Science management is not administration. It is not grant chasing. It is not herding academics toward deadlines. These are the artifacts of management, not its substance.
The substance is this: creating the conditions for discovery that would not otherwise exist.
At a Dutch university in 2026, this question carries specific weight. The national research infrastructure landscape is being reshaped, by new large-scale funding programmes, by European coordination roadmaps, by the quiet revolution of AI that is changing not just what we research but how we research. As someone who sits at the intersection of infrastructure, strategy, and institutional coordination, I find myself asking: what does it actually mean to manage science well?
The Infrastructure View
Most people think of research infrastructure as equipment: microscopes, cleanrooms, compute clusters. But infrastructure is also coordination. It is the agreements between universities to share access. It is the data standards that make results reproducible. It is the career paths for technical staff that keep expertise from walking out the door.
Managing science means seeing the invisible scaffolding that makes research possible. A new microscope is a visible win. A shared data management policy is invisible, until it isn't, and then its absence is catastrophic.
Good infrastructure is invisible. Great infrastructure makes new kinds of science inevitable.
The AI Paradigm Shift
AI is not a tool you bolt onto existing workflows. It is a shift in the epistemic regime of science. When a language model can synthesise literature across 10,000 papers, when a vision model can read a microscope image faster than a trained human, when an autonomous lab can design and execute experiments without human intervention, the role of the scientist changes.
And therefore the role of the science manager changes too.
The manager's job is no longer just to procure the equipment and hire the people. It is to design the interface between human curiosity and machine-driven discovery. How do you build a self-driving lab that accelerates rather than replaces? How do you evaluate a model not on a benchmark, but on whether it produces better science? How do you redesign a career path for a researcher who spends half her time training models?
These are not technical questions. They are organisational design questions dressed in technical clothing.
Acceleration Is Not Enablement
A recurring observation: we confuse making science faster with making science possible. A better GPU accelerates an existing simulation. A shared facility enables a line of inquiry that had no home. These are different goods.
One programme I work closely with is interesting precisely because it tries to do both. It funds infrastructure (acceleration) but also the coordination layer (enablement). The question is whether the coordination layer gets the attention it deserves, because it is harder to measure, harder to defend in a budget meeting, and harder to claim credit for.
Managing science, in this view, is the art of defending the invisible. It is building the coordination layer and making it look effortless. It is connecting people across organisational boundaries not because someone told you to, but because the science demands it.
What I'm Learning
I did not arrive at this understanding through theory. I arrived through practice: through the daily work of designing new research infrastructure, of coordinating cross-institutional networks, of managing lab operations. Every stakeholder meeting, every grant deadline, every equipment purchase order has taught me something about where the leverage actually lies.
Three observations that structure my current thinking:
First, leverage is in the connections, not the nodes. The most valuable thing a science manager does is connect people who should be talking. The national lab that doesn't know the university has the same capability. The PI who doesn't know the funding call exists. The technician whose expertise could solve a problem three floors up. These connections are not automatic; they require intentional infrastructure.
Second, the AI paradigm shift rewards generalists. The specialists (the domain experts) are essential. But the people who can see across domains, who can translate between the language of materials science and the language of machine learning, who can spot the pattern in a funding landscape that looks like noise to everyone else: these are the people who will shape what comes next. Science management is generalist work.
Third, institutions are slow but durable. A startup can move fast. A university moves at the speed of shared governance. This is frustrating, but it is also a feature. The infrastructure we build today, if we build it right, will outlast the funding programme, the political cycle, and the hype wave. The question is whether we have the patience to build for durability rather than impact.
The Work Ahead
Managing science at a Dutch university in 2026 means holding multiple tensions at once: national ambition vs local reality, acceleration vs enablement, the urgency of AI vs the patience required for institutional change.
I don't have a unified theory. I have a practice, a set of questions, and the privilege of working with people who care deeply about getting this right.
This site is a place to think out loud about what I'm learning. Essays, observations, signals worth sharing. If any of it lands, I'd welcome the conversation.