Deerfield Green — Whitepapers

Workforce Research Published

The AI Layoff Illusion

How COVID Overhiring Became “AI Efficiency” — and Why It Matters for Your Workforce Strategy

Challenger Gray tracked 54,836 jobs cited as AI-related in 2025. Gartner found less than 1% of layoffs were actually caused by AI. This paper separates genuine AI displacement from pandemic-era overhiring repackaged under a financially rewarding narrative.

Less than 1% of recent layoffs are genuinely caused by AI — the overwhelming majority are corrections to pandemic-era overhiring repackaged under a narrative that rewards companies with higher stock prices.

< 1%

of layoffs actually caused by AI

54,836

jobs cited as “AI-related” in 2025

55%

of employers regret AI-attributed cuts

10 : 1

AI job creation-to-displacement ratio

27 Sources 6 Figures ~5,000 Words BLS / Challenger / Gartner / Forrester

Contents — 8 Sections

1. The Headlines vs. the Numbers

54,836 AI-cited cuts versus Gartner's finding that less than 1% were actually caused by AI productivity gains.

2. The COVID Hiring Binge

How Meta (+143%), Amazon (+148%), and Salesforce (+127%) built workforces they couldn't sustain.

3. The Narrative Pivot

From “we hired too many people” to “AI made them redundant” — traced through earnings call data.

4. Follow the Money

AI stocks account for ~75% of S&P 500 returns since ChatGPT launched. The financial incentive to frame cuts as AI-driven.

5. Where AI Actually Displaces Workers

Klarna, Duolingo, and Block case studies — separating genuine displacement from narrative cover.

6. The Counterevidence

ITIF's 10:1 job creation-to-displacement ratio. PwC's finding that AI-exposed wages rise 2× faster.

7. The Regret Cycle

55% of employers regret AI-attributed layoffs. 50% will rehire customer service staff by 2027.

8. What Enterprise Leaders Should Do

Build a workforce taxonomy. Watch for real displacement signals. Resist the pressure to frame restructuring as AI-driven.

AI Economics Published

Token Consumption Patterns Across the Enterprise AI Stack

How 300× Price Deflation Led to 320% Spending Growth — and What It Means for Your AI Budget

Per-token prices have fallen 300× since 2023, yet total enterprise AI spending surged 320% in 2025. This paper maps the token economy across six dimensions to explain the Jevons Paradox playing out in real time.

Costs fell 1,000×, but bills tripled. Token consumption is emerging as the unit of economic activity in the AI era — and most enterprise budgets were built for a different world.

300×

per-token price deflation since 2023

1.5Q

tokens processed monthly, globally

10–40×

agentic token multiplier per task

320%

enterprise AI spending surge in 2025

25 Sources 9 Figures ~6,000 Words OpenRouter / OpenAI / Google / JPMorgan

Contents — 8 Sections

1. The Token Economy at Scale

1.5 quadrillion tokens per month, 50 trillion per day. Token volume as a leading economic indicator.

2. Where the Tokens Actually Go

Programming now consumes 50%+ of all tokens. Average prompt length quadrupled to 6,000+ tokens.

3. The Agentic Multiplier

Coding agents consume 10–40× more tokens per task. Claude Code: 33K tokens median; Cursor Agent: 188K.

4. The 300× Deflation

GPT-4 at $30/1M tokens (2023) to Gemini Flash at $0.10/1M today. Hidden multipliers in output and reasoning tokens.

5. The Jevons Paradox

When costs drop 1,000×, 1,000× more use cases become viable. A structural feature, not a temporary anomaly.

6. The Global Token Map

China: 140 trillion tokens/day domestically. DeepSeek and Qwen: 61% of consumption on global platforms.

7. Cost Per Useful Output

AI support at $0.99–2.00/ticket vs. $5–15 human. Coding at $0.28–0.67 per resolution.

8. Planning for the Token Economy

Model tiering strategies, agentic cost governance, and budget modeling that accounts for Jevons dynamics.

Labor Market Published

The Job Description Signal

What 1.3 Billion Postings Reveal About AI’s Real Impact on Workforce Strategy

AI mentions in U.S. job postings surged 134% since 2020 while total postings grew only 6%. The structural finding: augmentation-prone roles are growing in complexity and pay, while automation-prone roles are shrinking in scope and disappearing at entry level.

Job descriptions are the earliest measurable signal of AI's structural impact on the labor market — and the data shows a widening split between roles being augmented and roles being hollowed out.

134%

AI mention growth in U.S. postings

28–56%

AI skills wage premium

−73%

entry-level hiring rate collapse

+264%

Chief AI Officer adoption in 3 years

23 Sources 8 Figures ~6,500 Words Indeed / LinkedIn / PwC / Lightcast / HBS

Contents — 8 Sections

1. The 134% Signal

AI mentions surged 134% since 2020 while total postings grew 6%. The headline number obscures critical distinctions.

2. Augmentation vs. Automation

HBS research reveals the asymmetry: augmentation roles get longer, more complex descriptions; automation roles get shorter ones.

3. The Sector Divergence

51% of AI-skilled postings now sit outside IT. Manufacturing, finance, and marketing show the fastest premium growth.

4. The Wage Premium

28–56% salary premium that widens with seniority: 6.2% at entry level to 18.7% at staff level.

5. The Entry-Level Crisis

Entry-level hiring collapsed 73.4%. AI automates the tasks that previously served as on-ramps for junior talent.

6. New Roles, Dying Roles

Chief AI Officer up 264%, Prompt Engineer up 136%. Degree requirements down 7–9 percentage points since 2019.

7. The Global Regulatory Split

EU classifies AI in recruitment as “high risk.” U.S. has no federal framework. Singapore leads at 3.2% AI skill concentration.

8. Workforce Architecture Playbook

Audit job descriptions as diagnostic. Restructure roles along augmentation/automation axis. Rebuild entry-level pathways.

Strategy Published

The AI Adoption Curve

Why 88% of Enterprises Use AI and Only 5% See Returns

Four major maturity frameworks converge: most enterprises are stuck in early stages, with only 5–7% achieving substantial financial returns. The gap is not technology access but workforce readiness and organizational design.

The companies that invest in complementary capabilities now will be the 5% that capture outsized returns as the J-curve inflects upward.

88%

of organizations use AI in at least one function

5–7%

see substantial financial returns

$5.5T

unrealized value from skills gap (IDC)

3.8×

higher returns with mature training programs

15 Sources 8 Figures ~6,500 Words MIT CISR / McKinsey / BCG / Gartner / WEF

Contents — 8 Sections

1. The Adoption Paradox

88% adoption vs. 5–7% returns. Four maturity models converge on the same finding: most are stuck early.

2. The J-Curve Is Real

Census Bureau/MIT/Stanford: 1.33 percentage point initial productivity decline before gains materialize.

3. Pilot Purgatory

Five failure modes: no business ownership, data quality, process rigidity, integration gaps, measurement collapse.

4. The Workforce Readiness Deficit

Only 12% use AI daily. Only 20% “talent ready.” 84% have not redesigned jobs around AI.

5. Industry Scorecard

Sector-by-sector: financial services leads ($3,200/employee), healthcare grows fastest (36.8% CAGR).

6. The ROI Timeline

Only 6% see payoff in under a year. Most: 2–4 years. Organizations with mature training: 3.8× returns.

7. From Access to Competence

Four levers: job redesign, structured enablement, business ownership, measurement frameworks before pilots.

8. The Harvest Phase

U.S. productivity growth hit 2.7% in 2025. The window to build complementary capabilities is closing.