TL;DR

The unlikely journey from insurance clerk to vibe-coder — via boardrooms, startups, and a pandemic

Two and a half years ago, I had never written a line of code in my life. Not because I lacked exposure to technology — I’d spent 25 years in and around it. I’d managed international IT projects, led an AI-native startup, contributed to Bitkom working groups on AI and data strategy, and shaped digital transformation strategies for companies across Europe.

The unlikely journey from insurance clerk to vibe-coder — via boardrooms, startups, and a pandemic


Two and a half years ago, I had never written a line of code in my life. Not because I lacked exposure to technology — I’d spent 25 years in and around it. I’d managed international IT projects, led an AI-native startup, contributed to Bitkom working groups on AI and data strategy, and shaped digital transformation strategies for companies across Europe. But writing actual software? That was always someone else’s job.

Today, I have seven production applications running. Real products, with real users, solving real problems. One manages marinas across Europe with AI-powered contract processing. Another turns any screenshot into calendar events. A third generates professional social media content with five different AI image providers and video generation. One monitors smart energy meters in real-time with prepaid billing.

I built all of them without a development team. And this is the story of how a Versicherungskaufmann ended up here.

30 Years of Almost-But-Not-Quite Coding

My career started where many German careers do: with a proper apprenticeship. I trained as a Versicherungskaufmann — insurance clerk — at R+V Versicherung in the mid-90s. It was solid, traditional work. Not exactly Silicon Valley.

But technology kept pulling me closer.

At Pironet NDH (1998-2001), I became an International Project Manager. This was the dot-com era, and I was deep in digital infrastructure — but always on the business side. I managed the technology; I didn’t build it.

Then came Mummert Consulting (later Steria, now Sopra Steria). This is where I was actually supposed to learn programming. It was part of the consulting toolkit — everyone should be able to code, they said. But I found ways around it. I was better at understanding what needed to be built than building it myself. Instead of code, I invested my energy elsewhere: I completed my BWL degree (business administration) while working full-time.

Looking back, dodging that coding requirement was one of the most ironic decisions of my career.

The Bertelsmann Years: Scale, Complexity, Leadership

From 2005, I spent nearly a decade at arvato infoscore / Bertelsmann, climbing from consultant to Vice President and Location Manager. This was enterprise at scale — risk management, eCommerce fraud prevention, credit scoring. I spoke at Internet World about risk management fulfillment. I managed large teams across multiple locations.

I was as close to technology as you could get without actually touching code. I understood architectures, data flows, APIs, scoring models. I could design systems, write specs, argue with developers about what was possible. But write a for-loop? Not me.

DATAlovers: My First Taste of Building Something New

In 2015, everything changed. I became CEO of DATAlovers AG — a startup in Mainz that was doing AI and Big Data before it was cool. This was 2015, years before ChatGPT. We built what we called „bearch“ — a B2B data intelligence platform. We were featured in Bitkom’s „Excellence in Big Data“ study. We explored blockchain for company identification. We raised two investment rounds.

DATAlovers was, in many ways, an AI-native company before the term existed. And running it taught me something crucial: the gap between vision and execution is where startups die. I had unlimited ideas and zero ability to prototype them myself. Every experiment required convincing a developer, writing a spec, waiting for a sprint, reviewing the output. By the time we tested an idea, the market had often moved on.

I didn’t know it yet, but this frustration was planting a seed.

Kerkhoff, Corona, and the Decision to Go Independent

After DATAlovers, I joined Kerkhoff Group as Partner for Digitalization — bringing digital transformation to procurement. It was exciting work, and then COVID hit.

I won’t go into the details, but let’s say that the pandemic revealed differences in how organizations treat their people. I made a conscious decision to step away and go freelance. It was scary, it was liberating, and it gave me something I hadn’t had in years: time to explore.

CRIF — And The Itch That Wouldn’t Go Away

I landed at CRIF — first as a freelancer, then permanently as Executive Director of Product Management. CRIF is where I am today, shaping the future of business information and prediction across Europe. I also became active in Bitkom’s working groups on AI and Data Products and served five years on the plenary board of the Chamber of Industry and Commerce Rheinhessen.

On paper, my career was exactly where it should be: senior leadership, strategic influence, board-level engagement.

But that itch from the DATAlovers days never went away. The frustration of not being able to just build something when the idea hit. Of being dependent on development capacity, sprint priorities, budget approvals.

And then, around 2023, I discovered low-code. And everything started to change.

The Low-Code Gateway

Retool was my first real taste of building software myself. For those unfamiliar: it’s a low-code platform where you drag and drop components — tables, forms, buttons, charts — and connect them to databases or APIs.

It was revelatory.

Within days, I had built my first internal dashboard. It wasn’t pretty, and it did exactly one thing, but it worked. Something that would have taken a developer a few hours took me a few days — but the point is: I did it myself.

That feeling — of going from idea to working application without asking anyone for permission, budget, or sprint capacity — was addictive. After 25 years of designing systems on whiteboards and in PowerPoint, I could finally see my ideas come alive on screen.

Over the following months, I built several internal tools. None were revolutionary. But each one taught me something new about how software actually works — not theoretically (I already knew that), but practically.

A Note On „Vibe-Coding“

You’ll see me use the term vibe-coding throughout this series. Some developers find it dismissive — as if it trivializes real software engineering. I get that. And I want to be upfront about why I chose this word deliberately.

There’s another term floating around: citizen development or citizen coding. It’s the corporate-friendly version. It sounds like something from a Gartner report. It implies you’re a well-meaning amateur who builds small tools within guardrails set by „real“ developers.

That’s not what this is.

What I’m doing — what millions of non-developers are starting to do — is building production software. Full-stack applications with authentication, databases, payment processing, AI integration, and real users. Not internal dashboards. Not prototypes. Products.

Vibe-coding captures something that citizen development doesn’t: the creative, intuitive, conversational nature of building with AI. You describe what you want. You iterate. You follow the vibe. The AI handles the syntax; you handle the vision.

Is it the same as a senior engineer writing optimized code from scratch? No. Does the end user care? Also no.

I come from the non-dev world. I don’t pretend to be an engineer. But I refuse to call what I’m building „citizen development“ — because the results speak for themselves.

What Came Next

Around this time, the AI revolution was picking up speed. ChatGPT had launched, and suddenly everyone was talking about AI-assisted everything. I started experimenting with AI coding assistants, and what I found changed my trajectory completely.

But that’s a story for the next article.


This is Part 1 of a 7-part series about my journey from zero coding experience to building production applications with AI. In Part 2, I’ll talk about how Replit and AI-assisted development opened a door I didn’t know existed.

Follow me for the next installment, or check out what I’ve built:

  • 🧞 [CalendarGenie](https://calendargenie.replit.app) — Turn any screenshot into calendar events
  • ✍️ [Post-er](https://post-er.replit.app) — AI-powered content creation
  • ⚡ [Energy Backbone](https://energy-backbone.bigdataheaven-software.de) — Smart energy management
  • ⚓ [Marina Master](https://marina-master.bigdataheaven-software.de) — AI-powered marina management

Andreas Kulpa is Executive Director of Product Management at CRIF and a vibe-coder based in Wiesbaden, Germany. He’s been in tech for 30 years — but only started building software 2.5 years ago. Connect with him on [LinkedIn](https://linkedin.com/in/akulpa).