Blog
Working notes from the platform.
This is where we write about what it takes to put SIMOSphere AI to work: which systems get connected over MCP, what an AI teammate actually takes off your desk, how a team gets up to speed, and what running the platform in your own data center demands. Advisory topics such as data strategy, governance, and risk management live separately, at simo-online.com.
Topics
Six tracks through the archive.
Every topic has its own address. You can share it, link to it, and subscribe to it, and it works with JavaScript switched off.
All posts
Newest first.
MCP Connectors
Connecting CRM and ERP to a language model · the four-layer architecture
Why point-to-point integration collapses at the third data source, how an MCP service takes over the mediation, and the setup times we measured across completed rollouts.
6 min read
AI Training
AI literacy under Article 4 · why a one-page memo does not discharge the duty
Since 2 February 2025 the AI Act has required sufficient AI literacy among your own staff. What that means in practice, which level fits which role, and how to evidence where you stand.
5 min read
MCP Connectors
Building your own MCP server · and shaping it so it survives production
When a custom MCP service is worth it, what the protocol looks like on the wire, how to describe a tool, and the five rules that decide the quality of a connector.
7 min read
Platform and Operations
Sovereign AI is not a seal · four questions decide it
Four questions decide whether an AI setup is genuinely sovereign: where computation happens, who holds the keys, who can be legally compelled, and whether you can get back out. Plus what sovereignty costs.
5 min read
AI Teammates
Five teammates, one by one · role, skills, and the line where they hand back
Willi, Rita, Aml, Doro and Reggi each have a role, a remit and a boundary. What each one takes on, how it reaches your data, and what it explicitly does not decide.
5 min read
Platform and Operations
We are not a software vendor · when building your own actually pays
Four tests decide whether building your own software makes sense. In most cases at least one of them fails, and that is a good outcome.
5 min read
Platform and Operations
AI orchestration for mid-sized companies · why the next tool will not fix it
Everyone who experimented with AI in 2025 is now sitting on half a dozen disconnected tools. What orchestration means, which three patterns cover most of the work, and where adoption really breaks down.
6 min read
AI in Practice
The digital twin in construction · from a single building to the whole company
Construction made the digital twin what it is, first as a BIM model, then as a live picture of operations. What carries over to a whole company, and where orchestration actually begins.
9 min read
See it run on your own data.
A demo walks the platform through a workflow from your own company, not a sample data set. We prepare it together with you.