Skip to main content
9 – 17 UHR +49 8031 3508270 LUITPOLDSTR. 9, 83022 ROSENHEIM
DE / EN

CompanyGPT vs. Microsoft 365 Copilot: AI in Word, Excel and Outlook — even without a Copilot licence

Tobias Jonas Tobias Jonas | | 7 min read

Anyone using Microsoft 365 gets the AI question delivered by Microsoft along with it: Copilot is deeply anchored in Word, Excel, PowerPoint and Outlook and knows your files, calendars and mails. That is a genuine advantage, and we advise nobody against it.

This page puts the limits in context — and explains why integration into Office is no longer a unique selling point.

The two approaches in brief

Microsoft Copilot

Copilot — which Microsoft now refers to without the “365” suffix — works on the Microsoft Graph, so it knows the context from your tenant, and it is deeply anchored in the Office applications. Microsoft selects the models.

When it comes to licensing, one distinction is worth making that often gets lost: the web-based Copilot Chat is, per the Microsoft documentation, included at no extra cost in eligible Microsoft 365 subscriptions. The full feature set — grounding on your work data, Researcher, Analyst, editing inside the Office applications — by contrast requires a paid add-on licence per user on top of the existing subscription.

CompanyGPT

CompanyGPT is a platform that we build inside the customer’s cloud tenant — on Azure, Google Cloud or sovereign on STACKIT. It brings the interface, roles and permissions, a knowledge base across your documents, agents and workflow automation with it. You choose the models: Claude, GPT, Gemini and open models can be run in parallel and routed by use case. Billing is through a maintenance contract, not per head.

AI directly in Word, Excel, PowerPoint and Outlook — both routes

The objection we hear most often goes: “But Copilot sits right inside Office.” That is true — and it now applies to both sides.

Via the native Office add-in from companyM365, our MCP server for Microsoft 365, CompanyGPT is available as a task pane directly in Word, Excel, PowerPoint and Outlook. Technically this runs via Office.js, the official Microsoft interface for Office add-ins. That lets CompanyGPT not merely accompany the applications but work inside them: draft texts, analyse spreadsheets, build slides, answer emails — without leaving the application.

Three points matter here:

  • No Copilot licence per user required. We install the add-in together with you for your Microsoft 365 users.
  • Requests run through your CompanyGPT in your own cloud — with your models, your agents and your access rights.
  • It does not necessarily replace Copilot. Both routes can be run in parallel.

On our CompanyGPT page, a short video shows how an Excel file is created by AI directly from within Excel.

The place of processing: a point worth knowing

With the EU Data Boundary, Microsoft has built an extensive set of commitments for European customers, and Copilot expressly falls under it. Three points are still worth knowing, because they are stated in Microsoft’s own documentation and have to be decided administratively:

Flex routing. Microsoft describes the feature as allowing, for customers in the EU and EFTA, LLM inference outside the EU Data Boundary during peak load — specifically in the USA, Canada or Australia. According to the same documentation, the feature is enabled by default for tenants created after 25 March 2026; existing customers find their setting, per Microsoft, in the Message Center. Data at rest continues to stay within the EU Data Boundary, according to Microsoft. Administrators can change the setting at any time; if it is disabled, inference takes place within the EU Data Boundary even during peak load.

Anthropic models. Per the Microsoft documentation, they are excluded from the scope of the EU Data Boundary. In fairness, this belongs alongside it: for customers in the EU and in the United Kingdom, these models are disabled by default. Anyone who enables them thereby makes a deliberate decision that, on the record, leaves the EU Data Boundary.

Web search. Microsoft points out that the Data Protection Addendum (DPA) and the EU Data Boundary do not apply to generated web search queries. Also documented: prompts and responses stay within Microsoft 365, and the search queries go to Bing without user and tenant identifiers.

None of these points makes Copilot unsuitable — they are controllable. But you have to know about them in order to control them. We have collected the details, the sources and the concrete steps to check in your own tenant here: Microsoft Copilot and data processing outside Europe.

With CompanyGPT the question is framed differently, because the platform runs in your tenant: you choose the cloud, region and model access yourself and can document the place of processing accordingly. If you need maximum sovereignty, you run the platform on STACKIT.

Comparison table

CriterionMicrosoft CopilotCompanyGPT
Operating modelMicrosoft cloud, run by the vendorIn the customer’s cloud tenant (Azure, Google Cloud or STACKIT)
BillingCopilot Chat included in eligible subscriptions; full feature set as an add-on licence per userMaintenance contract, no per-user licence; tokens at the hyperscaler’s list price
Office integrationNative in the Office applicationsNative add-in via Office.js in Word, Excel, PowerPoint and Outlook
Model choiceSet by MicrosoftFree: Claude, GPT, Gemini and open models in parallel
Knowledge baseMicrosoft Graph (files, mails, calendars in the tenant)companyRAG across your documents, including mirroring of SharePoint permissions
Place of processingMicrosoft data centres; EU Data Boundary with documented exceptions (flex routing, Anthropic models, web search)Region chosen by the customer in their own tenant
AutomationCopilot Studion8n and the open Model Context Protocol (MCP)
ExtensibilityMicrosoft ecosystemMore than 100 MCP integrations, plus n8n, Make and Zapier

When Copilot is the right choice

We consider Copilot a good decision when:

  • your company is fully standardised on Microsoft 365 and the Graph connection to calendar, Teams and mail is the main benefit,
  • you have no particular requirements regarding the place of processing,
  • you need no model variety and can live well with Microsoft’s selection,
  • and you want a solution that requires no operations of your own to be set up.

When CompanyGPT is the better choice

  • When the place of processing has to be documented — professional secrecy holders, § 203 StGB contexts, HR data.
  • When you want to choose models freely and route them by use case.
  • When a dedicated knowledge base across documents is needed that goes beyond the Graph.
  • When you want to bring automation via n8n and MCP into your business processes.
  • When per-head licence costs become an issue as the user count grows — the maintenance contract model does not scale with headcount.

Conclusion

Copilot and CompanyGPT are not pure opponents. Copilot is strong where Microsoft 365 is the entire working world. CompanyGPT is strong where model freedom, a dedicated knowledge base and a documented place of processing count — and it now brings the Office integration with it.

In many projects we see both side by side: Copilot for part of the workforce, CompanyGPT for the cases that need more control. Anyone who wants to know what fits their own starting point gets an honest assessment from us — even when it reads “stay with Copilot”.

Sources

All statements about Microsoft Copilot are taken from Microsoft’s documentation, retrieved on 23 August 2026:

  • Flex routing for Microsoft Copilot — https://learn.microsoft.com/en-us/microsoft-365/copilot/copilot-flex-routing
  • Connect to an AI subprocessor — https://learn.microsoft.com/en-us/microsoft-365/copilot/connect-to-ai-subprocessor — on the Anthropic model exception and its EU default
  • Manage public web access — https://learn.microsoft.com/en-us/microsoft-365-copilot/manage-public-web-access — on web search
  • Microsoft Copilot licensing — https://learn.microsoft.com/en-us/microsoft-365/copilot/microsoft-365-copilot-licensing
  • Microsoft Copilot privacy — https://learn.microsoft.com/en-us/microsoft-365/copilot/microsoft-365-copilot-privacy

Note on the information given: All statements about other vendors’ products are based on their publicly available documentation as of the date stated. Vendors continuously develop their products, features and terms — the vendor’s current information is always authoritative. If any statement appears outdated or inaccurate to you, drop us a line at info@innfactory.ai; we will check and correct it promptly. This comparison is not a substitute for legal or data protection advice in an individual case.

Further reading

Tobias Jonas
Written by

Tobias Jonas

Co-CEO, M.Sc.

Tobias Jonas, M.Sc. ist Mitgründer und Co-CEO der innFactory AI Consulting GmbH. Er ist ein führender Innovator im Bereich Künstliche Intelligenz und Cloud Computing. Als Co-Founder der innFactory GmbH hat er hunderte KI- und Cloud-Projekte erfolgreich geleitet und das Unternehmen als wichtigen Akteur im deutschen IT-Sektor etabliert. Dabei ist Tobias immer am Puls der Zeit: Er erkannte früh das Potenzial von KI Agenten und veranstaltete dazu eines der ersten Meetups in Deutschland. Zudem wies er bereits im ersten Monat nach Veröffentlichung auf das MCP Protokoll hin und informierte seine Follower am Gründungstag über die Agentic AI Foundation. Neben seinen Geschäftsführerrollen engagiert sich Tobias Jonas in verschiedenen Fach- und Wirtschaftsverbänden, darunter der KI Bundesverband und der Digitalausschuss der IHK München und Oberbayern, und leitet praxisorientierte KI- und Cloudprojekte an der Technischen Hochschule Rosenheim. Als Keynote Speaker teilt er seine Expertise zu KI und vermittelt komplexe technologische Konzepte verständlich.

LinkedIn