Workflow Economics · Token Efficiency · Enterprise AI Capital Allocation

Token Efficiency Framework

Four-tab interactive framework for assessing AI displacement, signal thresholds, token-cost curves, and workflow-level capital allocation posture.

Panel 1 of 4 — the displacement equation

Workflow displacement crosses 1.0 when fidelity × human cost overtake AI cost × error penalty

Variables in green are moving in favor of automation. Variables in amber hold steady or rise. The ratio's crossing point is different for every enterprise workflow — set by its signal threshold and the training data available to it.

The displacement equation A ratio showing intelligence fidelity times human baseline cost over AI execution cost plus error rate times error cost. The ratio must exceed 1.0 for displacement. Human baseline cost rising — wages, overhead × Intelligence fidelity rising — algorithmic curve AI execution cost falling 10× / 12–18 months + Error rate tied to fidelity gains × Error cost workflow-specific > 1.0 displacement threshold For high-consequence workflows the path to crossing the threshold runs through fidelity improvement, not AI cost reduction.

Panel 2 of 4 — three scaling regimes

Compute economics run on three curves now, not one

Token cost efficiency improves roughly 10× every 12–18 months — orders of magnitude faster than Moore's Law ever delivered. A board applying a uniform decay assumption to a mixed AI-era portfolio will systematically misprice both the opportunity and the risk.

Token cost efficiency GPU AI performance Moore's Law
Token cost efficiency rises from 1x in 2022 to 4000x by 2028. GPU AI performance reaches 350x. Moore's Law reaches 8x.

Index: 2022 = 1.0 baseline · log scale · 2026 onward projected. Sources: synthesis of frontier model pricing (Anthropic, OpenAI, Google), NVIDIA datasheets, and historical CMOS scaling.

Panel 3 of 4 — the decision matrix

Two axes — signal threshold and position on the token efficiency curve — yield four capital postures

Each dot is a representative enterprise workflow placed by where its signal-to-noise threshold sits today and how soon AI economics cross the displacement line. The slope of the threshold curve is the analytical point: high-judgment workflows require disproportionately more progress before displacement becomes economic.

The decision matrix A two by two matrix with signal threshold on the vertical axis and position on the token efficiency curve on the horizontal axis. Workflows are placed in four capital allocation quadrants. PROTECT & INVEST 36+ months · deepen data moat BUILD & DEFEND 24–48 months · proprietary signal SEQUENCE NEXT 12–24 months · prepare infrastructure AUTOMATE NOW 0–6 months · cost overhang displacement frontier Strategic M&A judgment Litigation strategy Wealth advisory prep Compliance monitoring Contract analysis Standard code review Onboarding workflows Invoice processing Data entry Tier 1 support far from displacement below displacement line Position on token efficiency curve → ← Signal threshold of workflow high low

Panel 4 of 4 — workflow assessor

Score any workflow on four dimensions to place it on the matrix

The assessor is the operational instrument — what a CIO uses inside a portfolio review. Adjust the sliders to characterize a workflow; the placement, recommendation, and ROIC horizon update live.

35
70
40
60
Posture
Automate now
below displacement line
Displacement ratio
1.8
crosses 1.0 → displaceable
ROIC horizon
0–6 mo
capital deployment window
Capital allocation guidance