. Declare the allowed labels and connect each outcome.
The classification output label must match the declared enum. Use Branch for comparisons against values you already know.

​
Outcomes

​
Scenario: grade an interview

Bind an interview Conversation’s Transcript and the assessment rubric. Define outputs such as summary, strengths, and areas_to_improve. Instruct the Task to cite the candidate’s answers and mark missing evidence as unassessed. Connect complete to a feedback Conversation and bind the assessment outputs as its context. That Conversation can explain the feedback and answer follow-up questions while the Task remains responsible for the assessment. For a routing example, classify a support issue into billing or delivery and connect each label to its own support flow.