HiI wanna know best practice to map results of each subagents in 11 labs agent workflow.Please gu | ElevenLabs | Page 1

AI-generated Answer for choco_the_dev

Here’s how ElevenLabs recommends “mapping results” from subagents inside an Agent Workflow, using only documented features.

Treat “results” as structured data, not free text.
- Use Dispatch tool nodes (or tools inside a subagent) to produce JSON outputs, then write those values into Dynamic Variables via tool “assignments.” These variables become your portable state across nodes. (elevenlabs.io)
Route deterministically with Expressions whenever you can.
- Add Expression edges that read your Dynamic Variables (e.g., verified == true or score >= 0.8) to choose the next node/subagent. Prefer this over free‑form LLM evaluation for control flow. Expressions are supported in Workflows and called out in the changelog. (elevenlabs.io)
Use LLM Condition edges only for semantic checks.
- If you must branch based on unstructured dialogue (e.g., “did the user ask for billing?”), use LLM Condition edges; otherwise keep routing logic in Expressions tied to variables. (elevenlabs.io)
Capture tool success/failure explicitly.
- For API calls that gate the next step, use a Dispatch tool node and wire its Success/Failure result edges, rather than prompting the LLM to infer the outcome. (elevenlabs.io)
Persist and reuse results across subagents.
- Reference your Dynamic Variables in later subagents’ system prompts and tool parameters with {{variable_name}} so each specialized subagent can consume prior results. (elevenlabs.io)
Map escalations and handoffs with transfer nodes.
- Use Agent transfer or Transfer‑to‑number nodes for handoffs; transfer destinations can be filled from Dynamic Variables (runtime phone/SIP). (elevenlabs.io)
Validate your mapping before shipping.
- Create Tool Call Tests and Scenario Tests to assert that the workflow selects the expected branch, calls the right tool with the right parameters, and reaches the intended subagent. (elevenlabs.io)
Record final outcomes downstream.
- Enable Post‑call Webhooks to receive full transcripts, analysis, and metadata for external logging/analytics of which subagent path was taken and why. (elevenlabs.io)

Concrete pattern (documented example):

An auth subagent calls a verification tool that returns a JSON counter; an assignment stores it as auth_success_count; an Expression edge checks auth_success_count >= 3 and routes to a privileged subagent. (elevenlabs.io)

If you share the specific “results” you want

#HiI wanna know best practice to map results of each subagents in 11 labs agent workflow.Please gu