From LunarStorm to EuroWork

In 1996, there wasn’t much to do on the internet. You could read one page. Then you could read another one.

That year, Rickard Eriksson built Stajl Plejs Community, a place where you could leave a message for someone and get one back. Mark Zuckerberg turned twelve that year. By the turn of the millennium, the site had changed its name to LunarStorm, and at its peak, 1.2 million Swedes were members. For a whole generation, it wasn’t just another site. It was the internet.

The site won the Big Internet Prize, and the jury wrote something everyone who was there already knew: LunarStorm had taught young Swedes more about computers and the web than any government initiative, because they actually wanted to log in.

That’s the whole point, and it still holds true. Technology that people want to use does more than technology they’re told to use.

Thirty years later, the same idea

Now he’s back. The company behind LunarStorm was called LunarWorks. The new one is called EuroWork. Some patterns stick.

This time, it’s not about guestbooks but about what happens to the contract you pasted into an AI service last Tuesday.

Through Nischad Utbildning, Rickard Eriksson and his colleagues have worked with over 10,000 companies and public sector organisations on AI and digital communication. After enough training days, one pattern becomes impossible to miss: AI is already being used everywhere, but quietly.

Someone pastes a contract to understand a paragraph. Someone uploads a report with personal data to get a summary. Almost always with a private account, in their free time, because it was faster that way. It’s often called shadow AI, and no one does it out of malice. They do it because the workday is too short.

The problem is that no one can answer the simple follow-up question: where did the data go?

What he hears when he asks

The most common response he encounters isn’t someone who wants to ban AI. It’s someone who wants to be able to say yes without feeling bad about it. A manager who suspects what’s going on but doesn’t want to be the one to scold. A data protection officer who’d rather not know.

That’s what EuroWork was built for. A place to do it instead, with rules set by the organisation, not by each individual.

Here’s how it works in practice. AI processing happens with European providers by default, and nothing you type is used to train someone else’s AI. ChatGPT and Claude are available as options for those who consciously want them, never for sensitive content, and an administrator can turn them off completely.

Before anything is sent, it’s automatically reviewed. If it contains personal identity numbers, health data, or trade secrets, the protection level is raised, and you’ll see what triggered it. No one needs to remember a rule.

More than a chat window

A chat window would have been easy to build. What takes time is making AI something you use for real work.

You can build your own assistants on your own material. One that had the graphical manual as a source read an ad and pointed out the forbidden colour, the logo in the wrong corner, and the missing primary colour, with the rule quoted from the manual.

A document can be translated into 35 languages with headings, tables, and bullet points staying in place, so you can send it on without anyone having to put it back together.

The audio from a meeting becomes a record with decisions and responsible parties. Where something was missing in the audio, it says so, instead of guessing.

And under every answer, it shows what it cost. Usually a few pence. Management sees the total per employee and department and sets a monthly cap that the service stays within. No one has to wonder where the money went, which is one of the few questions more uncomfortable than the one about the data.

What he usually says

"AI won’t replace you. You’ll be replaced by someone who knows the right AI tools."

It’s an uncomfortable sentence, and it’s said with warmth. It’s not about scaring anyone into learning, but about the person who gets the tool in their hands and a little help along the way doing just fine. It was true for the fifteen-year-olds who learned HTML to make their LunarStorm profile look nicer. It’s true for the finance assistant who discovers that reconciliation takes twenty minutes instead of an afternoon.

From a guestbook to a contract is longer than it sounds. But the core question is the same: are you building something people want to open in the morning, or something they have to?