Not the tool. The decision that's actually stuck — a bottleneck, a risk, a process nobody owns. I design AI systems the way I've always worked: from what's breaking, backward.
"A five-minute delay isn't a delay. It's 200 missed connections by dinner."
Airline departure control systems — real-time decisions, zero tolerance for "we'll fix it later." A late gate agent isn't a minor inconvenience.
Building what sits behind a training platform. Different industry, same question: where does this actually break under real volume?
Same job, new system. Most businesses treat AI as something separate to try. It isn't — it's the operational work I've always done.
Pick a tool. Hope a use case appears. Retrofit it onto a process nobody re-examined.
The decision that's actually stuck — and design the system backward from there.
Name the real business decision under risk — not the tool. What's actually at stake if this stays broken.
Find where the system actually breaks — the cause, not the symptom: missing structure, missing ownership, missing priority.
Architect the AI system around that decision — what's automated, what stays with a human, who owns it.
Build it, hand it over, and settle who owns the infrastructure it runs on — before that becomes what stalls it.
A working conversation to name the real decision at stake — not a tool pitch.
Where the process actually breaks, and where AI creates the most value.
The architecture itself — decision logic, ownership, what a working pilot looks like.
Built out, handed over — a system your team actually owns and runs.
A custom AI system that flags contract risk under Kazakh commercial law for an industrial company's legal and operations team.
A validated AI assistant that answers strictly from a company's own approved course material — and says "I don't know" instead of guessing.
A concept package for a studio in Germany — real photos, a fixed budget, and a clear line between what's realistic and what isn't.
An AI agent that sorted a live inbox of 3,000+ emails — batch-verified, zero errors.
A full architecture for tracking AI subscriptions and one-off charges across mailboxes — live and running, human-in-the-loop by design.
Weekly research and positioning checks that run this practice — the same kind of system I design for clients.
SMB owners, founders, and operations leaders who have an AI pilot that stalled, or a process that's outgrown a spreadsheet.
Where — in person across the Netherlands, remote across Europe.
Helps SMB owners and operations leaders make the specific business decision that's stuck — which process to redesign, what to automate, what stays with a person — then builds the AI system around that decision. Not a generic AI workshop; a working system tied to one business outcome.
Tools are the last step, not the first. Most engagements start with a decision that's already broken — a bottleneck, an unowned process, a risk nobody's pricing in — and the tool gets chosen once that decision is clear. Work spans strategy, operations, knowledge management, sales, and client experience, not only workflow automation.
SMB owners, founders, consultants, experts, and operations leaders across the Netherlands and wider Europe — typically businesses with an AI pilot that stalled, or a process that's outgrown a spreadsheet and one person's memory.
With the OLDS Framework: Observe the real decision at stake, Locate where the system actually breaks, Design the AI system around that decision, Systemize it so it runs without ongoing involvement. Most relationships start with a paid discovery phase before any build work.
In person across the Netherlands, and remote across the rest of Europe.
Fifteen years fixing operational systems before AI entered the picture. The starting point is always the business decision under risk, never the tool — that discipline is the OLDS Framework.
"If your team has an AI pilot that stalled — let's talk."