How our own studio ended up building Rivilo.

-
October 2025
An agent called Bob.
We were curious about what the new models could actually do, and sceptical about what they were being sold as. The only way to find out was to put one inside our own team and watch. So: a Mac Mini under the desk, an API key, and a name to argue with.
Powered by OpenClaw
-
Early 2026
Bob went to work.
We put it inside the tools we already used. It read client tickets, wrote action plans, drafted the reply. Some of it was genuinely good. But most of it was disappointing and unreliable.
-
Then it broke. Repeatedly.
Telegram integrations. Background jobs that deleted their own files. Bob asking us at eight in the morning, for the third time, which list it was supposed to check.
-
We even tried giving it a body.
If something is part of the team, should it have a face? We put a robot on the table to find out. It danced. The answer was no, but the question was right.
-
So we stopped building and started asking.
What do we actually want from an AI in our team, and what should it never be allowed to do? We covered a wall in post-its before writing another line of code. One rule survived that wall: nothing leaves without a person approving it.
-
March 2026
Then we built our own.
Not a bot inside someone else's software. Our own board, our own agent, shaped around the way the studio actually works. We called it Ready.
Bob became Ready.
-
May 2026
We cancelled everything else.
Every project tool we had been paying for, gone. From then on the studio ran its own client work on the thing we built. If it broke, it broke on us first.
−€430a month
More than €5,000 a year no longer spent on software. Our work was starting to pay off.
-
Then we measured whether it was worth it.
Liking a tool is not the same as it saving time. So we tracked the two things that fill most of our week: writing client documents, and planning projects and roadmaps.
- 7 min
- median to write a client document, down from about 90 before Rivilo
- 25 of 36
- documents finished within half an hour
- 16 min
- on average to plan a project or roadmap, down from 3 to 4 hours
Document figures measured in our own account, 27 May to 13 September 2026. The before figures are our own estimates of the old way of working.
-
Summer 2026
And then it got its own name.
Ready was ours. Rivilo is the version built to work in a studio that is not ours: a new name, a new face, and the whole thing rebuilt around what a year of getting it wrong had taught us.
Ready became Rivilo.
-
We even had a meme phase.
A new name, an inbox that emptied itself, and apparently some time left over: the studio started making Rivilo memes.
-
September 2026
The first one from outside.
Rick at Uncode became the first person outside the studio to run his own work through it. Not a demo account. His actual week.
Rick and Robbin, before any of it worked properly. What do you do when people ask you to test their new AI tool that wants to read all your data? You say yes, of course.
Rick BossenbroekFounder, Uncode, Amsterdam
-
Now
Opening up, slowly.
A handful of early adopters are shaping it with us. We add them one at a time, because every new team still finds something we got wrong.
-
World domination.
(somewhere in the near future)