Introduction

Hello, I’m Mia Sato from the AI Research team.

In this article, I’ll introduce Workspace agents in ChatGPT, announced by OpenAI on April 22, 2026.
Workspace agents are a feature that allows teams to create agents inside ChatGPT and connect them with internal tools and files, making recurring team workflows easier to manage.

In this article, we’ll cover three key points:

• The main features of Workspace agents in ChatGPT
• How they differ from Projects and Custom GPTs
• How they can be used in EC operations and DX workflows

Inside GDX clients’ organizations, conversations like the following often happen:

“Which period was this number calculated for?”
“The inventory sheet has been updated, but it may not have been reflected in the promotional materials yet.”
“Then let’s summarize everything again before the meeting.”

The data exists.
But teams have to collect it again every time.
Just aligning on the basic assumptions takes time.

This is quietly exhausting.

That is why a system that allows teams to share recurring workflows as agents, while connecting them with the necessary tools and procedures, has strong potential.
In this article, we’ll look at Workspace agents from the perspective of reducing pre-meeting alignment work and unnecessary rework.

Introducing Workspace Agents

In simple terms, Workspace agents in ChatGPT are business agents designed for team use.

OpenAI describes Workspace agents as an evolution of GPTs. They are built on Codex, run in the cloud, and are designed to be used from ChatGPT or Slack. Another key feature is that they can be shared within a team and improved over time through actual use.

Their main features can be summarized as follows:

• Create agents that can be shared across a team
• Call agents from ChatGPT or Slack
• Run agents on a schedule
• Connect agents with apps such as Google Drive, Slack, and SharePoint
• Combine Skills and files
• Ask for human approval when needed
• Analyze usage

In other words, this is not a feature for asking a one-time question.
It is better suited to work that happens every week, every day, or every time a certain process is repeated.

For example:

• Collecting numbers and writing a report every Friday
• Reviewing the promotional calendar and inventory sheet to align assumptions before a meeting
• Checking product registration details from the same perspective each time

It may be easier to understand Workspace agents as a feature that turns “work that requires the same explanation every time” into a shared team workflow.

At first glance, Workspace agents may seem similar to other ChatGPT features.
So let’s briefly look at how they differ from Projects and Custom GPTs.

How Workspace Agents Differ from Projects

A Project is like a workspace where related conversations and files are collected.
For example, it can be used to store materials, past conversations, and planning notes for an EC improvement project.

Workspace agents, on the other hand, use that context and external tools to carry out specific recurring tasks.

A Project is a place to store context.
A Workspace agent is a team execution layer that uses that context to get work done.

This distinction makes the role of each feature easier to understand.

How Workspace Agents Differ from Custom GPTs

A Custom GPT is a dedicated version of ChatGPT built for a specific purpose.
It is useful when you want to fix the response style or knowledge area, such as for internal FAQ responses, product description writing, or inquiry classification.

Workspace agents are more workflow-oriented.
They can connect with apps, run on a schedule, include approval steps, and be shared with a team as part of a broader process.

A Custom GPT is like a dedicated consultation desk.
A Workspace agent is like a team member responsible for recurring operations.

Thinking about it this way makes it much easier to apply Workspace agents to real business workflows.

GDX Perspective: Where Can Workspace Agents Be Used in EC Operations?

Workspace agents are especially effective for tasks that teams repeat many times.
They seem particularly well suited to work where information is spread across multiple places and needs to be summarized in a form that helps people make decisions.

Here, we’ll focus on three use cases that are easy to try in EC operations.

Use Case 1: Creating a Weekly Advertising Report Agent

Key feature: Scheduled execution
Task: Advertising performance report creation
Goal: Reduce pre-meeting review time
Why it works: The same report format is used every week, making it easier to summarize changes and next actions

In advertising operations, teams regularly review metrics such as impressions, click-through rate, conversion rate, CPA, and sales.

However, it is not enough to simply look at the numbers.
Teams also need to understand what changed and what should be checked next.

By assigning this preparation work to an agent, the pre-meeting workload can become much lighter.

Example workflow

1. Select “Create” from Agents in ChatGPT
2. Enter that you want to create an agent for weekly advertising reports
3. Specify where the advertising reports and reference files are stored
4. Set the agent to run every Friday afternoon
5. Define the output format as “Conclusion, Evidence, Next Actions”

Example prompt

“Every Friday, review the advertising data and summarize the metrics that changed significantly compared with the previous week, the initiatives that performed well, points to watch, and items to check next week. Please output the result as a short memo that can be read in a meeting.”

Use Case 2: Creating an Agent to Organize Assumptions for Promotions, Inventory, and Price Revisions

Key feature: App integration and cross-tool workflow
Task: Preparing assumptions before promotional meetings
Goal: Separate confirmed information from unconfirmed information
Why it works: The agent can review multiple materials and create a pre-meeting memo

In EC operations, inventory sheets, promotional calendars, price revision notes, and product master data are often stored in different places.

What teams really want to discuss in meetings is not where to find the materials.
They want to decide which products to promote, which initiatives to stop, and which points require confirmation.

Workspace agents can review related materials and summarize confirmed information, inconsistencies, and unresolved items into a single memo.

Example workflow

1. Collect this week’s related materials in a shared location
2. Create a “Pre-Promotion Meeting Assumption Summary Agent” in Agent Builder
3. Specify the folders and files the agent is allowed to reference
4. Define the output format as “Confirmed Information, Unconfirmed Information, Inconsistencies, Questions to Confirm”
5. Run the agent before the meeting and share the memo

Example prompt

“Review this week’s inventory sheet, promotional calendar, and price revision notes, and create a pre-meeting assumption memo. Please divide the output into confirmed information, unconfirmed information, inconsistencies between materials, and questions that should be confirmed with the person in charge. Do not make decisions about price changes or product suspension; escalate those items for human review.”

Use Case 3: Sharing a Product Registration Pre-Check Agent with the Team

Key feature: Skills, files, and team sharing
Task: Pre-release checks for product registration, image registration, and price updates
Goal: Reduce variation in pre-release checks
Why it works: The team can use the same checking procedure

Product registration involves many items to review, including product names, descriptions, images, prices, SKUs, sales periods, notes, inventory, and publishing timing.

If each person checks these items in a different order or from a different perspective, omissions are more likely to occur.

By using Workspace agents, teams can share product registration pre-check rules.
Similar to the concept of Skills, this allows teams to reuse procedures and reduce the cost of explaining the same process every time.

Example workflow

1. Organize the product registration pre-check procedure
2. Define required items, prohibited expressions, and conditions that should stop publication
3. Create a “Product Registration Pre-Check Agent” in Agent Builder
4. Add product registration rules and checklist files
5. Share the agent within the team
6. Team members provide product materials and run the pre-release check

Example prompt

“Review the submitted product registration materials and create a pre-release checklist. Please divide the output into required items, missing information, points to confirm with the person in charge, and risks that may require stopping publication. Pay special attention to price, sales period, image replacement, and missing SKU information.”

Three Points to Check Before Adoption

1. Permissions and Approval: Decide in Advance What the Agent Can Do

Workspace agents can work with connected apps and tools.
This is convenient, but it also means teams need to decide in advance what agents are allowed to do.

The areas that require particular care are sending, editing, posting, and deleting.
Sending emails, editing spreadsheets, adding calendar events, or updating published pages can directly affect business operations.

A safe approach is to start with viewing, summarizing, and drafting only.
Then, after that, move on to posting or updating with approval.

This order is important.

2. Preview Status: Start with Drafting and Review Support

Workspace agents are currently in research preview.
Rather than placing them at the center of production workflows from the beginning, it is better to start small.

A recommended order is:

• Summarizing existing materials
• Listing items to confirm
• Drafting reports
• Creating assumption memos
• Posting or updating with approval

Do not start with automatic publishing, automatic sending, or automatic updating.
Start with areas where people can easily review and correct the output.
That is the key point.

3. Data Handling: The More Shared the Agent Is, the More Carefully Data Should Be Limited

One of the strengths of Workspace agents is that they can be shared across a team.
However, because they can be shared, data handling requires careful attention.

In EC operations, data that requires caution may include:

• Customer information
• Inquiry history
• Sales details
• Inventory numbers
• Cost information
• Purchasing terms
• Unreleased campaigns
• Business partner conditions

One point that requires particular attention is uploading entire CSV files.
It may be fast and convenient, but if personal information, cost information, or partner-related information is included, internal confirmation can become more complicated later.

At first, it is better to start with publicly available information, aggregated data, or dummy data.
Before using real CSV files, limit the file to only the necessary columns, remove personal information, and exclude cost information.

Handling this carefully can significantly lower the barrier to adoption.

Conclusion

Workspace agents in ChatGPT are a feature designed to help teams manage recurring work more smoothly.
It is useful to think of Projects as places to store context, Custom GPTs as dedicated consultation desks for specific tasks, and Workspace agents as workflow agents that connect with tools and schedules to support team operations.

In EC operations, Workspace agents seem especially easy to try for tasks such as weekly advertising reports, organizing assumptions around promotions, inventory, and price revisions, and product registration pre-checks.

There is no need to hand everything over to agents from the beginning.
Starting with drafts, assumption summaries, and lists of confirmation items is already enough to create value.

Reducing repeated explanations, aligning the points that need to be reviewed before meetings, and minimizing variation in checks may seem like small improvements.
But in day-to-day operations, these small time savings can make a meaningful difference.

References

Official reference: Introducing workspace agents in ChatGPT / OpenAI / https://openai.com/index/introducing-workspace-agents-in-chatgpt/

Official reference: ChatGPT Workspace Agents for Enterprise and Business / OpenAI Help Center / https://help.openai.com/en/articles/20001143-chatgpt-workspace-agents-for-enterprise-and-business

Reference / Expert commentary: OpenAI now lets teams make custom bots that can do work on their own / The Verge / https://www.theverge.com/ai-artificial-intelligence/917065/openai-chatgpt-workspace-agents-custom-teams-bots

Reference / Source: What Are ChatGPT Skills? Differences from GPTs and Key Use Cases for EC Operations / GDX Co., Ltd.

Part of this article was created with the support of ChatGPT and then edited and revised by the author. The content reflects the author’s personal views and does not represent the official views or statements of GDX Co., Ltd. The information is provided for reference purposes only; please refer to official announcements and primary sources for the latest and most accurate information.