Agency in an Accelerated World
As AI systems become faster, more capable and more integrated into the real world, we are entering an era where managing complexity is not just a technical challenge. It is a human one. Séb Krier’s recent essay offers a starting point for understanding what it will take to retain meaningful agency in a world shaped by increasingly autonomous systems.
Trust without total understanding
Krier argues that agency does not require humans to understand every detail of an AI system. Trying to do so can quickly become counterproductive. We already see this trade-off today: using automated cancer-detection tools makes little sense if we insist on manually reviewing every scan. At some point, a system performs well enough to be trusted. That trust, however, must be earned, monitored and supported by the right oversight structures.
Coordination across cognitive scales
This is where organisations and institutions come in. Today’s world is not built to drop AI into workflows at scale. We lack the mature techno-organisational ecosystems needed to account for tacit human knowledge—the kind that underpins everything from hospital procedures to supply-chain operations. Integrating AI effectively means designing for human–machine coordination from the ground up.
That coordination will be most sustainable when humans and AIs both have a role in decision-making. This does not apply only within organisations; it extends to the structure of society itself. Governance, verification systems and trust-based mechanisms will be essential between humans and AI systems operating at different cognitive scales.
Cognitive impedance matching
AI systems will be able to react, adapt and act orders of magnitude faster than we can. Managing that speed differential will require what Krier calls cognitive impedance matching: systems that translate between AI activity and human strategic oversight.
Imagine an AI managing a global supply chain in real time. A person could not track every shipment or micro-adjustment. But the right impedance-matching interface might provide curated summaries, forward-looking alerts and simple levers for shifting strategy without losing control.
From hands-on control to structural influence
As with previous technological shifts, we will need to rethink governance itself. Industrial complexity gave rise to modern bureaucracies and market mechanisms. The speed and scope of AI—and, increasingly, AGI—will require us to automate parts of governance too.
This is not the end of human agency. It is a shift in how and where agency operates. The challenge is not to preserve low-level understanding of every action, but to design abstractions that reflect what we care about and remain adaptable over time.
Those abstractions will be difficult to design in advance. They will need to evolve in response to human input for far longer than many assume. The deeper the system, the more important it becomes that humans retain the ability to set direction, even if we are not steering every turn of the wheel.
The question is no longer whether AI will outperform us in narrow domains. It already does.
The real frontier lies in how we respond. Agency must evolve from hands-on control to structural influence: setting objectives, defining the contours of oversight and embedding adaptability into every control plane.
This is not a retreat from responsibility. It is a redesign.
Read Séb Krier’s original essay: Maintaining Agency and Control in an Age of Accelerated Intelligence.