Five questions to ask about your ChatGPT setup

ChatGPT is an excellent product. The fact that your employees are already using it is not a problem to fix, it is a sign that the need exists. The question is not whether AI should be used at work, but under what conditions.

Here are five questions we think every Swedish organisation should be able to answer about their current setup. If you can answer yes to all five, you do not need to switch.

1. Do you know where your data is processed?

In EuroWork, every response shows where it was processed, which model answered, and that the content is never used to train any model. Not in a contract, but in the interface, while the work is ongoing. The default is European models. If someone intentionally wants to use ChatGPT or Claude for a harmless task, that is possible, but it is visible, and sensitive content is always routed to the EU.

2. What does it cost, and who sees the cost?

Most AI services are sold per licence and per month. What you do not see is the actual cost of each question. EuroWork shows the price per response, in your currency, and the price per quality level before the question is sent. A balanced question to a European model costs a few cents. The same question to a top US model costs several times more. Now you can choose with full transparency and set a cap for the total monthly cost.

3. What happens to what is created?

In chat tools, the response becomes text in a thread. In EuroWork, a long response becomes a document with version 1, information classification, review, and approval. The same thing, but manageable: the person revising can see what was written, by whom, and which version applies.

4. Can you control what is sent?

EuroWork classifies content into four levels and reviews every question before it is sent. Content that appears to contain personal data or confidential information is automatically escalated to stricter handling, and prohibited information is blocked entirely. This control is built into the service, not in a policy stored on the intranet.

5. Do employees get help to become proficient?

The challenge with AI in an organisation is not the technology, it is ensuring that fifteen people with different levels of prior knowledge can benefit from the same tool. EuroWork includes a built-in course, a training bot that answers questions about both AI and the service, and a gamified progression path that lets users track their own progress. There is no equivalent in standalone chat tools.

What you do not need to give up

Everything you are used to remains: the same way of writing, responses that stream in, markdown, tables, file uploads, image generation, projects with custom source material, custom assistants, and web searches with sources. We have deliberately not reinvented a new way to chat, because there is no reason to change what already works.

The difference lies in the framework around it: where data is processed, what it costs, what happens to the results, and who can follow up.

If you still stay with your current setup

Then we recommend three things: write a policy for what data may not be pasted in, disable model training on user data in the settings, and provide employees with half a day of training. This will take you a long way, regardless of which tool you choose.