What Claude’s Team and Enterprise plans add over personal accounts, how Anthropic treats business data, what connectors, projects and skills do, and the order I would set it up in.
Claude for business means using one of Anthropic’s business plans, Team or Enterprise, instead of letting staff use personal Claude accounts for work. The business plans give you central billing, single sign-on and admin controls, and by default Anthropic does not train its models on your content. For most small and mid-sized UK firms, Team is the right place to start. The decisions that matter most come after you sign up: which of your systems Claude can reach, and how you teach it the way your business works.
Personal accounts, Team or Enterprise
Anthropic sells four personal plans, Free, Pro, Max and Ultra, alongside its business plans. The personal plans are fine for one person’s own use. For work, the gap is control: on a personal plan, each person makes their own privacy choices and the business has no central administration of the account. The business plans are:
- Team, for teams of 2 to 150. It adds central billing and administration, single sign-on, admin controls over the connectors staff can use, shared projects and organisation-wide skills. You can mix standard seats with premium seats, which carry more usage, for the people who use Claude most. Claude Code, Anthropic’s coding tool, is included, which matters if you have developers.
- Enterprise, with a 20-seat minimum. It adds the controls a larger or regulated firm is likely to need: role-based access, SCIM provisioning, audit logs, custom data retention and a compliance API. It is priced per seat plus usage at API rates, so the bill depends on how much people use it, and on which models and tasks.
Anthropic lists its current prices, excluding tax, on its pricing page, and they change, so I won’t repeat them here. What drives the cost is the number of seats, how many people need a premium seat and how much people use Claude. Start with standard seats for the people who have a real task for Claude, then move the heaviest users up once you can see who they are.
What Anthropic does with your business data
Anthropic’s Commercial Terms say the business owns the outputs and that Anthropic may not train models on customer content.
Three details are worth knowing before anyone puts client information in:
- Feedback is the exception. If someone rates a response with the thumbs up or thumbs down buttons, Anthropic says it may use that material to train its models. Ask staff not to rate chats that contain client information.
- Chats stay in the product. Anthropic keeps chats on the business plans so people can come back to them. A deleted chat leaves the person’s history straight away and Anthropic’s back-end storage within 30 days, unless it was rated with feedback, flagged under Anthropic’s Usage Policy or has to be kept longer by law, in which case Anthropic keeps it for longer. Enterprise adds custom retention periods.
- Your business is the controller. Where Anthropic processes personal data for a business customer, its privacy centre describes the customer as the controller. The ICO’s guidance on contracts says that whenever a controller uses a processor, there must be a written contract in place. Anthropic’s Commercial Terms bring in its Data Processing Addendum, and UK customers contract with Anthropic Ireland, Limited. Read both, and record Claude as a supplier that processes personal data for you.
Personal plans work differently. Each person decides in their own privacy settings whether Anthropic may use their chats to improve Claude, and if they allow it, Anthropic may keep that data in a de-identified form for up to five years. That is a reasonable choice for an individual and the wrong place for client work, so if staff already use personal Claude accounts for work, moving them onto a business plan is the first fix. This is practical guidance, not legal advice. For what should never go into any AI tool, see my guide to using AI tools with customer data.
What Claude can reach in your other tools
Connectors let Claude search, and with your permission act in, the tools your business already uses. On Team and Enterprise, an owner has to switch the Microsoft 365 connector on for the organisation before staff can connect their own accounts.
- Microsoft 365. After a Microsoft Entra Global Administrator approves it for your tenant, staff can search SharePoint, OneDrive, Outlook and Teams from Claude. The connector works with each person’s own permissions, so people reach only the data they could already open. Write tools, which let Claude send email and Teams messages, manage calendar events and create or update files, are a separate step. To keep a write tool off, set it to Blocked in your organisation’s connector settings.
- Google Workspace. Gmail, Google Calendar and Google Drive. By default, outside Auto mode (in which these actions can run without asking you), Claude asks for approval before it sends an email or shares a file, and on Team and Enterprise the owner decides whether staff may let those actions run without asking.
Working within each person’s permissions is only as safe as the permissions themselves. A SharePoint site of HR files that was opened to the whole company years ago, for convenience, becomes searchable from Claude by anyone with that access. Tidy access first, connect second, and keep write actions off until people trust the read-only version.
Projects and skills: where your know-how lives
Two features stop people explaining your business from scratch in every chat.
Projects are workspaces with their own chat history, uploaded documents and instructions. On Team and Enterprise, a project can be shared with colleagues, who can view it or edit it. One shared project holding your service descriptions, prices and house style, with an instruction to write in British English, gives everyone the same starting point.
Skills are folders of instructions, scripts and resources that Claude loads when a task needs them: producing a document in your brand style, say, or following your format for meeting notes. A simple skill is written in plain Markdown, with no code. On Team and Enterprise, an owner can provision a skill to everyone at once, and can stop staff creating their own if you would rather approve each one. Skills need code execution switched on in your organisation settings.
Two of my own products are built around skills. FundAdmin AI is a set of 62 Claude skills for private fund documents (there is more on the financial services page). I also designed and built The Counsel, a legal assistant for the laws of England and Wales with specialist skills and agents for reviewing and drafting documents, and those skills and agents run on Claude. The pattern I would use in any firm is one skill per task, with clear steps and a defined output, written so a person can read exactly what Claude has been told to do.
Turning a tidied-up workflow into a reusable skill inside your own account, loaded with your company context, is the heart of the monthly AI partner service. Everything is set up in your accounts, not mine.
The order I would set it up in
- Find out what is already in use. Ask who uses a personal AI account for work and what goes into it. Ask without blame, or you won’t get honest answers.
- Start small on Team. Standard seats for the people with a real task, set up with single sign-on, rather than a seat for everyone on day one.
- Write the rules down. Name the approved plan, what never goes in, who checks output before it reaches a client, and the rule about not rating chats that hold client data. The free AI acceptable use policy template has a space for the approved tool and plan.
- Build one shared project. Your services, prices, house style and the questions customers ask most, with instructions on tone.
- Connect one system, read-only. Check who can see what in that system, switch the connector on for the organisation, and set the write tools to Blocked.
- Turn one weekly task into a skill. Pick something repeated, such as a client update or the first draft of a proposal, write down the steps the person who does it follows, and provision the skill once it works for them.
- Review after a month. Look at who used it, for what, and what went wrong, then decide whether premium seats, more connectors or Enterprise controls are worth it.
Keep a person in the loop throughout. Anthropic’s Commercial Terms make it the customer’s job to judge whether outputs are appropriate, including where a person should review them, before they are used or shared.
Claude, ChatGPT or Copilot?
If you haven’t chosen a tool yet, start with where your work lives. Skills make Claude a good fit for firms that want to package repeatable work. If your business runs on Microsoft 365, compare it with Copilot before you buy: my comparison of ChatGPT and Copilot includes a two-week trial you can run with Claude in the mix. I don’t earn commission from vendors, Anthropic included, and AI training for teams is built around whichever tool your business has approved.
Want a second opinion on whether Claude suits your business, or help setting it up properly? Book a free 30-minute AI consultation. We will look at how your business runs and where AI could save time, with no obligation.



