visibus Blog
Copilot is not the only answer

by Lubomira Schmid · September 6, 2026 · 4 min read
If a company already runs on Microsoft 365, Copilot is the obvious reflex: one more checkbox in the license, AI right inside Word, Excel and Outlook. That is completely understandable. But "the default" is not automatically "the best fit for your business" - and that is exactly what is worth checking once, before you book a license for every employee.
What Copilot is genuinely good at
I want to be fair before we compare: Copilot has real strengths.
It lives natively in the document. No switching to the browser, no copying back and forth - the draft is written right in Outlook, the summary right in the long email thread. And on data protection, Microsoft is solidly positioned: under "Enterprise Data Protection", your inputs are not used to train the models and stay within the Microsoft 365 framework, with your permissions and retention rules. If that is exactly what you need - AI help right inside your Office documents -, Copilot is a legitimate choice.
So far, so good. Now for the points that get lost in the reflex.
Microsoft picks the models - not you
For a long time, Copilot ran exclusively on OpenAI models. Microsoft has since added to that: as of 2026, Anthropic models (Claude) are part of Copilot, in some Office apps even as the default. With that, Microsoft itself confirms what practice has long shown: no single model is the best one for every task.
Except: the selection remains Microsoft's selection. You get the providers Microsoft buys - OpenAI and Anthropic. No Gemini for reading out a stack of receipts, no Mistral, no European or local model if you want one.
That Copilot automatically decides which model answers is not the point of criticism - automatic selection is convenient, and nobody wants to run a model comparison before every question.
What matters are two questions: How large is the pool being chosen from? And can you override when you specifically want a particular model for a task? In Copilot, the pool is limited to Microsoft's purchasing, and targeted per-task selection is not available to users in most Office scenarios.
For Swiss and EU companies there is an additional catch, buried in the fine print: the Anthropic models in Copilot are excluded from Microsoft's EU Data Boundary - these requests are processed outside of it. That is exactly why they are disabled by default for companies in the EU, the EFTA (including Switzerland) and the UK; whoever wants them has to consciously enable them in the admin center. So Microsoft does offer model variety now - but for EU and Swiss companies only at the price of the processing location. The contractual commitments remain; the promise that everything is processed within the EU boundary does not.
This is vendor lock-in with a friendly face: you are tied to the provider selection and roadmap of one corporation. If a different model is the better one next month, you wait until Microsoft adds it - and then check the fine print for the conditions.
You pay per user, on Microsoft's terms
Copilot is licensed per employee, on top of the Microsoft 365 base license. Microsoft sets the prices - including limited-time introductory offers, after which the list price applies again. So you pay a flat per-user fee that another provider sets and adjusts - regardless of how intensively each employee actually uses the AI.
Data protection within the framework - but no anonymization
Here is the nuance that matters: Copilot keeps your data within the Microsoft framework, which is good. But the model sees your content in plain text - names, addresses, figures. There is no layer that removes personal data before it reaches the model. The protection is contractual and organizational, not technically placed in front.
The alternative: model freedom instead of the Microsoft default
This is exactly where visibus Chat comes in - not as a "better Copilot", but as a different approach:
- All leading models in one interface - Claude, GPT, Gemini, Mistral. Selected automatically if you like - and you can override per task at any time. When a better model arrives, it is simply there. No lock-in.
- Anonymization before the model. Personal data is detected and replaced before the request reaches the model - a technical layer, not just a promise.
- Your instance in the EU or Switzerland, history and documents stay with you.
- Usage-based instead of a flat per-user fee.
And to be fair here as well: Copilot lives inside the document, visibus Chat is its own interface. If your only need is drafting in Word and Outlook, stay with Copilot. For everyone who wants model freedom, an active data protection layer and provider independence, there is an alternative.
The point
Copilot is not a bad choice. But "we already have Microsoft" is a decision for one provider, one model family and one flat per-user fee - made without ever being noticed as a decision. It is worth making it consciously.
If you want to know what the alternative looks like for your company - write to us. We'll look at it together.