Agentic AI Systems Should Be Designed as Marginal Token Allocators

TL;DR

This paper advocates for designing agentic AI systems as marginal token allocation economies, unifying various layers under a common economic framework to improve efficiency and address failure modes.

Contribution

It introduces a marginal token allocation perspective that unifies multiple AI system layers and predicts failure modes, guiding future evaluation and resource management research.

Findings

All four AI layers optimize the same economic condition.

Identifies recurring failure modes like over-routing and congestion.

Predicts that local token minimization can misallocate resources.

Abstract

This position paper argues that agentic AI systems should be designed and evaluated as \emph{marginal token allocation economies} rather than as text generators priced by the unit. We follow a single request -- a developer asking a coding agent to fix a failing test -- through four economic layers that today are designed in isolation: a router that decides which model answers, an agent that decides whether to plan, act, verify, or defer, a serving stack that decides how to produce each token, and a training pipeline that decides whether the trace is worth learning from. We show that all four layers are solving the \emph{same} first-order condition -- marginal benefit equals marginal cost plus latency cost plus risk cost -- with different index sets and different prices. The framing is deliberately minimal: we do not propose a complete theory of AI economics. But adopting marginal token allocation as the shared accounting object explains why systems that locally minimize tokens globally misallocate them, predicts a small set of recurring failure modes (over-routing, over-delegation, under-verification, serving congestion, stale rollouts, cache misuse), and points to a concrete research agenda in token-aware evaluation, autonomy pricing, congestion-priced serving, and risk-adjusted RL budgeting.