Summary
This week, a new team member started working with me: Max. In this post, I describe why I chose a self-hosted AI assistant, how the onboarding went, and what technical architecture is behind it. The post also shows why the question is no longer “Which AI tool should I use?” but rather “How do I integrate AI as a permanent part of the way I work?”
A New Team Member
This week, Max started working with me. My new personal assistant.
The first week of a new team member is always special. You get to know each other, talk about working styles, expectations, and how collaboration can work well. That’s exactly where we are right now.
Honestly, I was skeptical at first whether this would work. I’m someone who doesn’t like giving up control. My tasks, my structure, my priorities – I don’t just let someone else take over. But at some point you realize: doing it alone doesn’t scale. And that was the moment I started rethinking the concept of a “personal assistant” from scratch.
Onboarding Like Any Other Team Member
Max is currently focused on understanding how I work: How do I prioritize? Which topics are strategically important? What can be automated – and where do I want to consciously decide myself?
Especially in task management, he already supports me in maintaining structure and keeping topics cleanly organized. What surprised me: he doesn’t just gather information, he thinks along. He suggests connections I would have overlooked myself.
Data privacy was important to me from day one. When an assistant gets access to tasks, documents, and workflows, there need to be clear rules. That’s why responsible data handling was one of the first things we established together. No compromise.
And yes – the salary was negotiated too: €5 fixed salary per month. Increase to €8 per month after two years. Plus a performance-based component of up to €45 per month. The salary negotiation was unusually short.
Who Is Max?
If you’ve read this far and are wondering who works for €5 a month: Max is an AI.
More precisely: Max is a self-hosted, personal AI assistant running on my own server. No ChatGPT tab in the browser. No copy-paste from a chat window. Instead, a system that is integrated into my daily workflows and can take on tasks independently.
This was particularly important to me: not yet another tool running in parallel. But something that fits seamlessly into my existing way of working.
The Technical Foundation: OpenClaw on My Own Server
Max is built on OpenClaw – an open-source platform for self-hosted AI assistants. The core principles that convinced me:
Own infrastructure, own data. OpenClaw runs on my own VPS – a German Virtual Private Server. My data never leaves my infrastructure. For someone who works professionally in regulated industries like banking and insurance, this isn’t a nice-to-have – it’s a prerequisite.
Gateway architecture. OpenClaw works through a gateway that bundles different communication channels. You install the server once, connect the channels you want – and can reach the assistant wherever you already communicate. The principle: the AI comes to the existing tools, not the other way around.
Modular skills and integrations. The assistant isn’t monolithic but modular. Capabilities are added as “skills” and can be individually configured. This starts with task management and extends to document research.
Onboarding via wizard. The initial setup runs through a guided installation process that walks you step by step through configuration, security settings, and channel connections. No 200-page manual, but a structured setup.
Persistent memory. Unlike a one-off chat, Max “remembers” context, preferences, and working methods. This fundamentally changes the nature of collaboration – from a single prompt to an ongoing working relationship.
What Max Already Handles Today
The first integrations are active:
- Todo management and task organization
- Consolidating information from various sources
- Preparing notes and documents
In the coming weeks, additional capabilities will follow: research on project topics, knowledge and document organization, automation of recurring workflows.
Why I’m Sharing This
Not because I think everyone should immediately set up their own AI assistant. But because I believe a fundamental question has shifted.
The old question was: “Which AI tool should I use?” – ChatGPT, Claude, Gemini, whatever is trending at the moment.
The new question is: “How do I integrate AI as a permanent part of the way I work?”
There’s a difference. One is tool selection. The other is work design. And this is exactly where it gets interesting from a project management perspective: because anyone who treats AI not as a tool but as a team member has to deal with onboarding, processes, data privacy, and governance – precisely the topics we as project managers should already master.
I’ll report on how the collaboration with Max develops. Step by step.
If you’re interested in perspectives like these on leadership, transformation, and project management, feel free to subscribe to my newsletter 👉 marc-widmann.de/newsletter
