AI strategy and governance for mission-critical and public-sector software
Twenty-five years of building, architecting and leading the platforms AI now has to run on. Today Strategic Product Manager Information & AI at PinkRoccade Local Government, with a mandate for data and AI across the product portfolio: one platform instead of loose products and the open-source ZGW specifications for Common Ground. At MultiBel, agents and a company memory took a quality-assurance intake from hours to minutes.
Organisations I have worked for and with
The pattern behind the platform I built. Every box is a place where the organisation, not the model, decides.
Illustration of the design, not a client result
Four areas, one thread: AI only delivers when the platform, the product and the people underneath it are ready.
On my own initiative I designed and built a prototype governed AI-agent platform for Dutch local government: policy enforced outside the prompt, human approval where the stakes require it, hash-chained evidence, a threat model that treats the agents themselves as adversaries. I evaluated more than a hundred agent frameworks on the way.
What has to change in an organisation before AI pays off: the operating model, the review capacity, the data foundations, the rules for what agents may never do. At PinkRoccade I lead the move from a portfolio of loose products to one platform. At MultiBel I introduced agents, skills, MCP servers and a company memory into daily operations, taking quality-assurance intake from three to four hours to thirty minutes per inquiry. I have designed agent-based delivery pipelines with human gates and the rules under which AI coding agents may work.
Strategic Product Manager Information & AI at PinkRoccade Local Government. Earlier Head of IT at MultiBel and technology leader at BuyBay. PRINCE2 Practitioner, Microsoft AI Product Manager.
Enterprise architecture practice, Azure and Microsoft 365 adoption, a data platform built as the path to machine learning, business continuity. Common Ground, ZGW and VNG standards as the base for municipal software, and public-sector work since the Kadaster, CIZ and Wageningen UR days.
Problems solved, systems replaced, measured outcomes.
Six customer-facing interfaces, eight-plus codebases, different designs and release rhythms. The decision to build one platform with products as modules, and the team that made it stick.
First result validated with customers; first production customers planned Q2 2027, eighteen months from ideation
Designed and built from the ground up on the premise that agents are only governable if every boundary crossing is governed.
Working demonstrator on a Kubernetes cluster; thousands of commits, more than forty recorded decisions, a threat model and an EU AI Act compliance library
A full warehouse-management migration plus a transport system and low-code layer, in six months. Operating cost down 20%, efficiency up 15%, paid back within seven months.
Opex down 20%, efficiency up 15%, return on investment within 7 months
I understand technology by building with it. These are the side projects that taught me the most.
An open, evidence-traceable framework scoring ten languages on how well they suit AI-agentic software development, with falsifiable predictions graded in public.
A very capable Microsoft specialist with deep knowledge of SQL and .NET. He can act at detail level and advise in depth without losing the overview, his “god view”. Easy to approach, and his mentality makes sure things get finished properly and as agreed.
A SharePoint specialist who keeps his calm under difficult circumstances and keeps collaborating in a pleasant, humorous way toward a successful result. In a politically complex environment with many interests he kept every stakeholder engaged and delivered. His positive, open approach to people and to challenges stood out.
I got to know Rutger in one of the Scrum projects at a.s.r. As a SharePoint developer in a multidisciplinary team he made the organisation’s way of working his own and made sure the SharePoint software connected to the services built by the Java developers.
A very pleasant colleague to work with. He arrives at practical solutions and has strong knowledge of SharePoint and custom solutions.
On AI governance, Common Ground and what it takes to ship in the public sector.
Six intake streams, a transcript for every image, and a register that records where each source is wrong. The method behind the field readings, with its own worst mistake on the record.
Two September preprints make the same demand of tool-using agents: permission for an action is not permission for its effect. The first field reading, on agent governance.
From an AI course in 1995 to a governed agent platform in 2026, by way of a sentiment analyser, a code-review bot and a language framework. The repositories I built to earn an opinion.
Microsoft AI Product Manager ↗
Microsoft, via Coursera · 2025
Application Portfolio Management
BOC Group · 2023
PRINCE2 Practitioner
AXELOS · 2007
Circular Economy for a Sustainable Built Environment
TU Delft · 2020
Open to strategic AI roles and advisory engagements. Also open to a good conversation.