Playbooks: how to apply it

Playbook: use AI without giving away your data

For businesses that want to use AI, but where the question of where the data goes keeps coming up. Five steps to a setup on EU servers under your own control.

Sound familiar?

Colleagues already use an AI service, but nobody knows what goes into it.
A customer asks in a tender where your data is stored.
Management wants to move forward and IT first wants to know what happens to the data.

Step 1. Sort your data

Decide for each type of data how sensitive it is. Public texts can go anywhere. Internal documents preferably not. Personal data and customer secrets belong only in an environment where you decide who has access.

Step 2. Choose where it runs

The more sensitive the data, the closer to home the AI should run.

A service from an American company is the quickest to set up. The data falls under American law, even when it is stored in Europe.
A European AI service keeps the data under European law.
An open model on rented servers in the EU gives you the most control: you decide who has access and what is kept.

Step 3. Start with one task

Pick work where text goes in and text comes out: summarising, a first draft of a quote, a reply to a customer question. For the first weeks, measure how much a colleague has to change before it can go out. That number says more than any demo.

Step 4. Rent computing power by the hour

An open model runs on an ordinary server for quiet work. For busy moments you rent a heavier server with a GPU by the hour and release it when things are quiet. You only pay for the hours you use.

Step 5. Record what runs where

Write down which service processes which data, in which country and under which data processing agreement. That is your answer to a customer or auditor, and it belongs in your record of processing activities. The Dutch Data Protection Authority has a template for small businesses.

What it costs

A rented EU server running an open model costs a fixed amount per month. Computing power for busy moments is paid by the hour. American or European AI services charge per use. Which route is cheapest depends on how much work goes through it.

How we do it ourselves

Our own AI employees run on servers in France, with open models we manage ourselves. For heavy work we rent a GPU by the hour. After twenty minutes without work it is released automatically.

Frequently asked questions

Is an open model just as good?

For summarising, writing and answering questions about your own documents it comes close. For the hardest reasoning tasks the big services are ahead. Test it on your own work before you choose.

Do I have to manage servers myself?

No. You can outsource the management, as long as the servers and the agreements are in your own company's name.

Read on

AI in your own environment Data Processing Agreement

In this series

Playbook: let AI answer your phone Playbook: keep track of your market and competitors Playbook: stop retyping data between your systems Playbook: connect your AI assistant safely with MCP Playbook: edge AI next to your existing PLC Playbook: get the basics of your security in order Playbook: getting your team to use AI All knowledge

Sources

Our approach

From advice to management, with one partner

The same rhythm for every project.

Advice

We start with a conversation, not with code. First clarity on what you need and what you do not.

Build

We build and integrate to measure, with technology that fits you. You own your data and your code.

Manage

We stay involved: monitoring, adjusting and growing with your business.

See what fits your environment

Thirty minutes is enough to see which route suits you.

Call 085 083 5775

A 30-minute intake