Why your AI receptionist should live in the EU
A phone call is one of the most personal things a business handles. People say their name, their address, why they need a doctor, which car they drive, when they will be away from home. When a machine starts answering those calls, the question "where does this go?" stops being a footnote and becomes the product.
Most voice AI products answer it with a region dropdown and a sentence in the terms. We think it takes four settings, and that you should be able to see all four without asking sales.
1. Where the servers are
NordTell runs in the EU. Not "EU on request", not "EU for enterprise". The application, the database, the recordings and the transcripts are there. That answers the GDPR's first question cleanly: your data does not leave the EU because of us.
2. Whether the call is recorded, and when
Recording is a per-agent choice with three positions: off, on, or on after the caller consents. The last one means the agent asks, and only starts the recorder if the answer is yes. A dental clinic may want off. A sales team may want on-after-consent. A hotel may not care about recordings at all and keep only the summary. The point is that it is a switch you set, per agent, and can show to an auditor.
3. How long it is kept
Retention is a number of days. Calls, recordings and transcripts expire on it. You can also delete a single call, a lead or your entire account yourself, from the dashboard, without writing to anyone. The audit log records the deletion, and then the data is gone.
4. Which model hears the call
This is the setting most products hide. A voice agent uses several AI models: one that listens, one that thinks, one that speaks. Each is made by a company, and each company has its own data terms. In NordTell, the model is a dropdown in the agent editor and the provider is printed next to it. Choosing a model is a compliance decision, and we think it should look like one. Some providers we simply do not offer for customer data.
What the agent itself does
Hosting and settings are half of it. The other half is manners. The agent says it is an AI when it greets the caller. It does not collect things it does not need. It never claims it did something it did not do, which matters more than it sounds: an agent that says "booked" when the calendar said no creates a data problem as well as a customer problem. And it always keeps a way to a human, because the right to talk to a person is not only polite, it is often the law.
Why we built it this way
Because we are from here. Scandinavian customers do not accept "trust us" as a privacy policy, and we did not want to build a product we would not use for our own phone. The GDPR wrote down some of what good behaviour looks like. The rest is craft, and craft shows in the settings screen.