Skip to content
Rutger Hemrika
Menu
← Work

Own initiative, on the Dutch municipal market · 2026

A five-year municipal transformation blueprint, drafted with AI agents

A complete architecture and product-strategy package on Common Ground and GEMMA, produced in days, and what that taught me about where the real work moves.

20 layers: 13 technical, 7 product managementY1Y3Y5five-year horizon25 capabilities, responsible AI staged from day onedrafted with AI agents, every claim validated by hand
Twenty architecture layers and twenty-five capabilities across a five-year horizon, drafted with AI agents and curated by hand.

The experiment

Could AI agents produce a credible, end-to-end strategy and architecture package for Dutch municipal digital transformation, and how much of it would survive scrutiny? I directed the exercise in February 2026, using the Dutch municipal reference architecture GEMMA and Common Ground as the frame.

What came out

  • A target architecture with twenty layers, thirteen technical and seven for product management.
  • Twenty-five capabilities, each with owner, dependencies and maturity path.
  • A five-year roadmap with year one, three and five horizons, SAFe portfolio structure, eleven personas and playbooks.
  • Governance, service-level and data-processing templates.
  • A development starter: API specifications, tenant-isolated database migrations, policy-as-code and deployment templates.

The responsible-AI capability sits in the model from day one: human oversight of automated decisions, bias monitoring, explainability and EU AI Act conformity, with AI maturity staged deliberately over the five years rather than promised in year one.

What I learned

Breadth is now cheap. A package that would have taken a team a quarter took days. What did not get cheaper was judgement: validating every claim, catching the places where generated text was confidently wrong, and deciding what to cut. The work of a strategist moved from writing to curating, and the quality of the result depended entirely on how hard I pushed back.

This remains a speculative blueprint, not a shipped programme. Its value to me is the method and the argument it sharpened about what municipalities actually need from a platform.