Skip to content
AI & Automation

Using Claude at work: Team and Enterprise plans and your data

What happens to your business data in Anthropic's Claude Team and Enterprise plans: training, retention, single sign-on, connectors and admin controls.

Geidi AI Team6 min read

In brief

What it is
On Claude Team and Enterprise plans your data isn't used to train Anthropic's models by default, and you get single sign-on and central admin.
Who it affects
Organisations whose staff already use Claude, some on free or personally paid accounts.
What to do
Move work use onto a business plan with single sign-on, write down which data can go in, and turn on the Microsoft 365 connector deliberately.
On this page
  1. Personal accounts versus business plans
  2. What the Team plan gives you
  3. What Enterprise adds
  4. How long does Claude keep your data?
  5. Connecting Claude to Microsoft 365
  6. Claude inside Microsoft Copilot
  7. What to do next
  8. Where Geidi fits
  9. Sources

A few of your staff have started using Claude, Anthropic's AI assistant, and they like it. Some are on free accounts, one or two pay for Pro on a personal card, and now someone has asked whether the business should buy it properly. Before anyone answers, it helps to know what actually happens to the documents and conversations people put into it.

Short version: on Anthropic's business plans, your data isn't used to train its models by default, you get single sign-on and central administration, and the Enterprise plan adds the audit and retention controls most compliance teams will ask about. On personal accounts, those decisions sit with each individual. The details below are as of September 2026.

Personal accounts versus business plans

Anthropic splits its products in two. The consumer plans are Claude Free, Pro and Max. The commercial products include Claude for Work (the Team and Enterprise plans) and the Anthropic API.

For consumer plans, Anthropic says it uses chats and coding sessions to improve its models if the person chooses to allow it, or if a conversation is flagged for a safety review. That choice is made in each person's own privacy settings. Your business can't see it, enforce it or audit it.

For commercial products, Anthropic says it won't use inputs or outputs to train its models by default. There's one exception people miss: if someone explicitly sends feedback, for example by pressing the thumbs up or thumbs down button, Anthropic may use that conversation for training. Tell staff not to rate conversations that contain client or personal information. It's a small thing, but it's the kind of detail a policy should spell out.

What the Team plan gives you

Claude Team is the plan most smaller organisations start with. It comes with Standard seats, and Premium seats for heavier users, which you can mix across the team. Anthropic lists these admin and identity features:

  • Central admin tools and billing management.
  • Single sign-on and domain capture, so people sign in with their work identity and accounts on your domain come under the organisation.
  • Just-in-time provisioning, which creates accounts when people first sign in.
  • Role-based permissions to control what each person can do.

Team also includes connectors for Microsoft 365, Google Drive, Gmail, Google Calendar, Slack and GitHub, plus an enterprise search feature across connected sources.

What Enterprise adds

Claude Enterprise includes everything in Team and adds the controls a compliance or risk team is likely to ask for:

  • Audit logs of user actions, system events and data access.
  • SCIM, so accounts follow your identity provider automatically, including when someone leaves.
  • Custom data retention for chats and projects.
  • A Compliance API that gives programmatic access to usage data, chat histories and file content.
  • Customer-managed encryption keys held in your own cloud provider.

Anthropic also notes that Enterprise pricing is usage-based: the seat fee covers access, and usage is billed separately at API rates, with spend limits admins can set. Budget for that differently from a flat per-seat plan.

How long does Claude keep your data?

On the business plans, Anthropic keeps chats in the product so people can come back to them. When someone deletes a conversation, it disappears from their history straight away and is deleted from Anthropic's back-end storage within 30 days. Incognito chats are deleted automatically within 30 days, unless they're flagged for a Usage Policy violation.

On Enterprise, an Owner or Primary Owner can set custom retention periods for chats and for projects. The minimum is 30 days. One catch: by default, projects are kept indefinitely, and a project's retention wins over chat retention for chats inside that project. If you set a short chat retention period and expect it to clear everything, check your project settings too.

Connecting Claude to Microsoft 365

Claude has a Microsoft 365 connector that lets people search SharePoint, OneDrive, Outlook and Teams from inside Claude. Anthropic says it's available on every plan, including Free and Pro, but it needs a Microsoft Entra tenant tied to a Microsoft business plan, and an administrator has to grant consent in Entra before anyone can use it.

Two things to decide before you switch it on:

  1. Who gets it. Anthropic's setup guide covers restricting access, so you can start with a pilot group rather than the whole tenant.
  2. Whether to enable write tools. These let Claude send email, manage calendar events, create and update files, and send Teams messages on a person's behalf. Useful, but it's a much bigger decision than read-only search.

The connector works through each person's own access, so the same rule applies as with Microsoft's own Copilot: if people can open files they shouldn't, the AI can find them too.

Claude inside Microsoft Copilot

There's another route to Claude that doesn't involve buying from Anthropic at all. Microsoft now offers Anthropic's models inside Microsoft Copilot and Copilot Studio, with Anthropic acting as a Microsoft subprocessor. Microsoft says its Product Terms and Data Protection Addendum apply, unless a model is labelled "Anthropic models with Data Retention". In some capabilities, such as Researcher, users can choose Claude as the model.

Microsoft makes these models available by default in some regions and off by default in the EU, EFTA and the UK. A Global Administrator or AI Administrator can turn them on or off in the Microsoft 365 admin center, under Copilot settings, "AI providers operating as Microsoft subprocessors". If you're in Microsoft 365 and haven't looked at that setting, it's worth a look this week.

What to do next

  1. Move anyone using Claude for work onto a business plan with single sign-on.
  2. Write down which data can and can't go into it. Our shadow AI policy guide has a starting point.
  3. Tell staff about the feedback exception.
  4. Decide on the Microsoft 365 connector deliberately, starting read-only with a pilot group.
  5. Check the Anthropic subprocessor setting in your Microsoft 365 admin center.

If you're still choosing between tools, our comparison of ChatGPT, Claude and Microsoft Copilot covers the trade-offs.

Where Geidi fits

Geidi AI is Geidi's newest division, offering practical AI and automation built safely on the Microsoft 365 environments, infrastructure and security Geidi already manages, with the same fixed scopes, review gates and audit trails as its managed services. If AI use and costs are spreading with no visibility or control, start with a conversation.

Sources

More on AI & Automation

One partner for the whole environment.

Tell us about your existing IT. We would be pleased to review your current system and give you an overview as to how Geidi may be able to deliver and enhance your IT operations.

Book a consultationCall 1300 855 049