Set up ChatGPT Business for your organization

A setup guide for nonprofits: accounts, discounts, admin settings, connected apps, shared skills, and a useful first staff workflow.

GuideGetting startedOperationsSecurity & privacyUpdated November 9, 2026

Why an organization workspace

Many organizations start with a few staff using their own AI accounts. ChatGPT Business brings access and billing into a workspace your organization manages. It also gives teams ways to share context and procedures: communications staff can reuse writing guidance, and operations staff can follow the same expense-reporting process.

This guide covers setup for an animal advocacy organization: choosing a plan, configuring access, and helping staff complete a first workflow. Our starting point is to give staff capable tools and train them to use those tools well.

Last updated November 9, 2026. Things may change. Use the official documentation when this guide differs from what you see, and adapt our suggested defaults to your organization’s work.

Setup checklist

  • Apply for the nonprofit discount, create the workspace, and buy initial seats.
  • Name a workspace owner and backup.
  • Set up sign-in and invitations, keeping personal workspaces separate.
  • Give staff written examples of allowed data and work that needs human approval.
  • Enable capabilities; connect a service if the first workflow needs it.
  • Test the first workflow with an ordinary member.
  • Hold onboarding and give staff one contact for help.

Before you start

Apply for the nonprofit discount before subscribing. Start with OpenAI for Nonprofits.

Choose the initial staff members and one task to teach first, such as drafting from a public campaign brief. That task determines which files, connections, and instructions to prepare.

Already have a departmental workspace through your org domain? You can consider expanding it if the same owners, billing, domain, and settings can serve the wider organization. You can create a new workspace if needed.

Seats, discounts, and spending

Business requires at least two paid seats. Current USD prices are:

SeatMonthly billingAnnual billing, per month
Standard$25/person$20/person
Standard with approved nonprofit discount$10/person$8/person
Premium (no nonprofit discount)$125/person$100/person

Both seat types include ChatGPT, Work, and Codex; Premium provides higher usage. Start with Standard and upgrade when limits repeatedly interrupt useful work. In Codex, sign in with ChatGPT to use this subscription; you do not need an API key. Business overview, Business FAQ.

If you already subscribe, you can also apply for the nonprofit discount using your account email.

Choose monthly billing if uptake is uncertain, or annual billing for a stable seat count. Annual prices above are monthly equivalents, billed annually.

When someone leaves, remove their membership and reduce paid seats if you will not reuse the capacity. Removing a person alone does not reduce the bill; seat reductions take effect at the next billing cycle, including annual renewal. Manage billing and seats.

Start with automatic credit reload off. If staff need extra usage, agree on a budget and follow the spend-control setup instructions. Set a monthly recharge limit before enabling reload; leaving it blank allows unlimited automatic purchases.

Create the workspace and invite staff

For a new workspace, have the intended owner sign in with their organization email and follow the Business signup steps.

In Workspace settings → Members, have the owner assign a trusted backup as Owner so billing and access do not depend only on one person. Assign the Member role to ordinary staff and the Admin role to those handling routine administration. Roles and member setup.

Give each person their own login so you can remove access individually. Have them enable multi-factor authentication using the account-security instructions.

Invite staff through Members → Invite member, using their intended work-account emails and Standard seats. In Identity & access, we suggest leaving Automatically accept join requests off during the initial rollout. Tell staff to route invitations through the owner: Business also lets members invite colleagues, and manual join-request approval does not disable that route. Invitations and discovery.

Ask staff to keep personal workspaces separate when joining. Merging personal content into Business is not required and cannot be undone back into a personal workspace. Staff may lose access to that content when they leave the organization. Migration options.

Single sign-on (SSO) is optional. If you want staff to sign in through your organization’s identity system, such as Google Workspace, have its administrator follow the Business SSO setup and test before requiring it. SSO does not automate adding or removing Business members.

Give staff clear data and review rules

Before staff start, write a clear AI policy so people know what they can and can’t do. You can use Create your AI policy if you need a place to start. Two of the biggest questions most staff have are about what software they can connect to AI and what data they can input into AI. Give examples from your work and name the person who can approve exceptions. Here’s a useful starting point:

Staff questionSuggested starting rule
What can I use now?Public campaign materials, published research, and invented practice data. Add internal materials once your organization has approved them.
What needs a decision first?Ask your policy owner before uploading or connecting private documents or personnel records. Avoid using identifiable donor records; remove identifying details if the task can use anonymized data.
What should stay out?Keep sensitive investigation materials, confidential source identities, and legal strategy out of ChatGPT Business. See What data can you put into AI? for the full guidance.
What needs human approval?Publishing, sending external messages, spending money, and changing access permissions, unless your organization has explicitly approved a defined workflow.
Where does finished work go?Save reviewed documents in your team’s usual shared drive or other work system so colleagues can use them without finding your chat.

Explain the privacy implications alongside those rules:

  • Use the Business workspace for work covered by this policy. OpenAI excludes Business data from model training by default. That does not mean the data is never stored or that connected services have the same terms.
  • If you share chats with others, share only what the recipients should see. Read the conversation and included material before sharing. Start a fresh chat with just the shareable context if the thread contains restricted information.
  • Keep personal matters in your personal workspace. Coworkers do not automatically see ordinary private chats, but an owner can export member data where that option is available. Do not treat your work account as a guarantee of privacy from your organization. Business privacy and sharing.

If your organization has specific storage or deletion requirements, establish whether Business meets them before using the affected data. Regional storage does not mean all processing stays in that region, and temporary safety copies still go to the US. Storage details, Retention and deletion.

Capability settings: our suggested defaults

We generally recommend enabling Work and Codex, paired with training, so staff can work directly on documents, spreadsheets, or code. If your team prefers a more cautious approach, you can leave local access disabled.

Open Workspace settings → Permissions & roles to configure Work access. Work and Codex have separate local and cloud controls; follow the setup instructions for your workspace.

CapabilityOur suggested starting choice
Work CloudEnable for research, documents, and analysis.
Work LocalEnable for local files and desktop workflows, with training on what it can access and change.
Codex Local and CloudEnable for development workflows, with training on code review and repository access.
Remote Control, Computer History, access tokensLeave out of the initial setup unless a specific workflow needs them.

For practice with local files, use the ChatGPT desktop app, select Work, and open a folder containing copies of the task’s files. Local execution can still involve cloud storage of messages and context, so the same data rules apply. Codex users should work on a development branch and review changes before merging or deploying.

Apps and plugins: connect the tools your team uses

Connect a service when it helps staff complete their first workflow, such as Google Drive for drafting from a campaign document. An app connects to the service; a plugin can bundle apps and reusable instructions called skills.

Business apps are enabled by default. Focus on the services staff use; disabling the entire catalog is not a prerequisite. You can choose a smaller allowed set if your organization prefers tighter controls. Plugin management.

  1. Make it available. In Admin Console → your workspace → Plugins, set the plugin to Available and enable its included app if needed. Installed installs the plugin for eligible staff automatically.
  2. Choose allowed actions. Where the app offers Actions, enable the reads and writes your workflows need and disable those your policy excludes. New actions only governs future additions. Admin setup.
  3. Connect the work account. Have staff open Settings → Plugins, select the integration, then Install plugin and Connect where prompted. Authorize the intended work account and only permissions your organization accepts. Some apps use an administrator-managed connection instead. Connection setup.

We suggest “Allow low-risk actions,” paired with training. Set it under Workspace settings → General → Settings → Plugin permissions. It reduces repeated prompts for routine work, including some writes; higher-risk actions may prompt or be denied. Workspace permission setup.

If you want confirmation before changes, choose Allow read actions. Always ask also prompts before reading app data and can help newer users learn. Individual staff may not be able to override the workspace’s setting; supervised practice on test files is another way to start.

Allow all actions may be available for an individual app. The standard workspace-wide menu does not offer it. Automatic approvals do not replace your organization’s human-review rules. Naming a folder or restriction in a prompt does not enforce access controls. App permissions.

Projects and skills: share what works

Use a Project when people need the same background material; use a skill when they need the same procedure on different inputs. Choose one to start.

For a campaign Project: create a Project, add a public campaign brief and writing guidance, then use Share to invite colleagues. Give chat access to people using the material and edit access to those maintaining it. Editors can change files and instructions and invite others; everyone in the project can see its chats and files. Create and share a Project.

Shared Projects use project-only memory, where Work is currently unavailable. Use Chat for this exercise, or start a separate Work chat with approved source material to try Work.

For an expense-drafting skill: open Plugins → Plugin Directory → Skills → Create → Create with chat. Supply your template and ask for a skill that turns a receipt into a draft entry, flags missing information, and leaves submission to a person. Install it and test with an invented receipt before sharing. Create a skill.

Before sharing a skill or plugin, remove confidential examples and unintended actions from its files. Have someone technical review any code you cannot assess.

When a skill is ready for colleagues to use, select Share from its menu, then tell recipients to find it under Skills → Shared by workspace or Shared with me and choose Install. Sharing a standalone skill does not install it for everyone. Skill sharing.

Name someone to maintain shared instructions and templates. See Manage shared AI workflows for ongoing upkeep.

Rolling it out to staff

Test with someone who has the Member role before onboarding the wider team. Have them:

  1. Sign in, select the Business workspace, and open the Project or install the skill you prepared.
  2. Produce a draft, compare its claims with the source, and save the reviewed result in your team’s usual work system.
  3. For local or connected-file work, edit a test copy, inspect the change, and undo it.

Resolve access problems or confusing instructions before teaching the workflow more widely.

Send a short onboarding note. Adapt this with your workspace name, links, and decisions:

We're using ChatGPT Business to help with [specific tasks].

Accept the invitation using [account email], then select [workspace name]. Keep your personal workspace separate if asked about merging.

Start with [workflow link and task]. You may use [approved examples]. Ask [policy owner] before using [restricted examples].

For this workflow, connect [service] using your work account, or use [approved file]. Follow [setup link].

Review the result against its sources. Get human approval before [actions your policy requires]. Save finished work in [shared location].

Join [onboarding session] to try it together. Bring access problems, mistakes, and requests for another workflow to [person/channel].

Hold a live session where staff complete the example, catch a mistake, and practice stopping a task or undoing an edit. Explain that instructions in a webpage or document can try to redirect the AI; unexpected requests to share information or change unrelated files are reasons to stop. Leave room for staff concerns about AI and questions about their work.

After a week or two, ask what helped and where people got stuck. Use that feedback to improve the instructions, adjust seats or usage limits, and choose the next workflow to share.

Keep going

Related Resources

Want helpputting this to work?

Vegan Hacktivists offers free AI and automation support to animal protection organizations — from a first conversation to a full build.

A beaver carrying a branch through the water