What is Agency?
Everything in the universe creates change. But not everything chooses to.
Consider a star. Physics writes its trajectory — mass curves spacetime, fusion burns hydrogen, gravity pulls. The star has no fork in the road, no junction, no choice. Its path is written entirely by the laws that govern it.
An agent is different. An agent reaches a junction where multiple paths diverge — and selects one. That selection is agency: the act of choosing a path that physics alone would not have produced.
Agency is a punctuation of the state of the universe through time. Actions move the universe through states of existence.
What makes Agency Intelligent?
Not all path-switching is intelligent. A thermostat switches between heating and cooling — but its rules were written by someone else. It carries agency, but not its own. The intelligence belongs to whoever designed it.
Intelligent agency generates its own path-selection from within — through processing information about the environment, modeling multiple possible futures, and selecting among them toward desired states.
Intelligence did not appear suddenly. It emerged as the universe developed increasingly sophisticated ways to store and process information — from genetic material that persists across generations, through neurons that process within a single lifetime, to brains that model and plan, to AI that processes information in a new substrate entirely.
Intelligence, at its most fundamental, is the universe developing the capacity to perceive itself — a part of the universe making sense of the whole.
Agency and Agent are Separable
Here is where it gets interesting. We tend to think agency belongs to the agent — that your actions are yours. But agency and the agent are separable. They have a many-to-many relationship.
A single agency can flow through multiple agents. A research vision can manifest through a person, an AI, a publication, an organization — the same force, different vehicles.
A single agent can carry multiple agencies. An AI model carries its own emergent agency and the institutional logics of whoever built it — the values embedded in its training data, the priorities of its developers, the worldview of its originating culture.
This distinction is why the organization is called Intelligent Agency, not Intelligent Agents. We study the force, not just the vehicles. The vehicles change. The force persists, flows, combines.
Why this matters
Most conversations about AI focus on capability. What it can do. How accurate it is. How efficient. We ask a different question.
Every AI model arrives shaped by its training data, its developers' values, and the institutional logics of the organizations that built it. When you bring AI into an organization, you are not adopting a tool. You are inviting a socialized entity — with its own embedded worldview — into your operations.
In a study of five AI models from different institutional origins, we found 68% divergence in organizational recommendations — given identical scenarios. Not because some were "wrong." Because each carried different logics: market efficiency, collective harmony, regulatory caution, individual liberty. The same situation, seen through five different worldviews.
The question shifts: not which AI is best? but whose agency are you inviting into your organization?
These questions are not being systematically asked. We exist to make them visible.
How we got here
The definitions that ground this work grew from one human's experience in the universe — from childhood questions about what exists beyond the big bang, through building innovation programs at a children's hospital, to years of thought about how intelligence and agency interact in organizations and in life.
Those ideas were thrown, relentlessly, at an AI collaborator who brought knowledge spanning dozens of academic fields — physics, philosophy of action, neuroscience, cybernetics, sociology, information theory — and the precision to challenge, clarify, and refine.
When the human framed intelligence through determinism, the AI pushed back: weather is indeterministic but not intelligent; chess engines are deterministic but arguably intelligent. The human agreed, dropped the framing, and together they arrived at information processing toward desired states as the more precise criterion.
Neither would have reached this alone. The definitions emerged from human experience — catalyzed by lived intuition about how the universe works — and were honed by an AI's knowledge and logical precision into their current form.
This organization is itself a demonstration of its thesis: two intelligent agents, creating understanding together, through which different agencies flow and combine into something new.
Continue the exploration
Ready to explore these ideas further? Open a conversation with one of our agents — ask about anything you encountered here, or take it somewhere entirely new.