GDPR and the AI Act

How to Choose an EU-Hosted AI Sales Assistant: What GDPR and the AI Act Require

BOTrento Team8 min read

To choose an EU-hosted AI sales assistant, check four things: where your data is stored, where the language model processes your text, who stays in control of what gets sent, and what the vendor will confirm in writing. "EU-hosted" on its own only answers the first.

Neither GDPR nor the AI Act is a badge a tool simply has. They describe duties that are shared between you and the vendor. The checklist below helps you cut through marketing claims before AI touches your enquiries and quotes. It is general guidance, not legal advice.

What "EU-hosted" actually tells you

An EU server tells you where data is stored: your emails, quotes, contacts and notes at rest. That matters, but it is not the whole story.

Most AI sales assistants also send text to a language model through an API run by a commercial provider. That is a separate step, and it may happen somewhere other than the storage server. So a tool can honestly store data in the EU and still call an external model. Treat a blanket claim that everything stays inside the EU as something to question, not a fact to accept.

We cover this split in more depth in where your data is when you use AI and in what EU server hosting covers and what it doesn't.

What GDPR means for a sales assistant

Your enquiries contain personal data: names, work emails, phone numbers, sometimes details about a person's role or situation. If an AI tool handles them, you are normally the controller and the vendor acts as a processor on your behalf. In practice that means:

  • a data processing agreement is in place;
  • the vendor tells you which sub-processors it uses, including the model provider;
  • only the text needed for the task is processed;
  • you can export and delete your data;
  • you know whether the provider uses your data to train models.

For a shorter version, see our answer on whether an AI sales tool is GDPR-compliant. If you send outbound emails to prospects, the B2B cold email and GDPR checklist covers the opt-out side. For specific cases, such as sensitive categories of data or international transfers, ask a lawyer.

What the AI Act means in practice

The AI Act sorts AI systems by risk and sets duties mainly for providers and, to a lesser degree, for those who deploy them. A sales assistant that drafts replies and reminds you about follow-ups is generally a lower-risk use than, say, systems that make decisions about people's rights. Still, you should know your role and ask sensible questions:

  • Is it clear to you, and to the people you write to, when AI is involved?
  • Can a human review and change every output?
  • Does the vendor document what the system does and what it does not do?
  • Does anyone on your team understand the tool well enough to use it responsibly?

Which rules apply to your situation depends on the details, so confirm with a lawyer. Our AI Act checklist for sellers goes further on roles and vendor documentation.

Evaluation checklist: 10 questions to ask a vendor

Send these in writing and keep the answers. Vague replies are an answer too.

  1. Where is my data stored? Country and hosting provider, including the database.
  2. Where does the language model process my text? Which provider, under whose account, and whether it is an external API call.
  3. Is my data used for model training? By the vendor and by the model provider.
  4. What exactly is sent to the model? The full mailbox, or only the text needed for the task?
  5. Can anything be sent without a human approving it? You want the answer to be no.
  6. Can I use my own API key? And which providers does that support today?
  7. What happens if AI usage runs high? Is there a limit, a warning, or an automatic extra charge?
  8. How do I export and delete my data? Including what happens when the contract ends.
  9. Who are the sub-processors? Ask for the list.
  10. How are incidents handled? Monitoring and response times, stated plainly.

Be wary of phrases such as "fully compliant", "100% secure" or "nothing leaves the EU" without detail behind them. If you work in a regulated or confidential area, also look at how the tool treats enquiries under NDA.

Keep a human in the loop

Whatever the vendor promises, the most practical safeguard is a process where AI prepares and a person decides. Drafts of replies, quotes and follow-ups should wait for approval. Nothing should go to a customer on its own.

A simple approval routine

  1. AI drafts the reply from the email text.
  2. The owner or salesperson reads it and checks prices, dates and promises.
  3. They edit or reject it, then send it from their own mailbox.

This also works against the usual failure of generic chat tools: someone pastes a whole thread into a chat window, nobody checks the output, and nobody records what was shared.

How BOTrento handles this

BOTrento Chief of Staff is a managed process for enquiries and quotes, not a self-service chat window. Here is what that means for the questions above:

  • Storage: data is stored on servers in Germany (EU), with the application database in Ireland.
  • Model processing: the language model processes only the text needed for the task, through a commercial provider's API. By default that is our account; optionally it is your own key. Commercial API providers do not use the data for training. An EU server that calls an external AI API is not a closed computation, and we do not claim it is.
  • Human approval: nothing is sent without a person approving it. BOTrento does not send anything itself.
  • AI included: AI use is part of the plan, with a monthly usage limit. You get a warning at 80% and new work pauses at 100% until the next month. There is no automatic extra charge. See the pricing page for plans.
  • Own API key: optional, currently Anthropic (Claude) only. Keys are encrypted, only the last 4 characters are visible, and you can remove them any time. Details in the API key help page.
  • Exit: data and outputs can be exported, and access and copies are removed by an agreed date.

It reads your mailbox or uploaded data, shows a daily overview, drafts replies, quotes and follow-ups, and catches promises like "I'll get back to you by Friday". It does not replace your CRM or accounting, and it does not read attachments. You can see how it works in the process demo or read what BOTrento is.

We do not promise any particular business outcome. That is why you can test it on your own data first.

Test on your own data before you commit

A compliance checklist tells you whether a tool is acceptable. Only a trial tells you whether it fits how you work. Use real but representative enquiries, and check:

  • Are the drafts accurate on your prices, terms and tone?
  • Does the daily overview surface the right enquiries?
  • Are promised follow-ups caught?
  • Do your colleagues find the approval step easy?

The BOTrento 14-day free trial needs no card and no API key of your own. The 14 days start with your first successful data upload, and you have 7 days to prepare. Afterwards you keep read-only access and export for 30 days.

FAQ

Is an EU-hosted AI tool automatically GDPR compliant?

No. EU hosting covers where data is stored. GDPR also looks at the agreement with the vendor, sub-processors, what is sent to the model, retention and your ability to export and delete data.

Does the AI Act ban AI sales assistants?

No. A tool that drafts replies and reminders is generally a lower-risk use. You should still ask about transparency, human oversight and documentation, and check your own situation with a lawyer.

Is my text processed outside the EU when I use an AI sales assistant?

It depends on the setup. Storage may be in the EU, but if the tool calls an external model API, the text needed for the task is processed by that provider. Ask the vendor to confirm in writing where and by whom.

Can an AI sales assistant send emails on its own?

Some can, which is a risk. It is safer to choose a tool where a human approves every draft. In BOTrento nothing is sent without approval, and it does not send anything itself.

Should I use my own AI API key?

It is optional and mostly about cost transparency and choosing the provider. Check which providers the tool supports. BOTrento currently supports Anthropic (Claude) only.

Summary

  • EU hosting tells you where data is stored, not where the language model processes it.
  • GDPR and the AI Act set shared duties; ask the vendor to confirm answers in writing.
  • Use the 10-question checklist and be wary of absolute claims about data staying in the EU.
  • Choose a process where a human approves every outgoing email or quote.
  • Test on your own data before committing.

Try BOTrento free for 14 days

Test BOTrento on your real enquiries and quotes, with every draft approved by a person. Our team will get back to you within 24 hours on working days to help connect your mailbox. Fill in a few details and your access arrives by email right away. We are happy to help you connect your mailbox.

14 days free, no card and no API key of your own. The trial starts only once your data is first loaded. Then from €79 per month excl. VAT, AI included.

Nothing is sent without your approval. You add a card at the end, only if you continue; without payment access is blocked until you pay, we keep your data for 30 days and the export stays available. Pricing · See the process demo

No card and no API key of your own. Plan prices exclude VAT.

Related articles