forward deployed · ai in your business · ulm region
I build AI into your business — until it runs.
I advise, build it in and show your team — and I'm only out once the AI use-case runs in daily operations.
the problem
“We should do something with AI.”
AI is in every meeting. Teams still barely use it. Some type “write me an email” and end up disappointed. Others are afraid of doing something wrong. The problem isn't the technology — it's that nobody shows you how AI fits into your day. And where it doesn't help.
recognise yourself?
You know this — despite all your systems.
Data in silos, decisions on gut feel, knowledge in people's heads. Tap what's stuck for you — we'll sort out the rest.
„Which customer is drifting away, nobody tells me — until they're gone.“
„We quote from experience. Which job we lose money on, nobody knows.“
„Three weeks before the ISO audit I scrape the evidence together.“
why me
The AI builder who embeds and ships.
Most AI consultants have never shipped a line of code. Most developers never get it into daily operations. I build it in — and make sure you can carry on without me.
In tech this is called a Forward Deployed Engineer: the developer who embeds in your business and builds AI until it runs in production — instead of just advising. My first deployment was myself: four companies, built and run with AI as my team. Now I do it in yours.
I build what others talk about.
15 years of development. Apps, tools and AI systems — every day, with real users. I don't talk about AI second-hand.
built, not googledI know how people learn.
DVNLP Master Practitioner and certified behavioural counsellor. Knowledge that sticks — and behaviour that actually changes.
didactics + psychologyNo corporation, no overhead.
You talk to the person who's also in the room. Direct, regional, on eye level — without consultant-speak.
one point of contactwhat i actually do with it
I don't talk about AI. I live with it.
My best proof isn't my CV — it's what I organise and build with it every single day.
Conpilo, Dave & Tabea, the MPU coaching and my software development: four ventures, run by a single person — with AI as the team.
The little helpers that save me the most time every day — like my AI cockpit maj7 — I build myself, often in a single afternoon.
maj7 · try it yourself
This is how I run my life with Claude Code.
A folder with a handful of Markdown files, Claude as the organising partner, a small dashboard around it. No database, no cloud, no SaaS. The empty starter kit is right here to try — tasks, goals and focus are yours to fill in, in conversation with Claude.
the trick
It's not the one tool. It's the chord.
AI only becomes a colleague when three things play together.
Model
The language model that understands, writes, decides.
Tool
Only through the connection to your software and data does the model actually do something. The open standard for that is called MCP.
Order
Your goals and your knowledge, within which the model works coherently.
Most people stop at the first note and wonder why it doesn't sound right. I teach you the whole chord.
the hands · mcp
AI that works inside your software — not next to it.
MCP is the open standard that lets ChatGPT or Claude dock to your software. From then on the AI stops asking for copy-paste: it reads the bank statement, builds the form, proposes the post — and you decide.
My operating platform Conpilo now has over 80 such tools: marketing, forms, finance, communication. Every business gets its own access with clear permission scopes — read, draft, propose; publish only with explicit approval. What the AI handles through it today:
- 01Read the bank statement and pull the missing-receipts list: where a receipt is really missing, where only the recurring one deviates.
- 02Propose a whole package of posts — the boss decides line by line, with before/after and undo.
- 03Build a multi-page registration form, show and hide fields, done.
- 04Create a short link and QR code for the poster, and render the print sheet along with it.
The AI shows. The human decides.
The latest: one click on the cast icon and the admin tab follows the chat. The AI jumps to the right page, highlights the spot, asks its question with buttons — or lays down an offer as a card: field, old, new. Accept, decline, one by one. Nothing is written until the human clicks, with their permissions, five minutes of undo. No sending, no deleting, no money without them.
Idea in the morning, first version by evening, two days later with guided tours and actions for courses, people and appointments. That’s exactly how small it starts at your place, too.
And the other way round: customers ask a studio directly from ChatGPT or Claude about courses, timetable and prices — every business has its own address for that. The business itself connects ChatGPT or claude.ai to its Conpilo via login, no token tinkering.
straight talk
What I believe.
What I stand for — and what I build toward. Not everyone will agree. That's the point.
“ChatGPT alone is a toy.”
The value isn’t created in the chat, but when the model is docked to your tools and your order — via MCP, with clear permissions and approvals. Everything else is an expensive toy.
“Fear of software is more expensive than software today.”
The little helpers that save me the most time every day cost an afternoon. Whoever never starts out of fear pays with lost time — every single day.
“The valuable thing isn't the app. It's the ability to rebuild it tomorrow.”
Tools are replaceable. The ability to bring them into being is not.
“Whoever waits for the perfect moment has already missed it.”
The head start belongs to those who begin now — better bold and unfinished than perfect and too late.
“AI replaces the tedious, not the human.”
It takes the assembly-line work off our hands — so there's time again for what actually matters.
“An AI project is only done when it runs in daily operations — everything before that is work in progress.”
A PoC nobody uses is wasted time. I build until it actually runs at your place — and you can carry it on without me.
Want to work exactly like this?
Then let's build it.data protection
“And data protection?” — good that you ask.
The first question is almost always data protection. Rightly so. Except here AI isn't the risk — it's the solution. The more interesting question: what can AI do for your data protection?
Where is personal data even sitting?
AI searches your existing files — real names in old PDFs, personal data in Excels on the file server — and answers a question almost no business can answer today.
Anonymise & redact automatically.
Before the list goes to the tax advisor, it runs through a redaction step: names, addresses, IBANs are detected and replaced.
Local instead of cloud.
Capable models run on your own hardware today — the data never leaves the building. That's exactly what I build in.
With me, data protection isn't the small print — it's the reason I come to your premises instead of selling you finished SaaS.
offer & pricing
Built, not discussed. On your real tasks.
Hands-on Sprint
In a few days we build your first AI use-case on your real data — you end up with a working prototype you can touch and test.
Built in
I embed in your business and build the AI use-case with your team on your real data — until it runs in daily operations. No PoC that fizzles out.
Enablement
We anchor AI in your processes, dock ChatGPT or Claude to your software via MCP — and keep building where off-the-shelf isn’t enough.
Prices are a guideline — we set the fixed price after the first call.
who it's for
For SMEs that don't want to wait.
You run a company with 10 to 200 people in the Ulm, Neu-Ulm, Memmingen or Augsburg area? You hear “AI” everywhere and want to get concrete instead of reading another whitepaper? Then let's talk.
how it works
Four steps, fixed price, no fine print.
Intro call
30 minutes to find where AI gives you the biggest lever — and whether I'm the right fit. No sales pressure.
Fixed price
You get a clear offer at a fixed price. No hourly billing, no surprise at the end.
We build it
We build your use-case on real data — you end up with something you can use right away.
You own it
You leave with skills that stay in the team. I'm out the moment you can carry on without me.
faq
Forward Deployed Engineer — explained.
- What is a forward deployed engineer?
- A forward deployed engineer (FDE) is a developer who embeds directly in a business and builds AI into production there — instead of leaving it at a recommendation. The term comes from Silicon Valley (Palantir, OpenAI, Anthropic). I bring the model to small and mid-sized companies.
- How is that different from an AI consultant?
- A consultant gives recommendations and leaves. I write production code and stay until the solution runs in daily operations — and hand it over so your team can carry on without me.
- What does it cost, and how does it work?
- It starts with a free first call. Then a fixed price: a hands-on sprint from €1,900, a built-in project at a fixed price, or ongoing enablement from €990 a month. No timesheets.
- Do you also work remotely?
- Yes. I build remote-first and plug into your systems — on-site only where it's truly needed. That makes it fast, without anyone sitting in seminars for weeks.
- What is MCP — and why does it matter?
- MCP (Model Context Protocol) is an open standard that lets AI models like ChatGPT or Claude dock to software: read, draft, propose, and with approval also execute. Only then does a chat become a colleague that works inside your systems. I build such connections — to your existing software, with clear permissions and approvals.
contact
30 minutes that save you a year of trial and error.
No sales pitch, no pressure. We look together at where AI has the biggest leverage for you — and whether I'm the right person for it.
Book a call