Which suits your team: ChatGPT, Claude or Microsoft Copilot? If your team works mostly in Microsoft 365 and your documents live there, start with Copilot, because it follows the permissions you have already set. If your team works outside it, choose between ChatGPT Business and Claude Team based on a test with your own tasks. The main condition for all three: decide first which company data is allowed in, and pick a business plan rather than employees' personal accounts.
That is a different question from "which model is smarter". Models change every few weeks, while a team's choice has to last longer. For more on the models themselves, read ChatGPT or Claude? With GPT-6 and Opus 5.5, that is the wrong question. This article is about what a team needs: plans, privacy and administration, and what fits which kind of work. The details were checked on 7 October 2026; plans and prices change often.
Claude vs ChatGPT vs Copilot: which fits which team?
| ChatGPT Business | Claude Team | Microsoft Copilot | |
|---|---|---|---|
| Team size | 2 to 200 paid seats | 2 to 150 seats | Microsoft 365 Business Premium: up to 300 users |
| Training on your company data | No, according to OpenAI | Not by default, according to Anthropic | No, according to Microsoft (prompts, responses and Microsoft Graph data) |
| Administration | Shared workspace, central billing and admin, roles for owner and admin | Central billing, SSO and control over connectors | Follows your existing permissions, sensitivity labels and retention policies |
| Price per user per month | $20 billed annually, $25 monthly | $20 billed annually, $25 monthly | €27.73 for Microsoft 365 Business Premium with Copilot, annual, excl. VAT |
| Fits best (our assessment) | Teams working outside Microsoft who want a broad assistant with its own workspace | Teams working a lot with documents, projects and integrations | Teams already working in Outlook, Word, Excel and Teams |
The ChatGPT and Claude prices are for the standard seat. Both also offer a Premium seat with more usage, at $100 per month billed annually. The Microsoft price is a bundle that includes the Microsoft 365 licence, so it cannot be compared with the stand-alone price of the other two. Prices can differ by country and currency; Anthropic and Microsoft quote them excluding VAT. Sources and dates are at the end.
Is your team larger than 150 or 200 seats, or do you need audit logs, SCIM or your own retention periods? Then you are looking at Claude Enterprise or ChatGPT Enterprise, and at an enterprise plan with Microsoft.
What should you look at when choosing for a team?
For an individual, what the model can do matters most. For a team, four other things count:
- Data. Which documents and customer details may go in, and where is that processed? You decide that with your privacy officer, not just with IT.
- Administration. Who adds users and removes them when someone leaves? Can you get one invoice?
- Integrations. Where does your team's work live: in Microsoft 365, in a CRM or in loose files? The best model helps little if it cannot reach your documents.
- Ownership. Where do shared instructions and examples live, and can everyone reach them? Not in someone's chat history.
One warning: "not used for training" is not compliance. It says nothing about a data processing agreement, retention or location. Read each provider's business terms.
What does ChatGPT offer teams?
ChatGPT Business, formerly ChatGPT Team, is a shared workspace for teams. According to OpenAI, every seat, Standard or Premium, includes ChatGPT, ChatGPT Work and Codex. You need at least two paid seats and can have up to 200. OpenAI says it does not train on your workspace data.
For GDPR purposes OpenAI can sign a data processing agreement with you. Note two points where OpenAI contradicts itself. Its help centre says admins and owners do not see all members' private chats by default and that a workspace cannot export data. OpenAI's general privacy page (January 2026) says instead that workspace admins can view, access, export and delete conversations. Ask for confirmation in writing before putting employee data in. And if you want your instructions and knowledge to move with you when you switch tools later, keep them yourself.
What does Claude offer teams?
Claude Team is meant for teams of 2 to 150 people. According to Anthropic it includes central billing and administration, single sign-on, control over connectors, Projects for sharing work, and no training on your content by default. Claude Code and a connection to Microsoft 365 are included too.
Notable on Anthropic's own plans page: the personal plans (Free, Pro and Max) show "Opt-out" under model training, while Team shows "none by default". That is a good reason not to let work run through personal accounts. Audit logs, SCIM and custom retention periods only come with Enterprise.
What does Microsoft Copilot offer teams?
The big difference from the other two: Copilot works in the apps your team already uses, Word, Excel, PowerPoint, Outlook and Teams. According to Microsoft, prompts and responses fall under the same contractual terms as email in Exchange and files in SharePoint, with Microsoft acting as processor. Copilot respects your existing access rights, inherits sensitivity labels and applies your retention policy. According to Microsoft, prompts, responses and Microsoft Graph data are not used to train foundation models.
Watch two things. Web searches go through Bing, with their own data-handling rules. And according to Microsoft, Anthropic's models (Claude) fall outside the EU Data Boundary. If you have agreements about where data may be stored, check which models Copilot uses for your organisation. For the cost of agents that are billed per use, read What will Microsoft Copilot cost now that agents pay per use?.
When do you choose several tools?
One tool is often enough to start with. Several side by side can make sense: Copilot for work in Microsoft 365, and ChatGPT or Claude for specialist work such as long analyses or programming. The cost is two subscriptions and an agreement on what is allowed where.
That agreement is the real risk. Using several tools without rules leaves you with knowledge in loose chats and data at different suppliers. What you keep under your own control helps with any choice: your data, your processes and your shared instructions, in one place. In an AI workplace we set that up for clients, so that switching models is a test and not a migration.
How do you test it with five colleagues?
A comparison on your own work says more than any table, this one included:
- Pick three recurring tasks, for example drafting a quote, summarising a report or answering a customer email.
- Pick five colleagues with different roles and give them the same tasks and instructions.
- Have them do each task in each tool (or in two, if you already have a preference) and note: is it correct, is it usable straight away, how much needs correcting and how long did it take.
- Ask about administration too. Could your colleague get in without fuss? Could you see who used what? Could your company data have gone in safely?
- Repeat at the next release. With your tasks and criteria on paper, that takes an hour.
Do you want to work out which tool suits your team? Book a conversation.
Sources
- OpenAI Help Center, ChatGPT Business: General FAQ, and Managing data, sharing, and privacy in ChatGPT Business, accessed 7 October 2026.
- OpenAI, Enterprise privacy at OpenAI, updated 8 January 2026, accessed 7 October 2026.
- Anthropic, Plans & pricing, accessed 7 October 2026.
- Microsoft Learn, Enterprise data protection in Microsoft Copilot and Microsoft Copilot Chat, updated 18 August 2026, accessed 7 October 2026.
- Microsoft, Plans and pricing for Microsoft 365 Copilot (Dutch market), accessed 7 October 2026.
