Productivity

Capabilities that change how an agent works, not what it can read or write

These capabilities don't add new data sources or actions — they change the way an agent gets things done. Right now this category contains agent-level orchestration features.

Agent Delegation

Lets the agent hand off parts of a job to your user-defined sub-agents and combine their results into a single answer.

When enabled, the agent decides on its own when to delegate — typically when a job is too large for one run, when a more specialized agent is clearly better suited for part of the work, or when independent slices can run in parallel. It only ever delegates to your user-defined agents, never to base agents.

Available on: Business Specialist, System Architect, Partner (and any user-defined agents derived from them). Off by default — turn it on for the agents you want to act as "leads".

Not available on: Advisor, Sales Executive, App Builder, Site Builder, Pure AI.

Why turn it on:

• Lets a lead agent coordinate a team of focused specialists you've already built.
• Keeps each sub-agent's context clean — useful for large or multi-part jobs.
• Enables role assistants (Sales Assistant, HR Assistant, Supply Chain Assistant, …) that route work to the right specialist on your behalf.

Trade-offs:

• Higher total AI credit cost than a single end-to-end run.
• Higher latency overall, though parallel sub-tasks can shorten wall-clock time.
• Best avoided for small conversational requests — the lead just answers directly anyway.

Limits:

• Maximum 5 sub-agents per delegation step.
• One level deep — sub-agents can't themselves delegate further.
• Sub-agents inherit the lead's instance connection, enterprise context, and reasoning level.
• Built-in agents cannot be targeted; only your user-defined agents can.

For the full walkthrough — setup, what you'll see in the conversation, confirmations, cancellation, and tips — see Delegation & Sub-Agents.