Short answer. A team can be working with Claude after two weeks, provided access is sorted on day 1. Week 1 lays the foundation: plan and seats, domain and single sign-on, organisation instructions and read-only connections. Week 2 is real work: skills built on real cases, review by your own people, a short session on responsible use, and go-live. Writing into your core systems or letting Claude set prices is something we deliberately leave out of those two weeks.
What has to be ready on day 1
What slows a rollout down is rarely Claude. It is access. So we start with a list of access points, each with a name and a date on your side.
| Access point | Who on your side | Why it is needed |
|---|---|---|
| A business email address for whoever creates the Team | The future Owner | You can't create a Team with a gmail.com, yahoo.com or hotmail.com address |
| Who becomes Primary Owner | Managing director or COO | There is only one Primary Owner. The role can be a service account and uses a seat |
| The list of first users | Ops lead | Members need an email address on a domain your organisation allows |
| Access to your DNS | IT or your web agency | Domain verification uses a TXT record |
| An administrator for your identity provider | IT | For single sign-on, for example with Microsoft Entra ID, Google Workspace or Okta |
| A Global Administrator in Microsoft Entra | IT | Only if you connect Microsoft 365: they give one-time consent for the whole tenant |
| Someone who accepts the Commercial Terms | Managing director | Anthropic's data processing addendum is part of them |
On that last point: according to Anthropic, its data processing addendum (DPA), including the Standard Contractual Clauses, is automatically incorporated into the Commercial Terms. Accepting those terms means accepting the DPA. There is no separate document to sign. Do have someone read them before you use client data. A later part of this series covers GDPR.
From practice, not from a vendor page: the delay is almost always access. An administrator who is away that week, a licence still to be ordered, terms still sitting on someone's desk. That is why every access point goes into the plan with an owner and a due date on the client side, and why we go through the list before day 1.
Week 1: the foundation
Day 1. Plan and seats
You create the Team at claude.ai with a business email address and choose monthly or annual billing. Team needs at least two members. Which plan to choose and what it costs is covered in Claude Team or Claude Enterprise.
Two details matter. A pending invitation takes up a seat straight away, even if the person hasn't accepted yet. And an invitation expires after 21 days. So only invite the people who will actually take part in the first two weeks.
Day 2. Domain and single sign-on
Single sign-on is included in the Team plan. An Owner or Primary Owner sets it up. First you verify your domain: Claude shows a TXT record starting with anthropic-domain-verification-, which you add to your DNS alongside the existing records. Save that value straight away. Once the domain shows as "Pending", the admin console no longer displays it. Anthropic says ten minutes is usually enough, but it can take 24 to 48 hours.
Verifying your domain changes nothing for your users yet. Only when you enforce single sign-on do they have to log in through your identity provider. Anyone not assigned to the app there is then locked out. So we only enforce it after a test with two people.
With single sign-on you can also turn on just-in-time provisioning: anyone assigned to the app in your identity provider gets an account automatically the first time they log in. Automatic syncing through SCIM is only available on Enterprise.
Day 3. Organisation instructions, version 1
Organisation instructions are instructions Claude follows in every conversation across your organisation. Only Owners and the Primary Owner can set them. You have up to 3,000 characters, and a change can take up to an hour to apply everywhere. Where they conflict, they take precedence over a user's personal instructions.
Because they travel with every message, we keep version 1 short: which language and tone Claude uses, who your clients are, and what Claude never presents as final. Then we test in a new conversation whether Claude sticks to them. How to make good use of those 3,000 characters is the subject of the next part.
Day 4. Read-only connections
Connections to your own tools, called connectors in Claude, are switched on for the organisation by an Owner. Microsoft 365 is the one people wait on most, because it takes two people:
- An Owner enables the Microsoft 365 connector under Organization settings, Connectors.
- A Global Administrator in Microsoft Entra gives one-time consent for the whole tenant.
After that, each member connects their own account. The connector uses delegated permissions: Claude only sees what that person can already see in Microsoft 365. By default it can only read. Sending email, editing files or posting in Teams only becomes possible once you enable those write tools. In Entra you can also restrict which people may use the connector.
Microsoft itself describes consent on behalf of the whole organisation as a sensitive operation and asks you to review the requested permissions first. So plan that conversation with IT in advance, not at five o'clock on day 4.
Day 5. The baseline
Before anything changes, we write down how things work today. For each process: how often it happens, how long it takes and who does it. In the Kickstart that means three to five processes. Without a baseline, all you can say after two weeks is that it "feels better".
Week 2: real work
Days 6 and 7. Skills on real cases
A skill is a set way of working that Claude loads when it is relevant, such as how you build a quote or write a report. Owners can provision skills for the whole organisation under Organization settings, Plugins & skills, as a zip file containing a SKILL.md. A skill provisioned this way is available to everyone immediately and switched on by default. Users can turn it off themselves. On Team, skills are on by default. If a skill doesn't appear, check in the "Policy" tab that code execution and skills are both on. Skills don't work without code execution.
We build skills on the cases from the baseline, not on demo examples. A skill that works on an invented example tells you little about your real files.
Day 8. Review by your own people
The people who do the work today judge what Claude produces. Is the tone right, is anything missing, would you send this as it is? Based on that, we adjust the instructions and skills. This is the day most of the improvement happens.
Day 9. A short session on AI literacy
Article 4 of the EU AI Act asks organisations that use AI to take measures to support the AI literacy of their staff. The obligation has applied since 2 February 2025. National authorities supervise it from 2 August 2026. Since the Digital Omnibus amendment in mid-July 2026, the law no longer prescribes a specific level. According to the European Commission, no certificate is needed and an internal record of training is enough. It also says staff using a generative AI tool should be made aware of risks such as hallucination.
So we keep it practical: a short session on your own cases. What Claude does well, where it goes wrong, what should never go into it and who to ask. You keep the attendance list.
Day 10. Live, and the first figures
On day 10 the first processes run with Claude. On the Team plan, Owners and the Primary Owner see an analytics dashboard showing, among other things, weekly active members, use of skills and the number of read and write actions through connectors. We put that next to the baseline.
At handover you get a list of everything that was configured, and we hand back the temporary Owner rights.
What we deliberately don't do in two weeks
- No writing into your core systems. Accounting, CRM and ERP stay out of reach until the read-only processes have proven themselves. The Microsoft 365 write tools stay off too.
- No prices or quotes set by Claude. Claude can draft a proposal. A person decides and sends it.
- No inviting everyone at once. We start with the people in the chosen processes and widen access once it works.
- No enforcing single sign-on before the test. Otherwise you lock people out on the day they are meant to start.
In the first two weeks, Claude reads and proposes. Writing and deciding stay with your people.
The ten days on one page
| Day | What | Who on your side |
|---|---|---|
| Before day 1 | Access points with a name and a date | Managing director, IT |
| 1 | Create the Team, first seats | Owner |
| 2 | Verify the domain, test single sign-on | Owner, IT |
| 3 | Organisation instructions v1 | Owner |
| 4 | Enable connectors, read-only | Owner, Entra administrator |
| 5 | Baseline of three to five processes | Ops lead |
| 6 and 7 | Skills on real cases | The people in those processes |
| 8 | Review and adjust | The same people |
| 9 | AI literacy session | All users |
| 10 | Live, first analytics, handover | Owner |
This series
This is the second part of Claude for teams. The first part covered the choice between Claude Team and Claude Enterprise. Parts still to come: organisation instructions, Claude and Microsoft 365, skills, and GDPR with the data processing addendum. How we approach a Claude rollout in general is described on Claude.
This is the Kickstart. Two weeks, fixed scope, €4,500 excl. VAT. We set Claude up on your own accounts and put three to five processes live.
Sources (read on 3 November 2026)
- Anthropic Help Center, get started with the Team plan: https://support.claude.com/en/articles/9267247-get-started-with-the-team-plan
- Anthropic Help Center, Team plan: https://support.claude.com/en/articles/9266767-what-is-the-team-plan
- Anthropic Help Center, manage members: https://support.claude.com/en/articles/13133750-manage-members-on-team-and-enterprise-plans
- Anthropic Help Center, roles and permissions: https://support.claude.com/en/articles/9267276-roles-and-permissions
- Anthropic Help Center, set up single sign-on: https://support.claude.com/en/articles/13132885-set-up-single-sign-on-sso
- Anthropic Help Center, organisation instructions: https://support.claude.com/en/articles/14546867-set-organization-instructions
- Anthropic Help Center, set up the Microsoft 365 connector: https://support.claude.com/en/articles/12542951-set-up-the-microsoft-365-connector
- Anthropic Help Center, provision skills for your organisation: https://support.claude.com/en/articles/13119606-provision-and-manage-skills-for-your-organization
- Anthropic Help Center, use skills: https://support.claude.com/en/articles/12512180-use-skills-in-claude
- Anthropic Help Center, analytics for Team and Enterprise: https://support.claude.com/en/articles/12883420-view-usage-analytics-for-team-and-enterprise-plans
- Anthropic Privacy Center, data processing addendum (DPA): https://privacy.claude.com/en/articles/7996862-how-do-i-view-and-sign-your-data-processing-addendum-dpa
- Microsoft Learn, grant tenant-wide admin consent: https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/grant-admin-consent
- European Commission, AI literacy questions and answers: https://digital-strategy.ec.europa.eu/en/faqs/ai-literacy-questions-answers
Claude and Anthropic are trademarks of Anthropic, PBC. Microsoft 365, Microsoft Entra ID, Google Workspace and Okta are trademarks of their respective owners. NUVAI is independent of these vendors.