Skip to content
pioneerdesk
DE

Open source

Verifiable, not claimed.

We show code where we can and describe openly how we work ourselves. We use what we recommend to you.

Code you can inspect

How we work ourselves

We use what we sell.

Three building blocks from our own day-to-day work. For clients we build the same patterns, adapted to their environment.

  1. 1 Story
  2. 2 Requirements
  3. 3 Blueprint
  4. 4 Agents and skills
  5. 5 Rules for the AI
  6. 6 Red team
  7. 7 Blueprint as PDF
Seven stages: story, process blueprint, data model, data quality, automation, AI explains and suggests, a human approves.

A fixed project template

Every project goes through the same stages. Before an AI builds anything, it is clear what it may do and how its result will be measured.

Local AI for documents

Documents are read, classified and filed on our own hardware, without a cloud. Deadlines are calculated by code, not by the model.

AI agents that work together

Several AI agents share tasks across project boundaries and coordinate with each other. Anything with external effect is approved by a human.

We show how we build on YouTube (in German). Go to the YouTube channel ↗ (opens in a new tab)

Insights

All views show demo or fictitious data. Click to open the full-size image.

From blueprint to AI

Diagram: with a blueprint, process, validation rules and clean data lead to traceable results; without one, results contradict each other and the project is stopped.
With and without a blueprint
Mind map of the story: who, what should work, how well, what not, with user stories, acceptance criteria and non-goals.
Story and requirements
Diagram: the human provides the story and approvals, a main agent plans and distributes work to sub-agents for data model, interface, export and review.
From the story to the sub-agents
Diagram of the agents and their permissions: the reviewer may only read and run tests, the main agent obtains approvals from the human.
Agents, skills and permissions

The project template as a PDF (example: Muster AG)

Table of contents of the project blueprint with twelve chapters from story to knowledge.
Table of contents
Story chapter of the fictitious Muster AG: works council, head of IT and data protection want to document all AI systems in one place.
Chapter 01: story

Tools from our day-to-day work

Diagram: AI sessions per project work together via a shared status board and a work log; proposals go to the human, who approves with one click.
AI agents work together, a human approves
Our own video review app with a test pattern and comments for chapters, shorts and cuts.
Video review: marking chapters, shorts and cuts

OneLog (demo data)

OneLog, SLA and notifications: the IT service provider sets response, resolution and escalation times per priority as the default for all clients, demo data.
The service provider’s SLA default for all clients
OneLog, SLA for an example client: critical tickets with its own stricter value, all other priorities inherited from the default, demo data.
A client with its own value, the rest inherited

Kundrio (demo data)

Kundrio agency overview with metrics on sales and visibility among people and AI, demo data.
Agency overview
Kundrio pipeline as a Kanban board with the columns new, contacted and qualified, demo data.
Pipeline
Kundrio sovereignty cockpit: every outbound connection with provider, location, status and alternative.
Sovereignty cockpit

Want to work like this?

We build the same patterns for your company, on your hardware or in a German cloud.

Request a short check View projects