Claude Implementation Guide for AI Native

Claude Enterprise, at $20 per seat per month on an annual plan with a 20-seat minimum, plus usage at API rates. Enterprise rather than Team, because the four things a compliance officer will ask about are Enterprise-only: SCIM so leavers lose access the same day, role-based access control so write permissions are not organisation-wide, audit logs so activity is evidenced, and custom retention so you can state a deletion period in your record-keeping policy. The Compliance API is the fifth. Team also stops at 150 seats. As verified 1 to 3 September and re-checked 1 October 2026.

Yes. Neither the FCA nor the Central Bank of Ireland prohibits it. Data on the first-party surfaces is stored in the United States, which is lawful for a UK or Irish firm under Anthropic's data processing agreement with Standard Contractual Clauses and the UK Addendum, recorded in a DPIA and a transfer risk assessment. It bites where a client mandate or a public-sector contract requires UK or EU-only processing, or where health data is involved. For those workloads the routes are Bedrock in Ireland or London, or Vertex in the EU, at a premium and with features missing. Position as verified 1 to 3 September and re-checked 1 October 2026.

No, and a proposal that says otherwise is selling rather than advising. Copilot is native to Word, Excel, Outlook and Teams, grounded in Microsoft Graph, and governed by the Purview and Entra controls you already run. Claude sits alongside it for the people whose work is long documents, spreadsheet models, encoded procedures or code. There is also no Claude Teams app, so Teams stays Copilot's surface. Anthropic models inside Microsoft 365 Copilot are a separate question and a separate governance decision, because they process outside the EU Data Boundary.

Yes, through the Microsoft 365 connector, which is available on every plan. An Entra Global Administrator grants tenant consent once, and all permissions are delegated, so Claude sees only what the signed-in user can already see. It searches and analyses SharePoint and OneDrive documents and can create and update files there, and it searches, drafts, sends and manages Outlook mail and calendar. Write actions need an expanded consent and an admin toggle, settable per user group. One limit is worth knowing: Claude cannot attach files to Outlook drafts it creates. It can now post Teams chat and channel messages, but only with write actions enabled and each messaging tool allowed by an admin, and there is still no Claude app for Teams.

Three cases, stated straight. Patient-identifiable work cannot run in Cowork, because Cowork sits outside Anthropic's BAA, so that work goes to a build with a tenant-resident runtime. A live event stream is not a job for any seat product on any platform, so real-time monitoring is a build and the seat is the review surface. And where a client mandate or a public-sector contract requires UK or EU-only processing, the residency routes that satisfy it remove the Agent Skills, connector or Files API features most of this guide depends on, so split the work rather than pretending.

Seats are days, with one Entra consent for the Microsoft 365 connector. Getting real value takes longer and depends on one thing: whether the firm's procedures are written down. If the policies and templates are not under version control, fix the library first, or the assistant will cite the wrong document with complete confidence. The sequence in the deck runs the three in-seat examples first, because they need a champion rather than a build, then the connected work, then one build once the firm has learned how it wants to review agent output.

Yes. Complete the download form with your name, work email, company and role. The thank-you page provides the free 21-slide PDF. The longer companion document is not included in this download.