Agentic engineering for engineering organisations

AI agents that actually do the work in your pipeline.

We turn scattered AI experiments into a dependable pipeline: agents with a clear brief, automated checks and measurement points — embedded in your repositories, reviews and releases. In weeks, not years.

First call is 30 minutes, free of charge. No slides, no sales funnel.

Where teams are

The tools arrived long ago. The way of working did not.

Almost every engineering department now runs AI tooling. Licences are paid for, a few developers are enthusiastic — and lead time, quality and predictability have not moved.

01

A few fast people, no method

Two or three engineers ship noticeably faster. The rest watch. What the fast ones do differently is written down nowhere — so it cannot be handed on and cannot be reviewed.

02

No baseline, no decision

Without numbers from before, every discussion is a matter of taste. Leadership asks about impact, the department answers with anecdotes. Volume wins the argument.

03

Agents without guardrails

Anything running without checks, scoped permissions and an audit trail surfaces at the first audit or the first incident. Then everything gets stopped — including the parts that worked.

Agentic engineering is not a tooling question. It is a question of how work is done — and that can be built, reviewed and handed over.

How we work

Four steps we never skip.

The same order in every engagement. It is the reason a number stands at the end instead of a story.

  1. Week 0

    Measure what is true today

    Lead time for a change, time to review, defects after release, effort per ticket. We record the current state before touching anything — otherwise improvement cannot be evidenced later.

  2. Weeks 1–2

    Automate one real path

    No sandbox, no demo repository. We take a path your team walks daily and build agents for it: clear brief, automated checks, and a human at the approval step.

  3. Weeks 3–5

    Wire it into the pipeline

    Agents run where your work runs: repositories, reviews, tickets and CI. With scoped permissions, a complete audit trail, and a kill switch any manager can find without us.

  4. Afterwards

    Hand over, do not stay

    Playbook, configuration in the repository, two of your people trained. We measure again and say plainly whether the effort was worth it — including when the answer is uncomfortable.

What "agentic" means here

An agent, to us, is not a chat window. It is a bounded piece of work that a model carries out on its own — with access to exactly the tools that job needs, with automated checks, and with a trail that shows afterwards what happened. That is what turns an impressive tool into a dependable part of engineering.

That is the work we do: scope the brief, build the checks, limit the permissions, define the handover to humans. And then measure whether it held.

Career background LetMeShip InnoGames Funatics Kilo Moana

What took a team of five two years ago, I now build alone with AI agents, in a fraction of the time. Companies that ignore that end up paying for their competitors.

Alexander Siemer-Schmetzke, founder of Moana Works
12+ Years of software engineering, six of them as Head of Software Engineering
4–6 Weeks from assessment to the first defensible measurement
EU Processing and hosting inside the EU wherever the task allows

What would an agent take over in your pipeline today?

Thirty minutes, straight with the people doing the work. You get an honest read — including if it is that you do not need us for this.

No sales funnel, no deck. If it is not a fit, we say so on the call.