KeizAI The economics
The boundary of the firm is a variable.
Coase asked why firms exist at all, and answered: because using the open market costs something — discovery, negotiation, coordination, monitoring, enforcement. When AI and structured records materially reduce those costs, the efficient boundary of the firm becomes smaller and more fluid. A temporary network can carry what once required a permanent agency, studio or general contractor.
Coordination, not introduction, is the product.
A company's margin was never paying for a role list. It was paying for the functions around the work: producer judgement, project management, quality control, accountability, records. Merely disintermediating a company leaves the buyer with a chaotic group of freelancers and the same costs, unabsorbed.
KeizAI is designed to replace those functions, not to skip them: AI carries the coordination — decomposition, candidate search, explanation, chasing, records — while every act of judgement stays with a named person. The role list is cheap; any modern model can enumerate the roles a music video needs. The connection path — who will actually do this, vouched for by someone who is prepared to stand behind it — is scarce, and that is what the system preserves.
This is stated in the project's own charter as a testable thesis, not a marketing assumption: KeizAI does not disprove Coase, it changes his variables — and the value is only real if the coordination is genuinely absorbed.
A budget you can believe is a budget that cost something to declare.
Recursive sourcing has a failure mode: a participant several referrals deep cannot tell whether the money exists to pay them, or whether someone upstream is taking an undisclosed cut. KeizAI's design answers with a declared envelope: one budget, declared by the client in one currency, that every participant can see in full — alongside their own share and nothing of anyone else's.
The declaration is designed to cost something, which is what makes it credible. A project is planned and staffed for free, but becomes active — engagements acceptable, work items openable — only when a management fee, carried inside the declared envelope, is settled. Raising the budget means settling the corresponding fee on a deadline; lowering it refunds nothing. Declaring a number you cannot fund costs real money, and declaring one too low costs the same money later. Neither direction is free.
If the fee on an increase goes unsettled, the project is deactivated — and the deactivation is visible to every participant. If a budget is in trouble, the people relying on being paid from it are the first to know, not the last. Through all of this, KeizAI never takes custody of the budget itself: settlement between actors stays between the actors.
- Declared
- One envelope, one currency, visible to all participants
- Your share
- Visible to you; no one else's is visible to anyone
- Activation
- Gated on settling the management fee inside the envelope
- Arrears
- Deactivation, visible to every participant
- Custody
- Never — settlement stays between the actors
Nobody works inside someone else's margin.
A KeizAI project is not one buyer and many workers. It is a forest of independent bilateral engagements: each one has its own seller and buyer, its own terms, its own currency, its own tax treatment and its own invoice numbering. An invoice always has exactly one seller and one buyer, and no total ever consolidates a chain — there is no netting, no set-off, no pass-through.
The design's core claim about billing: an invoice is a consequence, not an assertion. Every chargeable line must point back to accepted work and the evidence supporting it, and an accepted milestone cannot be billed twice. What a producer marks up between the engagement above and the engagement below is commercially sensitive, and stays confidential by default.
The incentive this builds for recursive sourcing: bringing in your own people does not bury them. Whoever an actor engages, that relationship is a first-class commercial record with its own terms — legible to its two parties, private to everyone else.
What this design admits about itself.
- Arithmetic
-
The envelope leaks by subtraction
A participant who knows the total and their own share can bound what the others receive; with two participants the bound is exact. The design accepts this leak rather than concealing it, because hiding the total would destroy the transparency that is the point.
- Honesty
-
Under-declaration is possible
Nothing prevents a client declaring less than they settle off-channel, and nothing is built to detect it. The counter-pressure is structural: shares are defined against the declared budget, so an under-declared project's own numbers fail to add up in front of the people being paid from them.
- Friction
-
Deactivation will misfire
Visible deactivation will sometimes hit a client who is merely slow rather than insolvent. That is the accepted cost of making budget trouble visible early — preferred to an invisible arrears state that lets work be commissioned against money that may not exist.
- Open
-
Questions that belong to counsel
The enforceability of a non-refundable activation fee under Japanese consumer and commercial law, the formal Japanese name of the fee, and the exact rounding rule for fee arithmetic are recorded as open questions for counsel and the money kernel's accounting policy — not decided on a marketing page.
KeizAI is in an early specification phase. This page describes design intent recorded in the project's charter, invariants and ADR-036 — not shipped software, and not an offer: no price, rate, feature or launch date is claimed here.