AI at a Glance

From Prompt to Running Process

A database application can be built in an afternoon by chat today. An executable business process cannot. We closed the gap.

Why BPM is the hard case for AI

Anyone who needs a database-driven application today has two options: buy an expensive standard platform – or vibe-code it from scratch. The result is a clean relational data model you can open and work with in any tool you like.

For business processes, that path does not work. BPMN 2.0 stands in between: a standard from another era, without a successor, with lanes, gateways and symbols that require training courses. Producing a diagram is one thing. Producing a process that actually runs afterwards – with forms, actions and program logic – is something else entirely.

That is precisely the gap our platform closes: vibe coding, but for executable business processes.

Other AI draws the process.

Ours builds it.

Prompt
- Process model
- Forms
- Logic
- Actions
Ready to run

Two proofs instead of a promise

AI capabilities are easy to claim. We demonstrate ours in two places: in our own product, and in a delivered customer project.

In our own product

AI Control Hub

The MCP server of our platform. A prompt produces a complete process: a diagram with actions, low-code logic and the matching forms in the form editor – directly executable. Through a WebSocket connection into Workflow Studio you watch it take shape live.

Explore the AI Control Hub

In a customer project

AI Chat for Business Applications

A grown business application gains a language interface: users query their data in natural language, compare it, have documents analysed and changes prepared – within the existing permission system and with explicit approval before every change. First delivered in a production equipment database.

Explore the AI chat

Ten years of machine learning, not ten months

We have been working with machine learning for around ten years – long before language models, back then with neural networks for individual, clearly defined special cases. That experience is why we were early: our platform became an MCP server back in 2025, when the Model Context Protocol was only just emerging. MCP is widely established today, and other vendors offer it for their diagram editors. Our difference sits one layer deeper: not just modelling, but producing a process that can actually be executed.

Back in 2022, German news magazine DER SPIEGEL reported on our work with generative AI and on what was, by our own account, the first Stable Diffusion extension for Adobe Photoshop.

DER SPIEGEL, 46/2022 – “Die Picasso-Maschine”

Three years before we turned our BPM platform into an MCP server, we had already built and shipped generative AI in production.

Your server. Your keys. Your logs.

With AI in the enterprise, the model’s capability is often no longer the decisive question. What matters is who gets access to data and actions – and who stays in control. Our answer is the same at every point:

No direct data access The language model never talks to your database. It works exclusively through clearly defined tools of a secured MCP interface.
No special privileges The AI receives no permissions of its own. Your existing rights system applies – checked before the action and again before saving.
No silent changes Changes are proposed as a before-and-after comparison and executed only after explicit confirmation.
Fully logged Proposals, confirmations, rejections and executed changes are logged and can be marked as AI-assisted.
On-premise and vendor-neutral Operated in your own infrastructure. The open MCP standard keeps the choice of model your decision – now and later.

See it live

It is most convincing in motion: one prompt, and the process appears in front of you. We will show you on your own use case.

Book a Demo Contact us