Replacing Teams: why one tool is never enough
Looking for a chat means replacing five products at once. Know that beforehand and you are done in eight weeks; miss it and you spend week three explaining why the project is stuck.
The request nearly always arrives in the same shape: "We are looking for a European replacement for Teams." The sentence already contains the mistake the project will later fail on.
Teams is not a chat. It is a chat, a video conferencing system, a file store, a phone system and a permissions model that happen to sit behind one interface. Replace one of them and forget the other four, and you will not notice on day one — you will notice in week three, at which point it is no longer a technical question but a political one.
The five parts, taken separately
Chat is the part everyone thinks of first and the easiest to replace. Element on Matrix is the serious option here: an open protocol, self-hostable, in production use inside public authorities. The vendor sits in the United Kingdom — but because operations stay with you and the protocol is open, that weighs less than it would for a subscription product. This is exactly what Open standards are for.
Video conferencing is the part to move first, not last. There is no data migration, the risk is one afternoon, and you get proof that a European tool is not automatically worse. OpenTalk is built for it; Nextcloud Talk is fine for smaller groups but scales poorly without an additional signalling server.
File storage is its own project and does not belong in the same week. Split the estate into the cold half — archives, closed projects, contract folders — and the hot half, where several people write at once. The cold half moves over a weekend. The hot half is really a document-editing problem and gets solved later.
Telephony is the part almost everyone overlooks. If your extensions run through Teams, the migration carries a telephony project with number porting attached. That is doable, but it is not something you want to discover halfway through.
Permissions decide the effort in the end. If groups, guests and shares come out of a directory service today, you need one on the other side too — otherwise permission management becomes manual work, and manual work does not survive two years.
What is actually on the table
| Tool | Base | Replaces | Worth it for |
|---|---|---|---|
| Element / Matrix | UK | chat | open protocol, operated in-house |
| OpenTalk | DE | video meetings | when meetings are a central working tool |
| Nextcloud Talk | DE | chat, video, files | small groups, everything from one place |
| Wire | DE/CH | chat | when encryption is the first criterion |
| Stackfield | DE | chat, tasks | subscription product, no hosting of your own |
| Mattermost | US | chat | open source, in your own data centre |
And Mattermost, which has a US parent, is open source, and runs in your own data centre.
That last case is the instructive one. By pure origin logic, Mattermost comes off the list. By the question that actually matters — what does leaving cost me — it ranks above some European subscription products whose data cannot be extracted without vendor assistance. That is uncomfortable, and it is the consequence of what Vendor lock-in really describes.
Three questions that belong before the choice
First: does the old history come along? Usually not in full. That is survivable when it was said in advance, and a breach of trust when somebody discovers it on cut-over day.
Second: how do external partners get in without creating an account? If there is no official path, an unofficial one appears within two weeks — and then you are no longer discussing sovereignty, you are discussing shadow IT.
Third: who operates this if the person who set it up resigns? That question decides more projects than the other two combined. A self-hosted service with only one person who understands it is a dependency you have moved, not removed.
What gets worse
The seamless coupling is gone. Someone who today comments on a file from inside the chat window will notice that there are now two tools. Search across all content is weaker. The mobile applications are rougher. And no combination reaches the same level of telephony integration.
Say this in advance and you get patience. Leave it out and you lose it at the first incident — and patience is the scarcest resource in a project like this.
The realistic timeframe
Video conferencing: one afternoon. Chat with directory integration: four to eight weeks, depending on how cleanly groups are maintained today. File storage: a project of its own. Telephony: likewise, with lead time for porting.
And if you stop after chat and video because the rest does not pay off, that is not half a result. It is two of five layers that were not in your hands before and are now. A Sovereign workplace never arrives on a single date anyway; it arrives in stages.
Sources
Read next
- From 12 September 2026, the machine has to hand over its dataThe Data Act has applied for close to a year. Connected products placed on the market from September 2026 also carry a design obligation — and with it, the end of closed interfaces.
- European is not a status. It is a subscription.ownCloud, DRACOON, Pipedrive, Affinity, MariaDB, Silo AI — recent European software history, read as a list of ownership changes. Why asking where a vendor is headquartered is the wrong question, and what to ask instead.