---
title: "Managed Agents: Subagents"
description: "Learn how Aesthetix CRM Managed Agents delegate work to specialist subagents, how to create and manage them, monitor their activity, and what permissions and safeguards apply."
url: https://help.aesthetixcrm.com/en/articles/2223-managed-agents-subagents
collection: "Agent Studio"
updated: 2026-11-20
language: en
---

# Managed Agents: Subagents

Learn how Aesthetix CRM Managed Agents delegate work to specialist subagents, how to create and manage them, monitor their activity, and what permissions and safeguards apply.

Subagents let a Managed Agent act as a coordinator that hands focused tasks to specialist helpers, runs independent work at the same time, and reviews the results before it answers. You build and manage them from **AI Agents > Agent Studio > Managed Agents** in Aesthetix CRM, by describing what you want in plain language.

> **Protected health information (PHI):** AI features in Aesthetix CRM are not approved for processing PHI. Do not put PHI into agent instructions, prompts, files, memory or knowledge, and do not ask a subagent team to research or update records that contain patient health details. Use subagents for business work such as lead pipeline review, competitor research and marketing preparation.

## What Subagents Are and Benefits

Subagents are specialized AI helpers that work under a main Managed Agent. Each gets a focused task, the relevant apps and tools, and only the information it needs. A Managed Agent "team" has three parts:

-   **Main agent (coordinator):** plans the work, delegates, reviews results and responds.
    
-   **Subagents (specialists):** carry out assigned tasks using the relevant tools and app connections.
    
-   **Task reports:** results returned to the coordinator for review and consolidation.
    

Example: a sales-pipeline agent for your practice can hand "research this referral partner" and "review open consult opportunities" to two different specialists, then combine their findings into one recommended next step.

Subagents work inside Managed Agents. Flow agents, which are built on a node-based canvas, are a separate approach.

Benefits:

-   **Complex job automation:** multi-step jobs are completed by coordinated specialists instead of one long run.
    
-   **Faster execution:** independent tasks can run in parallel.
    
-   **Lower processing per task:** each specialist works from a short, task-specific brief instead of the whole conversation.
    
-   **Better checking:** the coordinator reviews specialist results before replying.
    
-   **Specialized execution:** research, CRM analysis, opportunity updates and other authorized operations.
    
-   **Transparency:** you can inspect specialist tasks, reports, sources, statuses and changes.
    
-   **Controlled permissions:** each specialist is limited to the apps, records and actions its assignment needs.
    

Speed, cost and accuracy gains depend on the workload, the task complexity, the tools involved and how dependent the tasks are on each other. Do not assume subagents will always make a run cheaper or faster.

## How Delegation Works

1.  **Plan:** the main agent analyzes the request and identifies the tasks.
    
2.  **Delegate:** specialists receive focused briefs, relevant context and appropriate tool access.
    
3.  **Execute:** independent tasks run in parallel. Dependent tasks wait for the results they need.
    
4.  **Review:** the main agent checks the specialist reports and can request more work.
    
5.  **Respond:** the coordinator consolidates the findings or completes the authorized actions.
    

Example: ask the agent to compare three local wellness studios for a partnership. It can send the research for each studio to a specialist at the same time, review what comes back, request follow-up research if something is thin, and return one comparison.

![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-fnSfBtH45HLFywbyCq7aeXudzM5U8ZMNebcYMwyP0Rcb0JIJ5kxX3z33XTi5Aqh90DgFqbCh2Oa.png?v=1&k=us1&s=api&e=1795348800&sig=qhZ3yLeghtw52gsGiKRuuP_nD9wdokAam68UsVm6kBQ)

Not every request needs subagents. The builder decides whether delegation helps; simple requests may run without any specialists.

## Create and Manage Subagents

Follow these steps to set up a team.

1.  **Open Managed Agents.** Go to **AI Agents > Agent Studio > Managed Agents**. Open an existing agent or click **\+ Create Agent**.
    
2.  **Describe the job.** Enter a natural-language description of the outcome, the tasks and the apps. Example: "Build a pipeline team that researches new consult leads, reviews open opportunities, identifies stalled consult follow-ups and recommends next steps." The builder generates the configuration and decides whether specialists are needed. Another example prompt: "Build me a team of sales agents that can automate my sales pipeline." The builder might create a Sales Pipeline Team with specialists such as a Lead Researcher and a Pipeline Steward.
    
3.  **Review the team.** Read the generated description, find the **Subagents** row, inspect each specialist's role and responsibilities, and confirm the team matches your intended workflow.
    
    ![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-UyzUo1UP56FKfGWRzox5aUIwGaCXFPVptm8vQagKckahZ4bAiaqjdanU3UoUN0cYOWgyuoPTzqM.png?v=1&k=us1&s=api&e=1795348800&sig=56f3g7qDqudPTjEl-pXfaZ7-AKKZNklcCmPisqN__7Q)
4.  **Refine the team.** Ask the builder in plain language to add, remove or adjust specialists. Examples: "Split this into research and CRM update subagents," "Add a subagent that checks the calendar," "Create a specialist that reviews opportunity records," "Make the research specialist read-only," "Add a specialist that checks existing opportunities before recommending CRM updates." Check the Subagents row after every change.
    
5.  **Review apps and permissions.** Verify the main agent's connected apps and tools, review its instructions and enabled Custom Skills, confirm which tasks are read-only and which may update records, and restrict sensitive actions to what is truly needed. Subagents use the main agent's app connections within their task limits.
    
6.  **Test the agent.** Click **Test Agent**, submit a realistic multi-step request, review the response and the delegated tasks, click **View subagent activity**, and verify the specialist results, their scope and the final answer. Example test: "Review this lead's CRM details, open opportunities and recent notes, and recommend the next action."
    
7.  **Publish and run.** Review configuration and permissions, add any supported triggers, click **Publish**, run the agent through chat or another supported execution method, and monitor the activity. Subagents support chat, applicable triggers, waits, follow-ups and API runs, although individual tools can have their own execution limits.
    
8.  **Review and improve.** After a run, inspect Subagent Activity and Agent Logs, check task results and record changes, and refine instructions or specialist responsibilities. Test significant changes before using the agent in automated workflows.
    

If the agent should book appointments, remember that AI appointment booking works only with native Aesthetix CRM calendars, not EMR calendars.

## Subagents vs Custom Skills

<table style="min-width: 50px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><th colspan="1" rowspan="1"><p><strong>Capability</strong></p></th><th colspan="1" rowspan="1"><p><strong>Purpose</strong></p></th></tr><tr><td colspan="1" rowspan="1"><p>Subagents</p></td><td colspan="1" rowspan="1"><p>Specialized workers that carry out delegated tasks</p></td></tr><tr><td colspan="1" rowspan="1"><p>Custom Skills</p></td><td colspan="1" rowspan="1"><p>Reusable instructions that define how a task should be performed</p></td></tr><tr><td colspan="1" rowspan="1"><p>Apps and tools</p></td><td colspan="1" rowspan="1"><p>Connected capabilities used to retrieve information or perform authorized actions</p></td></tr></tbody></table>

A Lead Researcher subagent can follow a Custom Skill that holds your lead qualification rules. See "Managed Agents: Custom Skills, Files and Memory".

## Monitoring Subagent Activity

Subagent Activity shows how the main agent delegated work, which specialist completed each task, and how the final answer was produced.

### View activity

Click **View subagent activity** beneath a response that contains delegated work. The panel shows:

![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-5KQbOPaaVUnDwqp5fRWlLkCpzc6qnugaVAtCxee6XX0APNV4RHFo3N9Bgkoa77inipwNsDjLhxH.png?v=1&k=us1&s=api&e=1795348800&sig=D7q3NnPb-7jwdBySVP3H4BGkjVUIC2hPn3qFRRP9jvI)

-   **Team Flow:** coordinator, specialist tasks, review and final response.
    
-   **Task status and duration:** execution status and completion time.
    
-   **What changed:** a summary of actions or changes during the run.
    
-   **Specialist details:** the assigned subagents and their responsibilities.
    
-   **Task reports:** results, sources, execution steps and available briefs.
    
-   **Activity timeline:** delegation, reviews and replies in order.
    

### Review one task

1.  Open the Managed Agent conversation.
    
2.  Click **View subagent activity**.
    
3.  Select a specialist task from the flow or timeline.
    
4.  Review its result, steps, sources, brief and full report where available.
    
    ![](https://files.us.gleap.ai/pf/g/640067b236b432d4bc3c20bb-WvYXTxmF24DnGezHHgORXwu20elNQ1jZhNFEzAR2klCxtMBaIGPacNXMcmNKqc1j2X5obN8rTnb.png?v=1&k=us1&s=api&e=1795348800&sig=K0qehYbyyzvSgmEWjw0iKGf-xkcTZTsa5MX1iuQBMuE)
5.  Compare its findings with the coordinator's final response.
    

This is useful when you need to validate competitor research, CRM updates or any multi-specialist job.

### Trigger and API runs

For automated runs, use the agent's **Activity** tab, and use **Agent Logs** (AI Agents > Agent Logs) for step-level detail and troubleshooting. See "Managed Agents: Testing, Publishing, Unpublishing and the Activity Feed" and "Agent Studio: Agent Logs, Metrics, Monitors and AI Usage".

## Permissions, Safeguards, Costs and Usage

### Permissions and safeguards

-   **Read-only tasks:** research and lookup tasks cannot change records.
    
-   **Scoped updates:** update tasks can change only the records authorized in their brief and permissions.
    
-   **Shared connections:** subagents use the main agent's existing app connections, so they do not need their own.
    
-   **Restricted tools:** Computer (browser) operations and Custom API calls stay with the main agent. Subagents cannot use them directly.
    
-   **Message review:** the main agent checks any customer message a specialist drafts before it is sent.
    
-   **Loop protection:** a subagent that keeps calling the same tool without making progress can be stopped.
    

For example, a research specialist can analyze a contact without changing the CRM record, while a separate update specialist changes only the opportunity records it is authorized for.

Tool availability depends on the type of run. Computer currently supports chat and manual runs, not triggers. That limit applies to Computer only; it does not restrict all subagent delegation to manual runs.

### Costs and usage

Subagent work is billed as part of the parent Managed Agent run. There is no separate subagent billing category. Because specialists work from shorter briefs, runs can use less processing, but actual charges depend on your account's AI usage and the work done. See "Agent Studio: Agent Logs, Metrics, Monitors and AI Usage" to watch usage.

## Frequently Asked Questions

**Do all Managed Agent tasks need subagents?**

No. The builder decides when delegation helps. Simple requests may run without specialists.

**Can multiple subagents work at the same time?**

Yes. Independent tasks run in parallel. Tasks that depend on earlier results wait for them.

**Can subagents update CRM records?**

Yes, when authorized. Read-only tasks cannot change records, and update tasks are limited to the scope in their brief and permissions.

**Do subagents need their own app connections?**

No. They use the main agent's connections with task-specific restrictions.

**Can a subagent use Computer or Custom API calls directly?**

No. Those capabilities remain with the main agent.

**Can subagents run from triggers?**

Yes, through supported trigger types. Some tools have extra limits. For example, Computer does not run during trigger-based executions.

**Do subagents always cut cost and time?**

No. Results depend on workload, task complexity, tools and task dependencies.

**What happens if a subagent keeps repeating a tool call?**

Loop protection can stop a subagent that is not making progress. Check Subagent Activity or Agent Logs to investigate.

**Can I inspect what each subagent did?**

Yes. Subagent Activity shows specialist tasks, reports, sources and execution details.

**Is it safe to ask subagents to review patient records?**

No. AI features are not approved for PHI. Keep patient health information out of agent instructions, prompts, files, memory and knowledge, and use subagents for business tasks only.
