What are subagents?

1 of 6

How subagents work
The main agent delegates, reviews, and moves the job forward
Main task: A user or workflow starts with one larger goal.
Route the job: The main agent chooses which subagent should handle one part.
Do the part: The subagent performs its focused task such as research, drafting, or checking.
Return the result: The subagent sends back a draft, answer, or structured result.
Continue the work: The main agent combines the result and decides the next step.
Subagents do not replace the main agent: they help it split up the work.

Common subagent roles
Give each helper one clear job
Research: Gathers facts, files, or background context.
Drafting: Creates a first draft, reply, or summary.
Routing: Sends work to the right system, person, or queue.
Checking: Reviews format, policy, or completeness.
Data lookup: Pulls approved details from business tools.
Reporting: Turns results into a clean update or handoff.
The best subagents are specific, repeatable, and easy to evaluate.

What makes a good subagent?
Small scope, clear inputs, clear outputs
Narrow purpose: One job, not a vague mission.
Defined inputs: It knows what information it receives.
Clear output: It returns a result in a predictable format.
Simple rules: It follows instructions, limits, and success criteria.
Easy testing: You can quickly check whether it did the job well.
If a subagent cannot be explained simply, it is probably too broad.

Where to use care
More helpers mean more coordination
Too much overlap: Keep roles separate so helpers do not duplicate work.
Bad handoffs: Define what each subagent receives and returns.
Too many steps: Use only as many helpers as the task really needs.
Access control: Give each subagent only the tools and data it actually needs.
Review points: Add checks before sending important actions or messages.
Subagents help when coordination is clear, not when complexity grows by accident.

The small-business takeaway
Start simple and let specialists handle repeatable work
Subagents can make AI systems easier to organize, improve, and control.
Break one big workflow into smaller dependable jobs.
Reuse the same subagent across support, operations, or sales tasks.
Improve one helper without rewriting the whole system.
Keep permissions tighter by limiting each helper's access.
Start small: one main agent, one or two subagents, one clear business outcome.
The value: small specialists make complex AI workflows more manageable.
Start with one clear task, then add subagents only where they truly help.

Lesson done.

Up next

What is computer use?

Where does this fit your business? Get your AI plan

Swipe to see each slide.

The short answer

A subagent is a small, specialised AI helper that handles one part of a bigger job, such as research, drafting, routing or checking, and hands the result back to the main agent, which combines it and decides the next step. Use only as many as the task needs, with clear handoffs, tight access and review points before important actions.

Three things to remember

01Subagents do not replace the main agent: they help it split a larger job into focused parts.
02A good subagent has one narrow job, defined inputs, a predictable output and simple rules.
03Start with one main agent and one or two subagents, and give each only the access it needs.

Every slide, in words

01
What are subagents?
Small specialized AI helpers that handle part of a bigger job
A subagent is a focused helper that takes one part of a workflow so the main agent does not have to do everything itself.
Focused role: Each subagent is given a narrow responsibility such as research, drafting, routing, or checking.
Clear handoff: The main agent sends a task to the subagent, gets a result back, and continues the workflow.
Better teamwork: Breaking work into smaller jobs can make AI systems easier to organize, test, and improve.
Think of subagents like specialists on a team: each one handles a specific job.
02
How subagents work
The main agent delegates, reviews, and moves the job forward
Main task: A user or workflow starts with one larger goal.
Route the job: The main agent chooses which subagent should handle one part.
Do the part: The subagent performs its focused task such as research, drafting, or checking.
Return the result: The subagent sends back a draft, answer, or structured result.
Continue the work: The main agent combines the result and decides the next step.
Subagents do not replace the main agent: they help it split up the work.
03
Common subagent roles
Give each helper one clear job
Research: Gathers facts, files, or background context.
Drafting: Creates a first draft, reply, or summary.
Routing: Sends work to the right system, person, or queue.
Checking: Reviews format, policy, or completeness.
Data lookup: Pulls approved details from business tools.
Reporting: Turns results into a clean update or handoff.
The best subagents are specific, repeatable, and easy to evaluate.
04
What makes a good subagent?
Small scope, clear inputs, clear outputs
Narrow purpose: One job, not a vague mission.
Defined inputs: It knows what information it receives.
Clear output: It returns a result in a predictable format.
Simple rules: It follows instructions, limits, and success criteria.
Easy testing: You can quickly check whether it did the job well.
If a subagent cannot be explained simply, it is probably too broad.
05
Where to use care
More helpers mean more coordination
Too much overlap: Keep roles separate so helpers do not duplicate work.
Bad handoffs: Define what each subagent receives and returns.
Too many steps: Use only as many helpers as the task really needs.
Access control: Give each subagent only the tools and data it actually needs.
Review points: Add checks before sending important actions or messages.
Subagents help when coordination is clear, not when complexity grows by accident.
06
The small-business takeaway
Start simple and let specialists handle repeatable work
Subagents can make AI systems easier to organize, improve, and control.
Break one big workflow into smaller dependable jobs.
Reuse the same subagent across support, operations, or sales tasks.
Improve one helper without rewriting the whole system.
Keep permissions tighter by limiting each helper's access.
Start small: one main agent, one or two subagents, one clear business outcome.
The value: small specialists make complex AI workflows more manageable.
Start with one clear task, then add subagents only where they truly help.

More in part 6

Next step

Where does this fit your business?

Six quick questions about your business, then your AI plan: what to start with, what a person still checks, and the lessons that go with it. Or bring it straight to a free thirty-minute call.

Get your AI plan Book a free call