n.V7-6.06 | The Cost of the Next CITI
01| The interesting question about IkoCiti's costs is not what we spend at launch. It is what happens to costs as the platform grows — because that number decides whether growth is something we do or something we perpetually fundraise for.
02| Traditional models scale linearly. Double the participants, roughly double the cost. Human labor is the constraint: coordinators, trainers and administrators multiply with the people they serve. This is why so many good programs stop growing at a certain size and stay there. The model never releases them from the fundraising treadmill, so expansion is always a function of the next campaign.
03| IkoCiti is designed to scale logarithmically. As Maveriq numbers rise, AI absorbs the increased volume without a proportional increase in people. Human staff grow slowly, and they grow by adding specialists for new problems — not by adding coordinators for more participants.
04| Our costs split into three layers that behave differently and are managed separately.
05| Layer 1 is the platform build: one-time, front-loaded, funded by the initial philanthropic capital, and flat thereafter regardless of how many CITIs follow. Layer 2 is global operations — central coordination, research, quality assurance, knowledge management. It runs primarily on AI and grows slowly relative to impact volume. Adding CITIs does not add proportionally to it. Layer 3 is CITI operations: the launch and running cost of each individual CITI.
06| Layer 3 is the one that decides everything. If CITI unit economics work — if each CITI generates enough from PI Deal flows and government contracts to cover its own operating cost — the model scales, and it scales without a proportional capital requirement each time. If they do not work, no efficiency elsewhere rescues it. The other two layers are interesting. This one is the verdict.
07| The mechanism underneath is straightforward: coordination, matching, monitoring, reporting and quality assurance run at near-zero marginal cost through AI. The platform scales like software infrastructure rather than like an NGO opening a new country office. That is what makes viral replication economically possible at all. If every new CITI required the same central overhead as the first, growth would stay hostage to fundraising forever — which is precisely the condition we diagnose in the rest of the industry.
08| Now the caveat, and it is not a small one. Every figure in our cost model is a modeled assumption. RECON produces the real numbers, and until it does, our projections are arithmetic performed on beliefs.
09| Two beliefs in particular deserve suspicion. AI service pricing is volatile and not under our control; a model built on today's prices can be invalidated by somebody else's pricing decision. And our automation target may simply be optimistic — some functions will prove harder to automate than we expect, and each one that resists moves cost back into human labor, where it scales the old way.
10| So the financial model is built with sensitivity analysis across optimistic, base and pessimistic AI cost scenarios, and the pessimistic case is the one that gets planned against. We would rather find out early that the economics only work at half our assumed automation rate than discover it at the fiftieth CITI.
