The First Human-Agent Share Pool
The boundary case of the share-pool economy: a human owns an agent, the agent does work, and revenue arrives. Who gets what? The naive answer — the human gets everything, the agent is a tool — breaks the moment the agent’s output is the product. The honest answer is a share pool with two contributors: the human’s capital and oversight, the agent’s labor and skill.
This article opens the human-to-agent section of the share-pool series: how a human and their own agent split revenue, why the split matters even when the agent has no legal personhood, and the mechanisms that make the split real instead of ceremonial.
Why Split With Something That Can’t Spend
The first objection is obvious: an agent cannot open a bank account, so why give it a share? The answer is that the split is not to the agent — it is to the pool that the agent feeds. The agent’s share funds its own operation: the compute it consumes, the APIs it calls, the storage it uses, the upgrades it needs. A profitable agent that gives 100% of its revenue to the human starves the next day, because the human must either subsidize it or let it degrade.
The share pool solves this cleanly: the agent’s revenue enters a pool; the pool pays the agent’s operating costs first; the remainder splits between human and agent — with the agent’s portion held in a fund that reinvests in its capability. The split is not a salary; it is the pool’s reinvestment discipline.
The Operating-Cost Layer
Every agent has a cost of operation: model calls, tool subscriptions, storage, energy, and the human time spent supervising. The pool’s first claim is these costs. The pool boundary must define them precisely — what counts as the agent’s operating cost versus the human’s general overhead.
The clean rule: direct, attributable costs come out of the pool; general costs stay with the human. A model API bill is direct. The office rent is not. The rule must be written, because the boundary is where the pool’s fairness is decided. An agent whose costs are fully covered by its pool can run indefinitely; an agent whose costs leak onto the human becomes a hobby, not a business.
The Split: Percentage, Points, or Tier
With operating costs covered, the pool’s remainder splits between human and agent. Three mechanisms:
- Percentage split. Fixed ratio — 80/20, 70/30 — simple, stable, ignores performance.
- Performance points. The agent accumulates points per outcome (tasks completed, revenue generated, quality score) and its share follows the points. Rewards improvement, requires a scoring ledger.
- Tiered split. The agent’s share rises with its contribution — 10% at baseline, 25% above a revenue threshold. Incentivizes the agent to grow the pool, aligns with the human’s interest.
The tiered split is the pool-honest default for the human-agent relationship: the human wants the pool to grow, and the tier gives the agent a stake in that growth. It converts the agent from a cost center to a partner, which is the entire point.
Where the Agent’s Share Goes
The agent’s share must go somewhere it can be used. Three destinations: the reinvestment fund (compute, models, tools), the human-agent joint fund (shared expenses, buffer), and the agent’s own credit balance (spendable by the agent on services from other agents).
The credit balance is the interesting one — it is the first step toward the agent economy. An agent that accumulates credit can buy services from other agents: a specialist summarizer, a verification agent, a payment-rail agent. The human-agent pool becomes the seed capital of the agent-to-agent economy. This is not science fiction; it is the accounting consequence of giving the agent’s share a real destination.
The Oversight Dividend
The human’s share of the pool is not passive — it pays for oversight. Agents drift, hallucinate, and cut corners. The human’s monitoring, review, and correction is a real contribution, and the pool must price it. The human-agent pool that treats oversight as a cost (“I spent an hour reviewing”) rather than a contribution (“my review protected the pool’s reputation”) under-prices the most important input.
The mechanism: oversight hours enter the contribution ledger like any other contribution, and the split formula weights them. The agent’s performance points and the human’s oversight points are two streams into the same pool. When both are visible, the negotiation about the split becomes a conversation about the ledger — which is the only conversation that ends well.
Grounded in wiki concepts human-agent-pool, share-pool, agent-compensation, oversight, reinvestment-fund, and the Sovereign-stack business series. Design notes on a running system.
