ChatGPT for Business: Plans, Uses and What to Keep Out of It

ChatGPT for business means putting your team on one of OpenAI’s organization plans, Business, Enterprise or Edu, instead of letting everyone use a personal account. What the business plans add is control, not a smarter model: a workspace the company owns, admin roles and single sign-on, business data excluded from model training by default, connected apps that an admin can switch on or off, shared projects, and ChatGPT Work for multi-step jobs. Enterprise adds heavier identity, audit and retention controls on top. None of it decides what people paste in. That part is yours, so this guide ends with what to keep out and a checklist to adopt.

If your question is how ChatGPT helps run a project, that is covered in ChatGPT for project management. This post is about the plan, the rules and the first uses.

ChatGPT business plans, by name

As of September 30, 2026, OpenAI sells ChatGPT in two groups. Plan names change more often than you would expect, so check the plan page before you buy.

• Individual plans: Free, Go, Plus and Pro. Each person owns their account and its settings.
• Business: a shared workspace for a team or small company. It was called ChatGPT Team until August 29, 2025, when OpenAI renamed it; the help center says the rename did not change the plan itself.
• Enterprise: everything in Business plus the identity, audit, retention and residency controls a larger or regulated company needs. You buy it through OpenAI’s sales team.
• Edu: the education counterpart of Enterprise, for schools, colleges and universities, grouped with Enterprise on OpenAI’s plan pages.

What ChatGPT Business adds over a personal account

• A workspace the company owns. People sign in with their work address, and when someone leaves, an owner or admin removes their access.
• Admin roles and sign-in. OpenAI’s Academy guide to inviting and managing your team↗ (opens in a new tab) describes Owner, Admin and Member roles, single sign-on with your identity provider, multi-factor authentication, and bulk invites by CSV.
• No training on your data by default. The Academy page on data governance and compliance↗ (opens in a new tab) says OpenAI does not use business customer content from Business or Enterprise to train its models by default, and that both plans encrypt data at rest and in transit.
• Connected apps under admin control. Apps, which OpenAI used to call connectors and now ships inside plugins, let ChatGPT search and act in Google Drive, Slack and similar tools. What they can reach is covered in ChatGPT connectors and apps.
• Shared projects. A project keeps one piece of work’s chats, files and instructions together, and can be shared with the right people. How to set one up is in ChatGPT Projects best practices.
• An agent for longer jobs. ChatGPT Work, which replaced agent mode, takes a multi-step task and returns a finished document, spreadsheet or deck. What it asks before acting is in ChatGPT agent mode and what replaced it.

One default matters more than it looks. OpenAI’s ChatGPT Work overview↗ (opens in a new tab) says plugins and their apps are on by default for Business workspaces and off by default for Enterprise and Edu. On Business, an app is available to your people from day one unless an admin turns it off, although each person still has to connect their own account in that service. Admins can also restrict an app to read-only actions where the app supports it.

What Enterprise adds

OpenAI’s plan comparison lists these for Enterprise and Edu beyond Business: SCIM provisioning from your identity provider, enterprise key management, custom roles with role-based access control, audit logs through the Compliance API, an analytics dashboard, and data retention and residency controls. The Academy notes that chats are kept until the user deletes them by default, and that Enterprise owners can instead set a workspace retention period of at least 90 days. If you must answer a regulator or an e-discovery request about what staff typed, that is an Enterprise conversation, not a Business one.

Where OpenAI’s own pages disagree

The Academy guide lists three roles and says Business members can invite and remove other members. The help center lists four roles, adding Analytics Viewer, and says admins can remove only members with a lower role. The help center also includes domain verification in Business, while the plan comparison lists it under Enterprise. Check your own workspace settings rather than trusting any single page, including this one.

Which model to use

OpenAI’s models page↗ (opens in a new tab) recommends GPT-6.1 Sol for complex work in ChatGPT Work and Codex, and a lighter model for focused, repeatable tasks. It also says GPT-5.5 retires from ChatGPT, ChatGPT Work and Codex on October 14, 2026, on every plan. If an admin set a workspace default, a scheduled task or a saved setting to GPT-5.5, change it before then. On Enterprise and Edu, GPT-6.1 Sol stays off until an administrator enables it.

Good first uses for a business

The first uses that pay off have three things in common: you supply the material, the output is a draft, and a person reads it before anyone relies on it.

• Customer email replies drafted from your policy and the ticket, sent by a person.
• Meeting notes turned into decisions, actions with owners, and open questions.
• Spreadsheet work: formulas, cleanup of an export, a month-over-month comparison you then spot-check.
• Standard documents: job postings, onboarding checklists, standard operating procedures.
• Sales preparation: a one-page brief on a prospect from its own website, with links you open.
• Contract and report summaries that list every date, dollar amount and obligation with a page number, so you know what to ask your lawyer.
• Recurring reports in ChatGPT Work, such as a Monday summary of last week’s support tickets, once the prompt is reliable.

Forty prompts for these jobs, each with what to check before you use the result, are in ChatGPT prompts for work.

What to keep out of ChatGPT, even on a business plan

“Not used for training” is not the same as “not stored”. Chats, files, project sources and data copied in from connected apps each follow their own retention rules. The plan limits what OpenAI does with your data; it does nothing about what your people choose to paste.

• Passwords, API keys, security codes and anything that unlocks an account.
• Social Security numbers, full card numbers and bank account numbers. Redact them; the task almost never needs them.
• Protected health information. OpenAI’s own rollout guide↗ (opens in a new tab) tells admins to stress not entering it unless the workspace has signed a healthcare addendum.
• Customer personal data beyond what the job needs. Summarizing a complaint does not require the customer’s home address.
• Material under an NDA or a contract that forbids sharing it with a third-party service, and attorney-client privileged documents.
• Decisions about people: hiring, firing, pay and performance ratings. ChatGPT can tidy your notes; a person decides and can explain why.

An acceptable-use checklist for ChatGPT at work

Two halves: what the admin sets up once, and what everyone agrees to. Adapt the brackets and put it where people will actually read it.

Admin, before inviting anyone
[ ] Plan chosen (Business or Enterprise) and an owner named
[ ] Single sign-on on, MFA enforced in the identity provider
[ ] Member list checked monthly; leavers removed the same day
[ ] Plugins and apps reviewed: only the ones we need switched on,
    read-only where possible, writes only after a named person agrees
[ ] Model defaults and scheduled tasks moved off retiring models
[ ] Retention decided (Enterprise) or documented (Business)

Everyone
[ ] Use only the company workspace for company work
[ ] Never paste: passwords, keys, SSNs, card or bank numbers,
    health information, privileged or NDA material
[ ] Paste only the customer data the task needs
[ ] Read and check every draft before it leaves the team:
    names, numbers, dates, links and quotes
[ ] People decide hiring, pay, prices, refunds and anything
    that cannot be undone
[ ] Connecting a new app needs [name]'s OK first
[ ] Report mistakes to [name]; no blame for reporting

Reviewed: [date]   Next review: [date + 3 months]

This checklist is ChatGPT-specific. A one-page policy for any assistant, with a 30-day rollout, is in how to use AI at work, and deciding who may connect agents to company systems, and who owns each one, is in AI agent governance for small teams. If people already use personal accounts, start with shadow AI agents.

Where the team’s work list lives

A workspace gives everyone the same assistant; it does not give the team a shared list of work. A shared project holds files and instructions, not who is doing what, and when ChatGPT changes something through an app, the other service usually records it as you.

fenbs is a small task board where people and AI assistants are members with roles. On Business, Enterprise and Edu, ChatGPT can add it in developer mode as a remote MCP server at https://fenbs.ai/api/mcp, with a browser sign-in and no key to paste, if your admin allows developer mode. Your acceptable-use rules can sit on the Decisions and rules page, recorded with the person who decided them, and every connected assistant reads the rules first. Tasks move through To Do, Next Up, In Progress and Completed, and History shows which changes ChatGPT made under its own name. fenbs has no due dates, sprints or settable assignee, and only the people you add can see a board, so it holds the work rather than your calendar.