Scaling AI in Production: AIOps, Architecture, and Human Oversight, Tue, Apr 7, 2026, 6:00 PM | Meetup
We are thrilled to welcome you to our 9th edition of AI Native Netherlands, hosted at the Miro headquarters in Amsterdam.
We look forward to seeing you again for another e
What I won't be doing is showing up with a neat framework and a confident answer. Honestly, nobody is a deep expert here yet, myself included. What I do bring is a healthy dose of skepticism and experience with software design, architecture, collaborative modelling and cognitive bias. My plan is to make this a conversation rather than a lecture, exploring how we keep architecture intentional and human-led when the tooling is moving faster than our understanding of it.
>>>
My session is called "Facilitating Software Architecture While AI Agents Write the Code." The core tension I want to explore: developers have always made implicit architectural decisions, but AI acceleration means those decisions happen faster, and increasingly get delegated. When AI generates code based on incomplete context, assumptions get baked in as if they were intentional, and before long, nobody is quite sure why the system looks the way it does.
>>>
The speed at which AI agents write code is no longer the bottleneck. Whether we understand what that code is actually doing, that's where it gets interesting.
On 7 April I'll be speaking at the AI Native Netherlands meetup at the Miro offices in Amsterdam.
>>>
The advice process doesn't give you worse decisions. It shows you the conversations you were never having.
#SoftwareArchitecture #DomainDrivenDesign #ArchitectureDecisionRecords #CollaborativeModelling #TeamTopologies
Separating information (the facts) from preferences (the opinions) is a move that only becomes possible when things are written down and transparent.
And if you can't have that conversation, if you can't dig into what someone's preference is really about, that's a different problem: the skill of having crucial conversations. No process fixes that for you.
>>>
Because opinions aren't the problem. Opinions packed in pseudo-rational framing are. When someone says "this approach doesn't fit our architecture" but what they mean is "this makes my life harder and I'm worried about my team's workload", those are two very different conversations. One stays abstract. The other gets somewhere.
>>>
This is one of the things collaborative software design is fundamentally about: making implicit knowledge explicit. The opinions, the preferences, the emotions behind a technical stance, they don't disappear when you ignore them. They just go underground and shape decisions anyway, invisibly.
>>>
When you write it in an ADR, something shifts. "I don't want X because I don't like working that way" suddenly sits there, visible, inviting a different kind of conversation. You can slow down. Move to system 2. Ask: what's the actual need behind this preference?
>>>
Those opinions were always there. They just lived in meetings, rarely written down, usually processed through system 1 reactions that reinforced existing power structures and gave you the same outcome you always got. The loudest voice won, or the most senior person's preference quietly became "the decision."
>>>
