Skip to content
Contact

Process

Understand first, then build.

Whether an agent is meant to pre-check your quotations or a solver has to compute your circuit, we run every project through the same four phases. Each one ends with a result you can hold in your hands, and with your decision on whether to continue.

The four phases

Analysis → Production
  1. → brief

    Analysis

    We look at your process where it actually happens and talk to the people who work with it every day. We record how much time it costs today and which systems and data are involved. After that you know whether the project is worth it and what we would solve it with.

    You receive

    • Brief with goals, assumptions and risks
    • Baseline: how much time the process costs today
    • Recommendation: fixed rules, an algorithm, an AI agent or a combination
  2. → running prototype

    Prototype

    Within a few weeks a first version is running, tested with your real data. If that data must not leave the building, we work with a local model on your hardware or with realistic test data. That way you decide on results rather than on slides.

    You receive

    • Running prototype on your systems or a copy of them
    • First results on your data, measured against the baseline
    • A clear line between what is hard-coded and what a model decides
  3. → your sign-off

    Presentation

    We show the prototype to the people who will work with it and put the numbers next to the baseline: how much time does it save, how often is it right, where does it still need a person? What works stays. What is missing or gets in the way goes on the list. Only once you are convinced do we move into production.

    You receive

    • Demonstration to your team
    • Before-and-after comparison of time saved and error rate
    • Effort, schedule and your sign-off for production
  4. → live operation

    Production

    We bring the solution into your operations and plug it in where your team already works: the ERP, the mailbox, the web interface or the machine. With permissions, logging and monitoring. Your team gets trained, and after go-live we stay on board.

    You receive

    • Release with a reproducible build, permissions and audit trail
    • Training and documentation your team actually understands
    • Monitoring, support and further development from a single source

Principles

  • If we cannot make it better, we say so

    Not every process gets better through software, and some problems are organisational rather than technical. We check that in the first conversation and advise against it when in doubt, before a quote exists.

  • Verifiable where it matters

    Whatever can be solved with fixed rules and algorithms, we solve that way, because then the result can be checked and comes out the same every time. Where money, safety or a measurement depends on it, no language model makes the call. A model only comes in where rules run out and the numbers show that it performs better.

  • Measure instead of assume

    Before we build, we measure how much time a process costs today or how accurate a calculation has to be. Afterwards we measure again. Whether the project paid off is shown by that comparison, not by our gut feeling.

  • Built in, not bolted on

    A solution that needs an extra window to be opened rarely gets used. We connect it to the systems you already have, settle the interfaces early and test the chain as a whole.

Collaboration

How we work with you.

A small team and short paths. Whoever plans your project also builds it. You steer the priorities with us and can see at any time what we are working on and what it costs.

Communication
Regular demos of runnable increments instead of status slides. One fixed contact person and direct channels to the developers.
Transparency
Effort, costs and open points are visible at all times. Changes in scope are discussed before they happen.
Your data
On request the models run entirely on your own hardware, so no data leaves the building. The repository, documentation and access credentials belong to you.
Documentation
It is written alongside the code, not at the end of the project. Architecture, usage and operation are described so that your team can carry on without us. After that you decide whether your team continues the development or we do.

Let’s talk

Which process costs you too much time today?

Describe it in a few sentences, whether it happens in the office or at the machine. We reply within two working days with a first assessment.

Start a conversation Explore our work