Research Notes on AI Adoption Depth

AI is often described as an inevitable force reshaping work. Yet when you look across organizations, industries, and stages of diffusion, the labor outcomes are strikingly uneven. Some firms redesign workflows and career ladders; others see only incremental efficiency gains from the same underlying technology.

This note summarizes the argument and findings from my paper, Adoption Depth and Organizational–Labor Transformation in the AI Era (Jormp LLC, March 2026). The central claim is simple: heterogeneity in AI's labor impact is driven less by capability than by adoption depth the joint alignment of organizational embedding and practitioner mastery.

The puzzle

Why does similar AI capability produce such different organizational and labor outcomes?

Capability defines the technological frontier. But realized transformation depends on how deeply AI is adopted not merely whether a tool is deployed, but whether it is embedded in how work is organized and performed.

I formalize this as adoption depth: a complementarity between two dimensions.

Organizational embedding (OE) captures integration into workflows, governance, and decision rights.

Practitioner mastery (PM) captures workers' ability to orchestrate AI in production and judgment.

Together they form an adoption depth index:

ADI = OE^θ₁ × PM^θ₂

The multiplicative structure matters. High embedding cannot compensate for near-zero mastery, and vice versa. Returns come from bundles of mutually reinforcing practices — not from tools alone.

Three regimes of transformation

Adoption depth implies regime differences in how AI shows up in labor markets:

1. Shallow: Tool-level automation with limited redesign
2. Intermediate: Augmentation dominates; workflows shift partially
3. Deep: Job architecture and hierarchy are reconfigured

These regimes predict nonlinear wage gradients and threshold-like behavior as adoption deepens not a smooth, linear diffusion curve.

What the data show

Using U.S. worker-level microdata from 2015–2025 (N = 869,273), I test four hypotheses mapped to adoption-depth predictions:

Exposure premium: AI exposure is positively associated with wages - confirmed.

Intensity gradient: Premia scale with exposure intensity - confirmed.

Complementarity amplification): Premia amplified for mid-to-high education - confirmed.

Regime dynamics: Nonlinear effects across diffusion stages - confirmed.

Baseline wage premium

In the baseline specification (log real wages, state and year fixed effects, demographic controls), AI intensity carries a coefficient of 0.4546 (robust SE: 0.0012). This is a conservative reduced-form premium positive and highly significant, consistent with early augmentation on average rather than uniform displacement.

Education as a mastery proxy

Education interactions tell the complementarity story most clearly. Wage premia rise materially for mid-to-high education strata, peak around the middle of the distribution, and taper at the highest levels. The pattern aligns with practitioner mastery capacity: workers who can operationalize AI as an orchestration layer not just run a prompt capture the largest gains.

This is human–AI complementarity in wage data, not automation alone.

Diffusion-stage dynamics

Year interactions reveal a nonlinear diffusion path: early-period compression (2018–2019), partial recovery in 2021, and renewed compression during accelerated diffusion (2022–2025).

Threshold estimation (Hansen 1999) finds marginal AI wage effects increase after roughly the 65th percentile of an adoption-depth proxy consistent with a regime transition rather than a single linear effect.

Conceptual model

AI Capability (technological frontier)
↓
Organizational Embedding × Practitioner Mastery
↓
Adoption Depth (ADI)
↓
Shallow → Intermediate → Deep regimes
↓
Wages · role design · hierarchy · mobility

AI exposure and intensity at the occupation level proxy the labor-market manifestation of underlying adoption depth and how task content and diffusion stages reflect deeper organizational choices.

Implications

For executives

Tool availability is insufficient. Realized returns depend on aligning workflow redesign (embedding) with workforce capability development (mastery). The strongest premia concentrate where workers treat AI as an orchestration layer integrated into judgment-heavy production — not as a one-off productivity hack.

For policymakers

AI-driven wage gains are distributional, not uniform. If mastery capacity (education, training, safe experimentation) lags diffusion, skill premia can widen. Policy should expand mastery capacity and enable responsible embedding — not treat capability shocks as deterministic displacement.

For practitioners

Career value increasingly reflects the ability to integrate AI into production and judgment. Credentials matter, but the mechanism is operational mastery: fluency, orchestration, and the judgment to know when AI augments versus when it misleads.

Contribution

1. Organizational theory: Adoption depth as a complementarity mechanism with regime implications, bridging capability to realized transformation.

2. Labor economics: Intensity-scaled, education-amplified AI wage premia documented at scale across nearly 870,000 workers.

3. Policy discourse: A framework emphasizing alignment and reskilling over technological determinism.

The study integrates organizational complementarity theory (Milgrom & Roberts; Teece et al.) with task-based and skill-biased technological change literatures (Autor; Acemoglu & Restrepo; Goldin & Katz), and provides a foundation for cross-country institutional and policy extensions.

The takeaway is disciplined but practical: shallow adoption looks like tools; deep adoption reshapes jobs. Labor outcomes depend on adoption depth, the joint product of how organizations embed AI and how practitioners master it.

For depth on data and result, click link.

—

Anuoluwapo Joshua