# How to assign tasks to Notion Custom Agents and review their work

> Assign tasks to Notion Custom Agents, hand work between agents and teammates, review changes in page comments, and report on productivity alongside credit usage.

_Published 2026-11-12 · by Alex Sherwood (aNotioneer)_

Source: https://anotioneer.com/blog/assign-tasks-to-notion-custom-agents

---

Ever since Notion launched Custom Agents, I’ve wanted to interact with them like teammates, rather than prompting them to take on work from a chat interface. This has several benefits:

- Simple collaboration with team members and agents
- To start multiple agent runs, just assign multiple tasks to your agents
- Your tasks that are ready to be reviewed can be added to a ‘queue’ which you can work through when you’re ready
- You can measure an agent’s productivity - not just the number of credits it’s consumed - based on the tasks it’s worked on

To set this system up, in many cases you simply need to make some minor changes to the configuration of your tasks database and to add a couple of sentences to the agent’s instructions.

Once you do, you can give a Notion Custom Agent a task, have a teammate review its work, then hand it back for changes. The task page can hold the brief, the result and the conversation needed to finish it.

Here’s how to set it up, using an agent that drafts and optionally sends emails as an example.

## 1. Assign tasks to your agent

Start with a task in your Notion tasks database. Give it a clear brief, just as you would when assigning work to a teammate.

For an email draft, that might include who the email is for, what it needs to achieve and links to the information the agent should use. Specify where the draft should go too, whether that’s the task page or a connected email tool.

Select your Custom Agent in the task’s assignee property.

Check the agent’s setup before relying on that handoff. Notion supports database property-change triggers and filters, so you can configure it to act when the task is assigned to it. Give it access to the tasks database, the reference pages and any tools it needs. Linking a page in the instructions doesn’t grant access to it. [Notion’s Custom Agent documentation](https://www.notion.com/en-gb/help/custom-agents) explains triggers and access settings.

Once the agent is working on a task, you’ll see an “\[Agent Name\] is working” indicator on cards in your board views, in the top right hand corner of your task page and when it’s responding to your page comments.

If you have the page open and click on the agent’s icon in the top right hand corner of the page, you can view the details of the agent’s run in a sidebar, rather than having to open the full page chat view.

<iframe
  src="https://www.youtube-nocookie.com/embed/5m7OjQtJ8NA"
  title="Notion Custom Agent task activity demonstration"
  className="aspect-video h-auto w-full border-0"
  loading="lazy"
  allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share"
  allowFullScreen
/>

Video from [@andrewaashen](https://x.com/andrewaashen) on [X](https://x.com/andrewaashen/status/2118746027754684573?a=02OzWnkZeAVWT1CtQI8&s=20).

## 2. Tell the agent how to update the task

Your agent’s instructions will already explain how to do the work: how to write the email, which sources to use and what a good draft looks like.

Add options to your task status property for blocked tasks that need intervention and tasks that need to be reviewed. Then add this section alongside those instructions:

> **Task status updates**
>
> 1. If there are any issues, set the Status to *Blocked* and describe the issue, as well as a proposed solution, in a comment on the page.
> 2. Otherwise once the email has been drafted, set the status to *Ready For Review*.

*Replace the status options in italics with your status names.*

These two instructions give the agent a clear way to finish its part of the task.

If it can complete the draft, it puts the task in your review queue. If it can’t, it flags what needs your attention and suggests a way forward.

For example, if the brief refers to a proposal the agent can’t access, a useful comment would be:

> I can’t access the proposal needed to draft this email. Please give me access or add the relevant scope and pricing to this task, then ask me to continue.

For other types of work, replace “once the email has been drafted” with the relevant outcome.

As an alternative to the second instruction, if you trust the agent to complete the task reliably, you could ask it to update the task’s status to mark it as complete.

## 3. Hand the task to a teammate, then back to the agent

Once a task is ready for review, if you want the agent to reassign it then add a separate instruction specifying how to identify the reviewer. Usually this will either be a specific team member or someone who’s tagged in another Person property on the task.

Make sure that team members who will need to interact with the agent can do so by sharing the agent with them.

The reviewer can then check the draft and decide what happens next.

If it’s ready, they can take the next step, such as sending the email, and mark the task complete when the full job is finished. Or they can simply comment on the page, tagging the agent with a message like “approved” and the agent can send the email and update the task’s status to mark it as done.

## 4. Review changes in the page comments

Use the task’s comments to explain revisions and resolve blockers. Notion supports mentioning agents in comments and comment-added triggers; configure the agent to respond to the review requests you want it to handle. [See Notion’s trigger documentation](https://www.notion.com/en-gb/help/custom-agents).

Be specific with feedback. For example:

> Please shorten the introduction, remove the discount and add a question asking whether Friday works for a call.

For a blocked task, provide the missing information or resolve the access issue, then ask the agent to continue.

## 5. Report on the work your agents complete

Once agents work through your tasks database, you can build reports around their contribution.

Start by setting up views showing tasks awaiting review, blocked tasks and completed tasks to enable you to manage the workflow. Then build summary reports in a dedicated dashboard page to monitor the agent’s performance.

If the name of your tasks will be different for each task (it’s usually best to assume that will be the case), then adding a task Type select (dropdown) property to your database allows you to identify every instance of certain tasks that agents are working on reliably.

The current assignee may change as work moves between the agent and your teammates too. If the assignee won’t be the agent when the task is completed, keep a separate property recording which agent did the work, so that contribution remains visible after reassignment. Record completion dates and revision counts if you want reports over time.

Treat those as fields you deliberately maintain through instructions or automations, rather than assuming the current status tells the whole history.

Review the results alongside credit usage. Twenty drafts that need extensive rewriting mean something different from twenty approved drafts that only need a quick check.

Start with one repeatable task, use the comments to improve the handoff, and build your reporting around the outcomes that matter to your team.