Price from the current rate card, never from an old quote
Delivery and installation on lines of their own
Payment terms written on every quote
A person checks the quote before it is sent
Where each one lets you write them down once, in each vendor’s own words
ClaudeTeam plan, by Anthropic
Organisation instructionsSet by Owners, applied to every conversation, and they take precedence over personal ones
ProjectsTheir own chats and knowledge base, shared to view or to edit
SkillsSpecialised knowledge and workflows, which Owners can provide to everyone
ChatGPTBusiness plan, by OpenAI
Company knowledgeAnswers from the connected sources each person can already open
ProjectsChats, files and instructions together, shared with chat or edit access
Skills, in betaReusable workflows, shared with people, groups or the whole workspace
The rules are an illustration. The places are named as each vendor names them on its own pages, read on 11 October 2026.
Once a team has been using AI on personal accounts, the owner has to choose which one the business will pay for and control. This page compares the business plans of the two, Claude Team and Enterprise from Anthropic and ChatGPT Business and Enterprise from OpenAI, on four questions an owner can check before paying: where the business’s rules and documents are written once, what an owner controls on each plan, where the data is stored for a business in the UK or Australia, and where a person approves before the assistant acts.
By The AI Adoption · Published on 11 October 2026 · Sources: the plan, help and privacy pages of Anthropic and OpenAI, read on 11 October 2026
Two business accounts, compared from their own pages
Anthropic sells Claude to businesses through two plans, Team and Enterprise, and OpenAI sells ChatGPT through Business and Enterprise. Both vendors state on their plan pages that, on these plans, they do not train their models on the business’s content by default. That commitment is one reason to move a team from personal accounts to a company account, as we explain in how to adopt AI securely, and it leaves open the other questions an owner has to settle before choosing.
Each plan includes more than the assistant itself. Claude Team includes Claude Code and Claude Cowork, and ChatGPT Business includes ChatGPT Work and Codex. This page leaves the coding tools aside and keeps to what an owner and a team use in their daily work, compared on the facts each vendor publishes, without ranking the models on quality or speed.
Where your business’s rules and documents live
A correction made in the middle of a conversation is not the same as a rule the business has written down. Memory can carry some of it forward, but the rules that should hold for everyone, such as how a quote is priced or what a reply to a customer must never promise, need a place of their own that the assistant reads every time and that only the right people can change. Both products offer such places, under different names and with a different reach.
In Claude, the Owners of a Team or Enterprise account can set organisation instructions, which apply to everyone in the organisation and to every conversation, and which take precedence over a person’s own instructions when the two conflict. Below that sit projects, each with its own conversations and a knowledge base of documents, shared with colleagues who can either view and use them or edit them, and skills, which give Claude specialised knowledge and workflows and which Owners can provide to everyone in the organisation.
In ChatGPT Business, company knowledge answers questions from the sources available to each person, such as connected apps or a Google Drive the administrator manages, and OpenAI states that ChatGPT can only access what each user is already allowed to view. Projects keep chats, files and instructions together, with instructions that apply only inside that project, and they can be shared with chat or edit access. Skills, marked as a beta on the plan page, are reusable workflows that can be shared with people, groups or the whole workspace.
The practical test is the same in both products: take one rule you have repeated to the assistant more than twice, decide whether it belongs to everyone, to one area of work or to one task, and check who can change it once it has been written down.
What an owner controls on each plan
Both first business plans come with single sign-on and central administration, and both vendors keep the finer controls for Enterprise: permissions that differ by role, accounts managed from the company’s directory, logs for compliance and the business’s own retention rules. The table sets the four plans side by side and gives, under each question, the label each vendor uses.
What comes with each plan
Question
Claude TeamAnthropic
Claude EnterpriseAnthropic
ChatGPT BusinessOpenAI
ChatGPT EnterpriseOpenAI
Your content used to train the vendor’s models“Model training” · “Content is used to train our models”
None by default
None by default
No
No
Single sign-on“Single sign-on (SSO)” · “SAML SSO”
Yes
Yes
Yes
Yes
Different permissions by role“Role-based access” · “Role-based access controls”
No
Yes
No
Yes
Accounts managed from your company directory“SCIM”, on both pages
No
Yes
No
Yes
Logs for compliance“Audit logs”, “Compliance API” · “Compliance API Logs Platform”
No
Yes
No
Yes
Your own retention rules“Custom data retention controls” · “custom data retention policies”
No
Yes
Not listed
Yes
Data stored in the UK or AustraliaAnthropic’s privacy centre · “Data residency”
Stored in the US
Stored in the US
No
Yes, among ten regions
According to the plan pages, neither first business plan includes permissions that differ by role, so on Claude Team and on ChatGPT Business what each person can reach depends on what is shared with them and on the access they already have in the connected tools. And on retention, Anthropic lists custom retention controls for Enterprise only, while OpenAI describes custom retention policies on its Enterprise plan and does not list them for Business.
Where the data is stored, for a business in the UK or Australia
OpenAI lists data residency for ChatGPT Enterprise in ten regions, the United Kingdom and Australia among them, and does not include it in ChatGPT Business.
Anthropic’s privacy centre states, for its commercial products such as Claude Enterprise, that data is stored in the United States, and that customer traffic may be routed to selected countries in the United States, Europe, Asia and Australia unless agreed otherwise. Anthropic’s regional options, which include Australia among the Asia-Pacific regions, are offered through Amazon Bedrock, Google Cloud’s Vertex AI and Microsoft Foundry. The page that lists them describes no such option for Claude’s own apps and does not mention the United Kingdom.
Where the AI processes the data, how long copies are kept and which connected tools hold copies of their own are separate questions from storage, and our guide on what you can safely share with AI goes through them. A business with a contract or a sector rule that requires storage in the UK or in Australia should check the exact plan and its terms before choosing.
Before it acts on its own
Both products now do more than answer questions: they work through a task across files and connected tools, and they can run recurring work on a schedule. What matters to an owner is the point at which a person approves before something is sent, changed or deleted, and both vendors describe it.
Where a person approves
Claude Coworkon the Team plan
You hand over a taskAcross your files and the tools you have connected
Writing to a connected toolApproved for each task, unless an administrator turns on Always allowA person approves
Automatic approvalA choice each member can make, which administrators can turn offThe business decides
Scheduled tasksIn cloud sessions, they keep running with the laptop closed
ChatGPT Workon the Business plan
You describe the workWith the context of your tools and files
Plan modeChatGPT gathers context, asks questions and writes a step-by-step plan
The plan, before the workYou suggest changes or approve it to startA person approves
Scheduled and team tasksOne-time or recurring, and shared work run with approved company tools
In Claude Cowork on Team and Enterprise, actions that write to a connected tool need approval for each task unless an administrator turns on Always allow, and members can choose to approve automatically, an option administrators can turn off. Anthropic’s announcement that Cowork and chat are becoming one Claude, rolling out to the Pro and Max plans first, describes the same principle in plain terms: by default Claude asks before taking an action, and a person can let it keep working and check in only when something needs a closer look.
In ChatGPT Work, Plan mode has ChatGPT gather context, ask questions and write a step-by-step plan, which a person can change or approve before the work starts. Work can be set up as one-time or recurring tasks, and team tasks let a team hand ChatGPT shared work, such as weekly reports and customer updates, using company tools the workspace has approved.
These settings decide when the assistant stops and asks. Which tasks should stop for a person is a decision the business makes task by task, and it is the one we settle first when we set up a workflow: the AI prepares and a person sends, or it acts on its own within written limits and passes the uncertain cases to a person, as we describe in handing a repetitive task to AI with a person’s approval.
The questions that decide it
Before a business account is chosen, these are the questions we work through with an owner. The facts above answer part of each one; the rest depends on the business.
Which tools does the team already work in, and which of them must the assistant read? Check each one on the vendor’s own list of connected apps before paying.
Which rules should hold for everyone, and who may change them? The answer decides how much you will rely on settings for the whole organisation, on projects and on skills.
Do different people need different access? Neither first business plan includes permissions by role, so either access is managed through what is shared, or the business looks at Enterprise.
Must the data be stored in a particular country? The storage options differ between the two vendors and between their plans, and the terms of the exact plan settle it.
Which tasks may the assistant carry out alone, and where must a person approve? Decide it for each task before setting any approval option.
Where we stand and how we start
We work in Claude ourselves and set it up for the businesses we work with: organisation instructions, projects and skills for the team’s work, connections to the tools the business already uses, and approval points where a person decides. When a team already works in another AI environment, we start from the account it has and the work it does there.
If you are about to choose, the first step with us is a free AI pre-audit: 45 minutes on a call about how you and your team use AI today, the rules you keep repeating to it and which work to hand over first. You leave with a clear next step, with or without us.
Every fact on this page comes from the vendors’ own pages, read on 11 October 2026. OpenAI’s pages were read in a browser. Plans and features change often; check the vendor’s page before buying.
The AI AdoptionFree AI pre-audit 
