I was a product owner. I took a consulting job on purpose, to get into AI. A year and two months later I received my MVP award from Microsoft.
That is the whole story in three sentences, and it is how I tell it when someone asks. It also skips everything useful. So here is the longer version, in dated steps, taken from my own site so I can't round anything up. I picked the beats that mattered, not every line of the CV.
1. It started in a chat window
In June 2021 I joined Assura in Lausanne as a Digital Insurance Advisor. The job was LiveChat support, in French, English and German. Customers typed, I answered.
In April 2023 I became Product Owner for the chatbot there, and Testing Lead. So I went from answering the chat to owning the thing that answers the chat. Vision, roadmap, the testing, the whole lifecycle from idea to training the model. My site lists what came out of it: a 25% increase in CSAT and a 43% sales uplift.
2. The move: consulting, to get into AI
In April 2025 I left the product owner job and joined Witivio as an AI consultant. I went in as a junior. That was the point. As a product owner I owned one product. As a consultant I would work across many clients and many agents. That was the whole reason for the move: to get into AI.
Here is the path from there, on one line. Dark dots are the job. Orange dots are things anyone could see.

3. The first thing people could see wasn't mine to publish
Six months into the job, in October 2025, Microsoft published a Customer Story featuring BuyerCompanion, a procurement agent built on Copilot Studio for a global energy company. I co-led its architecture, development and delivery. It automates the RFP process, from drafting specifications to identifying suppliers.
That work was visible because Microsoft and the client chose to talk about it. Most client work never gets that. It stays behind an NDA, which is right for the client and invisible to everyone else.
So if I wanted people to see how I work, I had to make things that were mine to share.
4. November 2025 to May 2026: putting things in public
This is the part that I think actually moved things. None of it was planned as a campaign. It was one thing, then the next.
- Guides, November and December 2025. Two playbooks on LinkedIn (an AI onboarding assistant on November 5, a framework for AI assistants on November 12), then a long one on building a Mermaid diagram agent on December 11.
- First talk, December 4, 2025. The Microsoft 365 & Power Platform Community Call, on generating Mermaid process diagrams with Copilot Studio. My first community call.
- Second talk, February 18, 2026. The Power Platform community call, on feature driven development with the VS Code extension for Copilot Studio. The same month I became Team Lead, ten months after joining as a junior.
- Prompt of the Week, March 31, 2026. Microsoft picked my "Prompt Architect" prompt from the Copilot Prompt Library for a shoutout.
- Third session, April 2026. A Microsoft Community Learning session on letting any agent explore Power Platform solutions through an MCP server.
- A decision framework, May 14, 2026. Posted in the Microsoft Tech Community: how we choose between Agent Builder, Copilot Studio and Azure AI Foundry on real client work.
- A newsletter, May 2026. Agentic Weekly, edition 001 covering the week of May 4 to 10.
- Another prompt, May 26, 2026. My "Use Case Qualifier" system prompt featured by Microsoft.
A prompt is a small thing to be thanked for. It still counts, because someone else can pick it up and use it the same day.
5. June 1, 2026
On June 1, 2026 Microsoft awarded me the Most Valuable Professional title for 2026-2027, in the M365 Copilot and Copilot Studio category. Here is the card as it sits on my site today.

I won't pretend to know how the program weighs each thing. The award text on the certificate talks about community leadership and sharing real-world expertise, and if I look at my own list, everything that fits that description happened in public. The customer story made a nice headline. The talks, prompts and guides are what someone could actually open and use.
6. What I kept doing after
The award didn't change the routine much. It mostly made it bigger.
- Talks kept coming. Community calls on process mining (June 17), an agent that governs other agents with the code on GitHub (July 23), an MCP server for debugging Power Automate flows (September 16) and what an agent can't read (September 24).
- A show. On August 11 I was a guest on the M365 Show, and then I became a host and commentator on it with Mirko Peters.
- A community. I founded the Copilot Studio Hub Discord and run the biweekly Copilot Studio Clinic, a free open session where anyone brings a real problem and we work through it live.
- More open source. On August 26 my MCP Server Review Skill was picked as Copilot Skill of the Week and published in the PnP copilot-prompts repository.
- A keynote in German. On September 18, at M365Con D-A-CH, on choosing between Agent Builder, Copilot Studio and Azure AI Foundry. That one is the Tech Community post from May, grown up.

Counting the talks page end to end, that is ten sessions between December 2025 and September 2026.
7. What I'd point at, if you asked me
I'm not going to hand out a formula. I don't have one, and the program is Microsoft's to describe. But if a colleague asked me what I did differently in those fourteen months, this is what I'd say.
| What I did | Where people could see it | Why it helped, I think |
|---|---|---|
| Turned project problems into talks | Community calls, Dec 2025 onwards | A recording keeps working after the call ends |
| Shared prompts and skills, not just posts | Copilot Prompt Library, PnP repo | People can reuse them the same day |
| Wrote the long version | LinkedIn guides, Tech Community | Depth is what people bookmark |
| Wrote every week | Agentic Weekly, from May 2026 | Showing up weekly beats one big moment |
| Put the code next to the talk | GitHub | A demo you can clone is a demo you can trust |
The common thread is boring. Everything on the list was useful to someone who has never met me, and a lot of it came straight out of client work: the decision framework was built on real engagements, and the German keynote says so in its subtitle, "aus echten Projekten". The job gave me the problems. Sharing them is what made them visible.
I still do the same things I did before June: a problem at work, then a talk, a prompt or a repo about it. The award is on the about page now. The routine hasn't changed, and I'm keeping it.




