Skip to main content
Orchestration defines how agents collaborate during an AI Workforce run or agentic workflow. Structured data defines when an agent should return predictable JSON instead of a normal conversational answer. In SketricGen V3, the main agent owns the user conversation. Sub agents help the main agent with bounded specialist work.

​
Main Agent

The main agent is responsible for:
  • Understanding the user’s request.
  • Planning the work.
  • Calling tools and connectors.
  • Calling sub agents when specialist help is useful.
  • Returning the final response to the user.
Write main-agent instructions that explain the task boundary, available tools, when to ask questions, and what a successful result looks like.

​
Sub Agents

Sub agents are specialist helpers. The main agent can call them for focused tasks and use their result in the final answer. Good sub-agent jobs include:
  • Classification
  • Extraction
  • Summarization
  • Research
  • Redaction
  • Validation
  • Short reusable subtasks
Sub agents cannot have their own sub agents. Keep the structure one level below the main agent so the workflow stays predictable.

​
Structured Data Outputs

AI Workforce supports structured outputs when a run needs predictable JSON. Structured inputs are no longer part of the public workflow model. The main agent talks to the user, and sub agents support that main agent behind the scenes. Use structured outputs when the result needs to be consumed by another system, rendered in a predictable UI, or reviewed consistently by your team. Good fits include:
  • Lead qualification summaries
  • Research briefs with fixed fields
  • Extraction from uploaded files
  • QA or validation reports
  • Connector-ready payloads
  • Final workflow outputs that should be copied into another system
Avoid structured outputs for normal customer-facing chat replies. If a user expects a human-readable response, keep the agent conversational. Structured outputs are most useful when:
  • The main agent needs to return a final result in a fixed shape.
  • A sub agent needs to give the main agent a predictable helper result.
  • A connector or downstream system expects specific fields.

​
Output Schema Setup

Open the agent settings and configure the output schema fields. For each field, define:
  • Name: the JSON key, such as lead_score or summary.
  • Type: string, number, boolean, or another supported schema type.
  • Description: what the field should contain.
  • Required: whether the field must be present.
Example: 
{
  "lead_name": "Maya Chen",
  "company": "Northstar Analytics",
  "intent": "Book a product demo",
  "lead_score": 82,
  "recommended_next_step": "Route to sales with the requested analytics use case."
}
Keep schemas small and focused. Re-test after changing schema fields, and check traces when a run fails to produce valid output.

​
Planning in AI Workforce

AI Workforce runs can keep a todo/planning state. This helps longer-running agents break a task into steps and continue work across tools, files, and artifacts. The plan is not a replacement for review. Check final outputs, workspace files, and traces before using results.

​
Traces

Traces show which tools and sub agents were used, what failed, how long steps took, and what artifacts were created. Use traces when a run takes the wrong path or a sub agent returns a weak result.