What Makes a Person Agentic?
Agency is the ability to notice, choose, and act. What changes when AI can strengthen — or quietly replace — each part?
I’ve been trying to name something I keep noticing.
In coaching sessions, team retrospectives, and strategy workshops, AI can make people dramatically more capable. It can hold context, surface patterns, and turn an unclear intention into a concrete next move.
But capability is not automatically agency.
Agency is the ability to notice what is happening, make a meaningful choice, and act with consequence. AI can strengthen all three. It can also short-circuit them: noticing becomes summarising, choosing becomes accepting, and acting becomes pressing run.
Agentic People is my attempt to think out loud about that boundary. Not how to make AI more autonomous, but how to help people remain authors as machines become more capable.
The frame that keeps helping me is sociometry: the study of who relates to whom, and how. When AI enters a group, it introduces new choice patterns. Athletes test their reflections with an assistant before talking to a coach. Teams ask a model to interpret a meeting before they compare their own interpretations. Leaders use AI to widen the options available to a group — or to narrow them before the group ever sees them.
These aren’t just efficiency moves. They’re changes in the structure of the group.
The structure of agency is changing: who gets to frame the problem, whose interpretation becomes legible, who chooses the next move, and who remains responsible for it.
I’m not making an optimism argument here, and I’m not making a pessimism argument. I think both framings get in the way of noticing what’s actually happening.
What I’m trying to do instead is be curious about the specific dynamics. Not “is AI good or bad for coaching?” but “does this reflection help the athlete hear themselves, or replace their own interpretation?” Not “does AI help organisations?” but “who gains the ability to act, and who becomes a recipient of generated decisions?”
These are group dynamics questions. They’re facilitation questions. They’re systems questions. And they happen to require some understanding of how large language models actually work.
This site is where I work through them. I come at it from a few directions: 16 years coaching runners, years working on org design and complexity at The Adaptavist Group, and a fair amount of hands-on AI building — pipelines, tools, home systems, this site itself.
None of those domains give me the answer. But each one gives me a different lens on the same question.
What helps people become more agentic with AI? I don’t know yet. But I think it is the right question to be asking.