AI is the most significant amplifier of existing power structures in human history. In the hands of centralised authority - corporate, state, or algorithmic - it accelerates control, narrows the range of acceptable decisions, and bakes monoculture into the infrastructure of collective life.
It's happening now.
The question isn't whether AI will reshape how humans make decisions together. It will. The question is whether that reshaping happens to us or with us. Whether it concentrates decision-making power further - into fewer hands, fewer models, fewer approved framings - or whether it finally gives ordinary groups the capacity to govern themselves clearly.
We are building for the second possibility. That's what μ democracy is for.
Inversion
For most of human history, the argument for concentrated decision-making has been essentially pragmatic: people can only act on what they know, and knowledge has always been scarce, slow, and unevenly distributed. Hierarchies formed around information. Institutions emerged to manage the fact that most people, most of the time, couldn't see enough of the picture to be trusted with consequential decisions.
This wasn't always cynical. It was often just true.
But it produced a deep cultural myth: that ordinary people are not really capable of governing themselves at scale. That democracy is something you do periodically in a booth, not something you practice daily in the groups you actually belong to. That decision-making authority belongs, fundamentally, elsewhere.
AI inherits this assumption. Most AI systems are built on top of it, trained by institutions that encode it, deployed in ways that reinforce it. The result is not liberation - it's the existing hierarchy with better tools. Smarter surveillance. More persuasive control. Monoculture at machine speed.
But the assumption is wrong.
The facilitator bottleneck
There is a practical reason why genuine participatory decision-making has always been hard to scale: it requires skilled facilitation. Someone who can hold a group's attention on the actual question, separate tangled issues, surface what's really in conflict, and guide people toward a decision that is clear enough to be remembered, revisited, and built upon.
Good facilitators are rare. They take years to develop. There are nowhere near enough of them for the number of groups that need them.
This is not a small problem. It means that most of the organisations, cooperatives, communities, and committees that are genuinely trying to self-govern - trying to practice democracy in the real, everyday sense - do it badly. Not because the people are bad. Because the skill is hard and the support isn't there. So they fall into the familiar failure modes: circular conversation, the loudest voice winning, vague agreements that dissolve before the next meeting.
AI can change this. Not by replacing human judgment - that's exactly the wrong use - but by providing the coaching capacity that groups have always needed and never had access to. A way to ask the question a good facilitator would ask. To notice when two people are talking about different things. To say: you have three issues here, not one - which do you want to decide first?
That capacity, available to any group willing to use it, is genuinely new. It didn't exist before. And maybe it is actually a valuable use of AI. Maybe it could actually make us better?
A new paradigm, or the old one faster
The danger is not that AI is powerful. The danger is that we apply powerful technology inside the assumptions of a culture that has always treated decision-making as something to be managed and controlled from above.
When you do that, AI doesn't democratise anything. It just makes the existing structure more efficient at reproducing itself. It selects for the decisions that fit the model. It flattens the diversity of how different groups think, what they value, what tradeoffs they're willing to make. It produces the appearance of participation while the actual power concentrates further.
This is what we mean when we say the current trajectory amplifies centralisation. It's not a conspiracy. It's just what happens when you hand a powerful tool to institutions whose logic has always been to limit the scope of who gets to decide.
The alternative isn't to refuse the technology. AI is here. That argument is over.
The alternative is to use it with a completely different intent - to build the capacity of groups rather than manage their outputs. To make clarity and accountability accessible at the smallest scale, not just the largest. To treat the people in the room as the actual decision-makers, and AI as the tool that helps them do that job well.
What μ democracy is actually for
We believe democracy fails at the top because it has been abandoned at the bottom.
Not through malice. Through efficiency. We just ran out of time. Organisations streamlined their processes. The messy, slow, unglamorous work of actually deciding things together - surfacing disagreement, sitting with complexity, reaching genuine understanding - got optimised away. We got better spreadsheets. We got tyrants.
This didn't happen to us. We allowed it. By ceding the daily practice of collective decision-making, by treating democracy as something that happens elsewhere and above us, we hollowed it out from the inside. The structures remained. The capacity disappeared.
A decision log, on its own, is not the answer to this. Records without awareness are just bureaucracy. Transparency without understanding is just surveillance with better filing. The tools are only as good as the practice they serve - and the practice requires something harder than software: a genuine reckoning with how decisions actually work. Who frames them. Who they include and exclude. What gets treated as settled before anyone speaks. How power moves through a room before anyone calls a vote.
AI cannot give us that awareness. Nothing can give it to us. But it can hold a space open for it - slow a group down long enough to ask the questions that get skipped, name the thing that everyone is circling, make visible the structure of what's actually happening.
That's what we're building toward. Not a logging tool. Not a governance platform. A practice - of clarity, of accountability, of genuine collective agency - that people can actually develop, in the groups they already belong to, starting now.
Democracy doesn't need to be reclaimed from governments. It needs to be reclaimed from our own habit of giving it away.
That habit can be broken. We are learning how.