--- description: >- Virtual Employee Economics: the Three Laws, the moving human-machine labor boundary, and why operator-agnosticism is the economically load-bearing property. --- # Virtual Employee Economics Every other page in this section asks whether governed delivery *works*. This one asks what it *costs* — and what happens to the economics of delivery when a "worker" can be a person or an AI agent, interchangeably. The research program calls an AI agent engaged this way a **virtual employee (VE)**: not a tool a person operates, but a unit of labor that takes on work a person would otherwise do. Once that substitution is possible at all, delivery leaders face a labor market with two kinds of workers whose costs and capabilities move on very different curves. The Three Laws of Virtual Employee Economics — developed by Vernon Keenan of Keenan Vision, in a research partnership with UC Berkeley's Haas School of Business — describe those curves. ## The Three Laws **Law 1 — Infinite Scale.** VE adoption follows a logistic curve: slow at first, then steep, then saturating at whatever ceiling a given market can carry. For a delivery leader the plain meaning is that workforce capacity stops being bound by hiring. Human capacity grows one recruit, one ramp-up at a time; VE capacity, once adoption inflects, grows faster than any org chart can — until it hits the market's carrying capacity, not yours. **Law 2 — Cognitive Commoditization.** The cost of a unit of cognitive work decays exponentially over time. Anything priced on the scarcity of a skill — analysis, configuration, documentation, first-draft anything — sits on a falling price curve. Margins built on charging for scarce cognition erode on a schedule set by the decay rate, not by competitive pressure. **Law 3 — Exponential Learning.** VE performance rises with adoption: the more the virtual workforce is used, the better it performs. Human teams improve too, but each new hire starts near zero; a VE workforce's gains accrue to every instance at once. Capability accrues with use. The laws carry a corollary. The **Gradient Value Corollary** models the quiet erosion of employment tiers: displacement does not arrive as one dramatic event but diffuses gradually across the skill ladder, tier by tier, as VE competitiveness — performance over cost — climbs while both curves move in the VE's favor. !!! note "How to read the laws" The Three Laws are a formal model, not a measurement. They are delivered as an interactive, parameter-driven application — adoption rates, cost-decay constants, and learning slopes are sliders to explore, not measured constants of nature. The instrument here is mathematics: the laws tell you what *follows if* the curves hold, and the research program's job is to estimate where, and how fast, they do. That is a weaker claim than field data — and a stated one. ## Why this matters for Rosetta If the boundary between human and machine labor is going to move — and the laws say it moves continuously, not once — then the economically decisive question is what happens to work when it crosses. Work that was briefed to a specific person, or wired into a specific tool, must be re-explained, re-scoped, and re-trusted on the far side. Work that was compiled for *a worker* crosses intact. ![A moving labor boundary between human workers and virtual employees: compiled work crosses intact, work briefed to a person stalls](../assets/ve-boundary-light.svg#only-light) ![A moving labor boundary between human workers and virtual employees: compiled work crosses intact, work briefed to a person stalls](../assets/ve-boundary-dark.svg#only-dark) That operator-agnosticism is the core design property of [governed intent compilation](../concepts/governed-intent-compilation.md), and Virtual Employee Economics is the argument that it is not a convenience but the economically load-bearing property of the whole system. The Cognitive DevOps Maturity Model working paper (first draft, June 2026) states this as a named hypothesis — **VE Economics**: the [Spec Manifest](../concepts/spec-manifest.md) is a unit of *economic* production, not just an engineering artifact. In plain terms — our reading, not the paper's — it is the thing an organization prices, allocates, and reassigns as the labor boundary shifts. The hypothesis is presented as a hypothesis, awaiting validation, and this site treats it as one. ## The honest counterweight: Quiet Erosion The same paper names the risk on the other side of the ledger — **Quiet Erosion**: organizations that delegate cognitive work may gradually lose the capacity to supervise the very systems they delegate to. (The corollary above uses the same phrase for employment tiers; this is the organizational version — erosion of oversight, not of jobs.) The research program treats this as a named risk with a validation agenda, not a footnote. And [governance](../concepts/governance.md) is the designed mitigation: delegation with recorded, inspectable reasoning — provenance, decision traces, an auditable trail — rather than delegation into a black box. --- *The Three Laws are Keenan Vision research, delivered as the interactive model described above; the Cognitive DevOps Maturity Model paper behind the VE Economics and Quiet Erosion hypotheses is summarized in [the research papers](papers.md); the Rosetta story starts at [the overview](../index.md).*