Services

We start with a single process, extend what proves itself and safeguard each stage. This page shows what you get from us along the way.

One process as a pilot

The starting point is a single process. We walk through it with your team, connect an agent, and your team checks the result against its own cases. This usually takes a few weeks.

Analyse processes and plan an AI pilot

We follow a typical process and its exceptions with the people who handle it. You get a recommendation on which step to automate and how to measure the result.

  • Identify duplicate entries, delays and follow-ups
  • Estimate the effort and data a pilot needs
  • Agree measurable acceptance criteria with the team

Connect AI agents to business systems

An agent can match an enquiry to an order or draft a reply. We integrate the applications it needs and agree when a person takes over.

  • Connect email, ticketing or ERP systems, for example via MCP or the command line
  • Define data flows, permissions and approvals
  • Test exceptions and make actions traceable

Develop business software with AI agents

Our Agentic Framework supports analysis, implementation, testing and code review. When modernising software, we trace its existing rules and check new behaviour against real cases.

  • Find business rules in the code and validate them with specialists
  • Implement features incrementally in .NET
  • Check changes with acceptance tests and independent reviews

From our own work: three weeks instead of an estimated nine months

The three weeks are calendar time: during them the agents worked on all seven days of the week, evenings and nights included. The nine months are our estimate for the same rebuild without AI, with two developers, five days a week and about ten hours a day.
The case in detail

Connect ERP, warehouse logistics and transport management

A pallet scanned at the loading dock has to be recorded in both the warehouse and the ERP system. We develop and modernise the applications that connect them.

  • Connect warehousing and shipping to SAP using IDocs
  • Integrate barcode capture, loading and transport planning
  • Verify data exchange with external warehouses

How we work in detail

What is behind the way we work: the difference from a prototype, the handling of personal data and our development process.

From prototype to software a business relies on

Many good solutions start with someone who knows a problem from daily work. It used to become a spreadsheet. Today vibe coding produces a prototype in a few hours. That is a good start.

Once a department depends on it, different questions come up. Who reviews the code? Who maintains the solution? What happens in an outage?

Vibe coding

  • An idea tried out in hours
  • Runs on one desk
  • Nobody reviews the code
  • Costs next to nothing

Professional development with AI

  • Requirements with acceptance criteria
  • Access rights, audit log and outage test
  • Test first, a second agent reviews, a person sets the scope and is accountable for the result
  • Costs far less than development without AI agents

We take the step from prototype to production software with you. With AI agents it costs far less than the conventional route.

Once customer data is involved

Vibe coding is great for trying out an idea quickly. With real customer or employee data, the GDPR comes into play. It is then no longer enough for someone to put something together over lunch and show an AI service whatever happens to be at hand.

Your business then has questions to answer. These are the four we always settle first.

Where does the data go?
Together we define which AI service sees which data, where it is processed, on what contractual basis and whether the service uses it for its own training.
Who can do what?
Each user and each agent gets only the rights needed for their step. Reading, changing and sending are controlled separately.
What happened?
Business-relevant actions are logged, so you can trace later who or what accessed data or changed something.
What if someone asks for their data?
We plan from the start how one person's data can be found, exported, corrected and erased or restricted, including at the AI service.

The legal assessment stays with you and your data protection officer or legal adviser. We provide the technical basis for it.

More about secure integration

How we build software

Agents build and review. A person decides beforehand how far they go on their own.

PersonAgent

  1. 1 (Person)

    Define the task

    We describe the task with acceptance criteria and define how far the agent works on its own.

  2. 2 (Agent)

    Test first

    An agent writes the test, which fails at first.

  3. 3 (Agent)

    Implement

    An agent builds until the test passes.

  4. 4 (Agent)

    Review independently

    A second agent reviews the code without having seen it built.

  5. 5 (Person)

    Acceptance

    Where tests cannot clearly show that the result is right, a person tests it. Otherwise the task ends with delivery, as defined in step 1.

More about software development

Which process takes up too much of your team’s time?

Bring us a process that regularly stalls, generates too many follow-ups or requires the same data to be entered in multiple systems. We will discuss where the effort goes and which first step makes sense for your business.

Arrange an initial conversation