Skip to main content
Help Center

Your AI, your data

Why teamspace brings no AI provider along and every tenant enters its own endpoint instead – and what that means for the GDPR and processing on your behalf.

Most providers handle AI like this: they sign a contract with a large model provider, build it in, switch it on for all customers and bill for usage. Convenient for everyone involved – and completely opaque for the customer.

teamspace does it differently: There is no teamspace AI. Every tenant enters its own endpoint. We supply the features; which model runs them and where it sits is your decision.

Why this detour?

Because it is the only way in which the decisive question gets an honest answer: Where does our data go?

If a software company brings the AI along, the answer is “to a sub-processor of our provider” – followed by a chain of contracts, assurances and standard contractual clauses that nobody checks in full. If you enter the endpoint yourself, the answer is simply: to wherever you put it.

This has four practical consequences:

No GDPR detour. Processing takes place at the provider you have a contract with – or on your own server. No additional processing on your behalf arises via teamspace, no sub-processing relationship you have to share responsibility for without being able to influence it.

You know the terms. Does the provider train on your data? How long does it store it? In which country are the servers? These are questions for your contract, not ours. And you can answer them without asking us.

You can stay local. Nothing forces you towards a provider on the internet. If you run a model in your own data centre, you enter its address – then not a single character leaves your premises. For public authorities, hospitals and anyone with strict requirements, this is often the only viable way.

You keep the choice. A better, cheaper or more suitable model? Change one field, save. You are not tied to what we negotiated two years ago.

What teamspace sends to the AI

Only what the invoked feature needs – and only when a human triggers it.

  • If you condense a ticket, that ticket goes to your endpoint.
  • If you analyse a report, that report’s data goes there.
  • If you touch no AI feature, nothing goes.

There is no background process continuously transmitting your data “for better results”. There is no pre-indexing of your tenant. Every call is a deliberate action by a user.

One further limit is set by the maximum input length field in the configuration: it caps how many characters a single call may transmit at most. See Connect an AI.

What that means for you to do

The price of this clarity is that an administrator has to decide something once:

  1. Choose a provider – or run your own model.
  2. Check the terms – training, storage, location. That is the actual work, and it has to be done whichever way you go; here it is simply visible.
  3. Enter the endpoint and access key – see Connect an AI.
  4. Enable the features your company wants to use.

After that it is invisible to all users. They only see that the features are there.

The effort is one-off, the clarity remains. If in two years someone asks where your ticket content is processed, there is a short, verifiable answer – instead of a search through someone else’s contracts.

And the model quality?

A concern that sometimes comes up: if you bring the AI yourself, you might get a worse one.

The opposite is true. You are not tied to the choice a software vendor once made; you can take whichever model fits best today – or the cheapest, if the task is simple. The model identifier is a field in the configuration.

Notes

  • No AI connected = no AI features. Without an endpoint, the features don’t appear. That is not a fault, it is the default.
  • It is one endpoint per tenant. Not per user and not per module.
  • The access key is yours. teamspace stores it for the calls, but it is your key with your provider – with your allowance and your billing.
  • This applies only to the built-in features. If you connect your own assistant via MCP, you bring that assistant’s AI yourself; then this client’s terms apply.