Start somewhere that has nothing to do with politics. A dairy co-operative with a dozen branches has to change its payout policy. It runs a postal ballot: every member gets a paper, ticks yes or no, and the papers are counted. Fifty-one per cent carries it. The other forty-nine per cent, and every reason anyone gave for being in it, vanish from the record the moment the count is announced. A year later, when the policy bites and someone asks how it was decided, there is a number and nothing underneath it.
That is the ordinary shape of deciding together across an organisation too spread out to sit in one hall. A union with worksites up and down the country ratifying an offer. A professional body revising its code of ethics across regional branches. A company adopting one AI-use policy over a dozen subsidiary boards. A church network putting a property question to every parish. Different rooms, same problem: a decision has to be made across many rooms, the rooms cannot all meet, and the usual instruments — a ballot, a roadshow, a consultation summarised by the secretariat — keep the numbers and throw away the reasoning.
The machinery in a citizens’ assembly was built for exactly this shape. It is worth saying plainly, because it is easy to miss under the national headline: the assembly software was never political software. It is what deciding-together looks like when you refuse to average. Strip the politics off it and what is left is general.
The whole of it turns on a single refusal. When many rooms run the same question and their results are drawn together, the map may never manufacture agreement by blending dissent away. It shows where the rooms converge, where they divide and along which lines, and what holds up everywhere against what is only local. A position one branch holds and no other is surfaced, not buried. A split inside a room is carried up into the whole picture verbatim, attributed to the room that held it. Nothing in the machine is allowed to collapse a disagreement into a number.
That refusal is the product, and it is politics-blind. It does not know whether the rooms are party branches or co-op branches. It knows only that they disagreed, that the disagreement should survive, and that the whole thing should be provable. Every record is sealed on sovereign New Zealand and EU infrastructure — no US-owned or China-owned cloud in the stack — write-once, its signature checkable offline by anyone, without trusting the platform. What came from a rehearsal and what came from real people is never blurred together.
So the useful way to name the capability is not “a platform for organisations”. It is narrower and sharper than that: provable, dissent-preserving decision-making across rooms that never sit in one hall — for any group whose disagreements deserve to survive the minutes.
A multi-branch co-op or credit union deciding a constitutional or payout change. Today the losing branches disappear into the tally. With this, each branch deliberates, its dissent stays attributed to it, and the consolidated result is provable to every member — not a number they are asked to trust, but a map they can read down to their own branch.
A union ratifying an offer across worksites. The permanent residue of a big ratification is the suspicion that the leadership stitched it up. Sealed, write-once records the leadership cannot edit — with each site’s dissent legible in the national result — turn that suspicion into something a member can check.
A board adopting a group policy over its subsidiaries — an AI-use policy, say. The engagement between each director and the tool is signed on the director’s own device and sealed; each subsidiary’s dissent is preserved in the group record. The minutes stop being written after the fact and start being provable at the time.
None of those is a hypothetical about a different product. They are the same machine at a lower setting than a nation.
| The room | How it is decided today | With sealed, dissent-preserving records |
|---|---|---|
| Multi-branch co-op or credit union | The losing branches disappear into the tally | Each branch’s dissent stays attributed to it; the result is provable to every member, not a number they must trust |
| Union ratifying an offer across worksites | The residue is a suspicion the leadership stitched it up | Write-once records the leadership cannot edit, each site’s dissent legible in the national result |
| Board adopting a group policy over subsidiaries | Minutes written after the fact | Signed on each director’s own device and sealed; each subsidiary’s dissent preserved — provable at the time |
If the general story is “any group that has to decide together”, the citizens’ assembly is that story at its most demanding: national stakes, rooms split along party lines, obligations to tangata whenua, and scrutiny that is actively looking for the seam. A tool that holds up there holds up in an AGM. That is why the democracy piece is the flagship and not a footnote — what survives a nation’s argument will survive yours. The full version of that argument, including how a federation of assemblies could produce a national read without a referendum’s uninformed yes or no, is set out in the companion piece, One question, many rooms.
Two things, kept apart on purpose, because running them together would be the same overclaim this whole approach exists to avoid.
The single-room version is real and usable now. An organisation can run a decision through the platform today — a committee or board decision recorded with maker-checker and signed engagement, a consent or ranked or quadratic poll, a deliberation sealed as a tamper-evident record — and walk away with something provable. That part is built, deployed, and in use.
The many-rooms version — the consolidation that draws a federation of sealed results into one dissent-preserving map — is built and has been validated end to end against our own live governance infrastructure: real synthesis, real seals, real offline verification. It is not running in production for any organisation yet. It can be demonstrated; it is not in anyone’s hands today. Anyone who tells you otherwise is selling.
And it decides nothing, at any setting. The output is a contestability map — where the rooms converged, where they split, where the evidence is thin, what people should stress-test — never an answer. Where a question touches tikanga or Māori data it routes to real people; a synthetic voice is barred from standing for tangata whenua, a refusal in the code, and that bar is provisional, subject to a confirmation from tangata whenua we have not yet had. The machine’s whole job is to make the human decision better informed and the record of it verifiable.
A ballot keeps the count and loses the reasons. A single room keeps the reasoning and cannot reach the numbers. Many rooms on one question, drawn together without any of them averaged away and sealed so the whole picture can be proven and read down to the last dissent, keep both — and nothing in that is specific to a country. The rooms can be branches, worksites, parishes, subsidiaries. The country is just the biggest room.
Same machine, different rooms. People deliberate and decide. The record proves what was argued.