Rolling out Claude to a 20-person team: plans, data settings and the first 30 days

For a 20-person company, the Claude Team plan is the right starting point: 20 Standard seats cost $400 a month on annual billing, your content is not used for model training by default, and you get single sign-on, central billing and admin controls. Buy Standard seats for everyone, add Premium seats only when usage proves the need, and spend the first week on settings rather than prompts. The plan is the easy part. The rollout decides whether people use it. This guide covers what the Team plan costs, how the usage limits work, which data settings to change, when Enterprise plans make more sense, and a 30-day rollout.

Prices and plans checked on 17 November 2026.

The short answer

Plan: Claude Team, Standard seats for all 20 people, billed annually once a one-month trial on monthly billing has gone well. Cost: $20 per seat per month billed annually, or $25 if you pay monthly (Claude pricing page). That is $400 or $500 a month for 20 people, before tax. Data: no model training on your content by default on Team. One exception matters: feedback sent with the thumbs buttons. First 30 days: week one for setup, week two for a five-person pilot, week three for everyone, week four for a review of real usage. When it changes: if you need audit logs, SCIM or custom data retention, you are looking at Claude Enterprise, not Team.

What does the Claude Team plan cost for 20 people?

Anthropic's pricing page lists the Team plan for 2 to 150 seats, with two seat types you can mix. Prices are shown in US dollars, exclude tax, and Anthropic says pricing and currency vary by region, so check what your own account shows at checkout.

Setup for 20 people Billed annually Billed monthly 20 Standard seats $400 a month ($4,800 charged up front) $500 a month 17 Standard + 3 Premium seats $640 a month $800 a month

Standard seats are $20 per seat per month billed annually, or $25 billed monthly. Premium seats are $100 per seat per month billed annually, or $125 billed monthly. Both figures come from the pricing page and from Anthropic's help article What is the Team plan?. One caution: a worked example in the help article on how the Team plan bill is calculated shows a lower annual Premium figure. I am using the pricing page. Confirm the number on the checkout screen before you commit.

Annual billing is charged in full on day one. Seats you add mid-term are prorated and charged immediately, and removing a member gives no refund: the seat simply becomes free to reassign (Purchase and manage seats on Team plans).

Usage limits on Standard and Premium seats: who needs which?

Is Claude Code included? Yes. Claude Code is included with every Team seat, along with Cowork, access to all available models, shared projects, connectors and enterprise search across tools such as Slack and Microsoft 365. The difference between the seat types is usage limits, not features.

For comparison, the pricing page lists Claude Pro at $17 a month on annual billing or $20 monthly. So a Standard Team seat costs roughly what Pro would cost for each person, with more usage, central billing, a shared workspace and admin controls on top. Stacking individual paid plans on company cards saves nothing.

According to the Team plan help article, Standard seats include 1.25 times the Pro plan's per-session usage, and Premium seats include 6.25 times. There is also a weekly limit. Limits are per member. They are not pooled across the team and cannot be transferred.

In practice, most of a 20-person company will never touch the limit on a Standard seat. The people who do are usually the two or three who run long coding sessions in Claude Code or heavy file work in Cowork all day. Start everyone on Standard seats, watch the usage analytics for three weeks, then upgrade the people who keep hitting the wall. Upgrades are prorated, so waiting costs you nothing.

Usage credits and extra usage: what happens when someone hits the limit?

Unless an Owner has turned on usage credits, a member who reaches their limit waits for it to reset. With usage credits on, people keep working past the limit.

Three facts from Anthropic's article on managing usage credits to know before you enable usage credits:

Extra usage is billed at standard API rates, and on Team the credits are bought in advance. Owners can set a monthly spend limit for the whole organisation, and per-user limits where that option is available. Purchased credits expire 12 months after purchase and are not refundable.

My rule: enable usage credits with a small organisation-wide cap and auto-reload off. You learn who needs more usage without a surprise invoice. If the same person burns through credits every week, a Premium seat is the cheaper answer.

The API credits are separate

Since 7 October 2026, Team plans also include monthly API credits: $20 per Standard seat and $100 per Premium seat, pooled and capped at $500 a month (Monthly API credits for Max, Ultra and Team plans). These are for building apps and agents on the Claude Platform, not for extending chat usage. They do not roll over. For a non-technical team this is a nice extra, not a reason to buy.

Does Claude train on our data, and where does the data sit?

Here is what Anthropic's own pages say.

Model training. The pricing page lists model training as "None by default" for Team and Enterprise, and as "Opt-out" for the Free, Pro, Max and Ultra plans. Anthropic's privacy centre confirms it will not use inputs or outputs from its commercial products for training by default (Is my data used for model training?).

The exception. If a member sends feedback, for example with the thumbs up or down button, Anthropic may use that chat for training and keep it for up to five years, de-linked from your user and customer IDs. Owners can switch member feedback off under the "Rate chats" setting. For a company handling client data, I would switch it off on day one.

Retention. Conversations stay in the product until members delete them. Deleted chats are removed from Anthropic's back-end storage within 30 days. Content flagged for a usage policy violation can be kept for up to two years (How long do you store my organization's data?). Custom data retention controls are an Enterprise feature. On Team you cannot set your own retention period.

GDPR role. For Team and Enterprise, your company is the controller and Anthropic acts as the processor (Anthropic privacy centre).

Region. Greece, the UK and all 27 EU member states are on Anthropic's list of supported countries. I could not find an EU data residency option for the Team plan on Anthropic's pages. If where data is stored is a hard requirement for you, ask Anthropic in writing before you buy.

The bigger leak is usually not the Team plan at all. It is the staff already using personal accounts, where training is an opt-out setting. Moving them onto company seats closes that gap. If you need this written into policy, that is the work I do under AI governance and the EU AI Act.

Claude Team plan or Claude Enterprise?

A 20-person company can now buy either. Claude Enterprise is listed as self-serve from 20 seats at $20 per seat per month billed annually, with usage billed on top at API rates. So the Claude Enterprise cost depends on how much your people use it, while Team is a flat price with limits.

Team Enterprise Seats 2 to 150 20 minimum Price $20 or $100 per seat per month billed annually, usage included up to limits $20 per seat per month billed annually, plus usage at API rates Single sign on Yes Yes SCIM, audit logs, Compliance API No Yes Custom data retention controls No Yes Custom roles (role based access) No Yes Model training on your content None by default None by default

Source: Claude pricing page and the Enterprise plan page.

For most teams of 20, Team is enough. Choose Enterprise when a regulator, an auditor or a large client requires audit logs, automatic provisioning and removal of users, or a fixed retention period. The pricing page also lists a HIPAA-ready offering on Enterprise only. Also note one difference in control of personal accounts: Team can verify your domain and block new personal accounts on it, but only Enterprise admins can claim and migrate existing ones (Claim and migrate accounts on your domain).

What changed recently that an admin should know?

Two things from the last few weeks affect default settings on Team:

Claude Meetings (beta, from 12 November 2026). Claude can join a Google Meet call as a visible participant, take notes and save a transcript. On Team plans it is on by default, and meetings with external guests are allowed by default (Use Claude Meetings). Decide whether you want that before a client finds a note-taker in their call. Chat and Cowork become one experience on 1 December 2026. For Team organisations it turns on by default unless an Owner chooses to keep it off, and the separate experience is retired on 1 February 2027 (Anthropic's admin guide). Review the approval and network settings that come with it.

The first 30 days

Week 1: set it up properly

Name a Primary Owner and one more Owner. Do not leave the company's AI account tied to one person. Complete domain verification and set up single sign on. Anthropic's SSO guide covers Team plans, and just in time provisioning can create accounts at first login. In Organization settings, decide on feedback ("Rate chats"), Claude Meetings, the unified Chat and Cowork experience, and usage credits with a spend cap. Choose which connectors to allow. For Google Workspace, an Owner must enable them before anyone can connect, and each person signs in with their own Google account, so Claude sees only what that person can already see (Use Google Workspace connectors). Write one page of rules: what may go into Claude, what may not, and who to ask.

Week 2: pilot with five people

Pick five people from different functions and give each one real, weekly task to move into Claude. Set up shared projects with the documents each task needs. Collect what worked at the end of the week.

Week 3: bring in everyone

Invite the remaining 15. Run one hands-on session where the pilot group shows their own work, not slides. Every person leaves with one task they will do in Claude that week. This is where most rollouts succeed or fail, and it is the reason I run AI training workshops around people's real work instead of generic demos.

Week 4: review and adjust

Open the usage analytics. Who is active, who is not, who hits limits? Upgrade proven heavy users to Premium seats, and move to annual billing only if the numbers justify it.

What I would do

I would buy 20 Standard seats on monthly billing for the first month. It costs $100 more than the annual rate and buys the freedom to stop. I would turn off member feedback, turn off Claude Meetings for external calls until we have a line about it in our client terms, enable usage credits with a low cap, and connect Google Workspace or Microsoft 365 only after the one-page rules exist.

After 30 days I would switch to annual billing and give Premium seats to the people the data points to, not the people who ask loudest.

This changes in two cases. If you are in a regulated sector or your clients audit you, start the conversation about Enterprise now, because audit logs and custom retention cannot be added to Team. If your company runs on Microsoft 365 or Google Workspace and people mainly want help inside those apps, compare against the assistant built into your suite before you standardise. When I brought AI into the processes and product of a 200-person fintech, the tool mattered far less than whether each team had one clear job for it.

What to check before you decide

The seat prices shown at checkout for your country, including tax. Whether any client contract or regulator requires audit logs or a fixed retention period. Whether you need data stored in the EU, confirmed in writing by Anthropic. The default state of Claude Meetings, feedback and the unified Chat and Cowork experience in your Organization settings. How many staff already use personal Claude accounts with a company email.

If any term here is new to you or your team, my AI glossary explains it in plain language.

Not sure whether a tool rollout is the right first move for your company? Take the free AI Growth GPS and see where AI would pay back first.