Approach

I pass on what I use myself every day.

AI is changing faster than any topic I have worked with so far. That’s why I don’t pass on theory, but what has proven itself in my own companies, in my own way of working and in my AI projects, and I also say where things still get stuck.

To the three stages To the principles

Hands shape layers of colorful material on a workbench

The three stages

The three stages build on each other.

Every stage is a good starting point, and at each one there is a next step worth taking. This is how I see the three stages, and this is how I work with you at each of them.

Stage 1

You chat

You work in a chat with Claude, ChatGPT or Gemini, maybe already with projects. You start every step yourself.

How I help you here

I show you how to give the AI the right context, how to check its results and which tasks are really worth it. After that you work with set projects and templates instead of single chats, and the quality becomes reliable.

What often gets stuck

The results vary, and every time you have to explain to the AI again what it’s all about.

The next step

As soon as you can describe a task you keep doing in the chat so clearly that the AI carries it out on its own, you are on your way to Stage 2. Often that is a first skill or a small workflow in n8n.

Matching formats

Course Room AI Strategy Workshop AI Audit

Stage 2

You build

You let the AI carry out steps on its own, for example with Cowork, Claude Code or workflows in n8n. You start it, and the rest runs.

How I help you here

We turn single automations into workflows with a clear goal, fixed review steps and a memory that grows with them. You learn when the AI may take the lead and when a fixed workflow is the better choice.

What often gets stuck

A lot of it works once, but not reliably, and in the end you still have to watch over it yourself.

The next step

You reach Stage 3 when several of these workflows start on their own, on a schedule or a trigger, and you only approve what the agents are not allowed to decide for themselves.

Matching formats

Macher Kreis Team Circle Revenue & Growth

Stage 3

You orchestrate

Several agents work together, on a schedule or a trigger, even when you are not there. You approve what they are not allowed to decide for themselves.

How I help you here

I share how I run my own system: how agents hand work to each other, how approvals and reviews work, how the system learns from mistakes and what it costs. With the A2A Benchmark we measure where your systems stand.

What often gets stuck

With every agent, oversight, costs and upkeep grow, and mistakes only show up late.

What this stage is about

At this stage it is less about adding more agents and more about running them reliably. That means measuring results, keeping an eye on costs and letting the system learn from its mistakes.

Matching formats

A2A Benchmark Macher Kreis Revenue & Growth

Six principles

What I stick to, whatever the stage.

Tools change almost every month. These six principles have stayed the same for me for a long time, and they apply to a single chat just as much as to a whole team of agents.

We start with what should be running at the end.

Before we touch a tool, we decide which result counts and how we will know it has been reached. An agent that doesn’t know its goal works very hard in the wrong direction.

The AI shouldn’t start from zero every time.

Every new chat is like a new colleague on their very first day. That’s why we build up memory, rules and knowledge so the AI doesn’t start over each time.

What has to happen reliably gets a fixed safeguard.

AI is strong where it has room to shape things, but not always reliable when something has to run the same way every time. So we decide what the AI decides on its own and what gets a fixed safeguard and a check.

Autonomy grows step by step.

First something runs once with a prompt, then as a skill, then as an automation, and only after that on its own. If you start with the last step, you often don’t know how to lead the agents.

You build it yourself, and the knowledge stays with you.

I guide, explain and review alongside you, but the systems belong to you and your team. That way nothing depends on me once I’m no longer involved.

Every mistake becomes a rule.

When something goes wrong, we look for the cause and record the fix so it doesn’t happen again. That way the system becomes more reliable every week.

Let’s find out what the next stage is for you.

In a call we look at where you or your team stand right now and which format fits. If you prefer to write, you can reach me at .

Book a call