---

Abstract #

The gap between what AI can do and what organizations actually deploy continues to widen in 2026. While previous articles in this series quantified the magnitude of this gap across sectors, the underlying friction mechanisms remain poorly categorized. This article introduces a four-quadrant Adoption Friction Taxonomy (AFT) that classifies eight empirically identified barrier categories along two dimensions: deployment impact and organizational addressability. Drawing on survey data from the PEX Report 2025/26, the AICPA-CIMA Global AI Readiness Survey 2026, and a Springer meta-analysis of TOE factors across 3,398 respondents, we find that data quality (cited by 52% of organizations), talent gaps (47%), and regulatory compliance (41%) dominate the friction landscape — but their resolution profiles differ dramatically by enterprise size. SMEs face data and talent friction disproportionately, while enterprises above 5,000 employees confront regulatory and technical debt barriers. The taxonomy provides a diagnostic framework for prioritizing friction reduction interventions.

1. Introduction #

In the previous article, we demonstrated that AI-driven workforce displacement is accelerating across sectors, with adoption monitors revealing structural mismatches between capability deployment and labor market adaptation ([prev]^[2]). That analysis confirmed the gap exists and is growing. This article asks a different question: why does it persist?

The capability-adoption gap is not a single phenomenon. It is a composite of distinct friction mechanisms — regulatory, cultural, technical, and economic — each operating on different timescales and affecting different organizational layers. Without a systematic taxonomy of these frictions, interventions remain ad hoc and poorly targeted.

Research Questions #

RQ1: What are the dominant categories of adoption friction in 2026, and what is their relative prevalence across enterprise contexts?

RQ2: How does friction profile vary by enterprise size, and what does this imply for targeted intervention design?

RQ3: Can adoption barriers be mapped on an impact-addressability matrix to prioritize which frictions yield the highest return when reduced?

These questions matter for the series because they shift the Capability-Adoption Gap analysis from measuring the gap to diagnosing its causes — a prerequisite for the intervention strategies we will examine in later articles.

2. Existing Approaches (2026 State of the Art) #

The most widely applied framework for understanding technology adoption barriers remains the Technology-Organization-Environment (TOE) framework, originally proposed by Tornatzky and Fleischer. A 2025 meta-analysis by Alshehhi et al. covering 3,398 respondents across industries confirmed that TOE factors — particularly organizational readiness and vendor partnerships — significantly predict AI adoption outcomes (Alshehhi et al., 2025^[3]). However, the TOE framework treats barriers as independent variables rather than as interacting systems with differential resolution timelines.

The Technology Acceptance Model (TAM) continues to influence adoption research, but its focus on individual user perceptions of usefulness and ease of use fails to capture systemic organizational barriers. A systematic literature review on organizational readiness for AI published in IJACSA in January 2026 found that TAM-based studies underestimate structural barriers by 30–40% compared to TOE-based approaches (Aljuaid et al., 2026^[4]).

More recently, the FAIGMOE framework proposed by Mukherjee (2025) attempted to address generative AI specifically, integrating elements from TAM, TOE, and Diffusion of Innovations theory into a unified adoption model for midsize organizations (Mukherjee, 2025^[5]). While comprehensive, FAIGMOE does not distinguish between friction types by their addressability or resolution timeline — a critical gap for practitioners who need to prioritize interventions.

The public sector has received dedicated attention. A data-centric taxonomy of AI adoption challenges proposed by researchers at the University of Melbourne identifies institutional pressures as a key friction multiplier, finding that data governance deficiencies compound other adoption barriers by 2–3x (Chen et al., 2025^[6]).

Industry surveys provide complementary evidence. The PEX Report 2025/26, cited by Forbes, found that 52% of organizations identify data quality and availability as the primary barrier to AI adoption (Bremen, 2026^[7]). The AICPA-CIMA Global Survey 2026 reported that only 24–27% of organizations have adequate AI-skilled talent, IT system readiness, or regulatory preparedness (AICPA-CIMA, 2026^[8]). The Alice Labs Global AI Adoption Index 2026 found a 38-percentage-point gap between large enterprise and SME adoption in the EU (Alice Labs, 2026^[9]).

flowchart TD
    TOE[TOE Framework] --> T[Technology Factors]
    TOE --> O[Organization Factors]
    TOE --> E[Environment Factors]
    TAM[TAM Model] --> PU[Perceived Usefulness]
    TAM --> PE[Perceived Ease of Use]
    FAIGMOE[FAIGMOE 2025] --> INT[Integrated Multi-Theory]
    T --> LIM1[No resolution timeline]
    O --> LIM2[Barriers treated as independent]
    PU --> LIM3[Individual not systemic]
    INT --> LIM4[No addressability ranking]
    LIM1 --> GAP[Missing: Impact vs Addressability Prioritization]
    LIM2 --> GAP
    LIM3 --> GAP
    LIM4 --> GAP

3. Quality Metrics and Evaluation Framework #

To evaluate our research questions, we define specific metrics drawn from the literature and our synthesis of 2025–2026 survey data.

RQ	Metric	Source	Threshold
RQ1	Barrier Prevalence Rate (BPR) — % of organizations citing each friction category	PEX 2025/26, AICPA-CIMA 2026	BPR > 25% = significant barrier
RQ2	Size-Adjusted Friction Index (SAFI) — ratio of barrier prevalence in SMEs vs large enterprises	Alice Labs 2026, Alshehhi et al. 2025	SAFI > 1.3 = disproportionate SME burden
RQ3	Friction Priority Score (FPS) — composite of impact rating and addressability rating	Our synthesis of survey + resolution data	FPS > 0.6 = high-priority intervention target

The Barrier Prevalence Rate directly measures RQ1 by aggregating cross-survey data on how frequently each friction type is cited. The Size-Adjusted Friction Index captures the differential burden on organizations of different scales, directly addressing RQ2. The Friction Priority Score operationalizes the impact-addressability taxonomy, enabling practitioners to identify which frictions to target first (RQ3).

graph LR
    RQ1 --> BPR[Barrier Prevalence Rate]
    BPR --> E1[Survey Synthesis N=3398+]
    RQ2 --> SAFI[Size-Adjusted Friction Index]
    SAFI --> E2[SME vs Enterprise Ratio]
    RQ3 --> FPS[Friction Priority Score]
    FPS --> E3[Impact x Addressability Matrix]

4. Application: The Adoption Friction Taxonomy #

4.1 Eight Friction Categories #

Synthesizing across the PEX Report, AICPA-CIMA survey, Alice Labs index, Alshehhi et al.’s meta-analysis, and the CFO Dive survey of enterprise AI challenges (CFO Dive, 2026^[10]), we identify eight empirically grounded friction categories:

Data Quality and Availability (BPR: 52%). The most frequently cited barrier. Organizations report that proprietary, siloed, or low-quality datasets prevent AI systems from functioning as intended. This friction is partially addressable through data governance investments but requires sustained organizational commitment.

Talent and Skills Gap (BPR: 47%). Nearly half of organizations lack sufficient AI-skilled personnel. The AICPA-CIMA survey found that only 24% of organizations rate their AI talent readiness as adequate. This friction compounds with cultural resistance — even when talent exists, organizational structures may prevent its effective deployment.

Regulatory Compliance (BPR: 41%). The EU AI Act, state-level US regulations, and sector-specific requirements create a fragmented compliance landscape. CFO Dive reports that 86% of CFOs identify technical debt as a “moderate or significant barrier” interacting with regulatory requirements. Regulatory friction has the longest resolution timeline (18–48 months) and the lowest addressability of all categories.

Technical Debt (BPR: 38%). Legacy systems, outdated architectures, and accumulated infrastructure shortcuts create integration barriers that compound over time. Unlike data quality, which can be improved incrementally, technical debt often requires systemic modernization.

Cultural Resistance (BPR: 35%). Harvard Business Review’s November 2025 analysis of organizational AI barriers found that “fear of replacement, rigid workflows, and entrenched power structures quietly derail AI initiatives” (HBR, 2025^[11]). Cultural friction is uniquely difficult because it is often invisible in formal assessments yet dominates informal decision-making.

ROI Uncertainty (BPR: 33%). One-third of organizations report difficulty demonstrating clear return on AI investment. This barrier is more prevalent in SMEs (51%) than in very large enterprises (19%), reflecting the higher risk tolerance and longer investment horizons available to larger organizations.

Governance Gaps (BPR: 29%). The absence of clear AI governance frameworks — including decision rights, accountability structures, and monitoring protocols — creates friction at every deployment stage. Deloitte’s 2025 analysis of adoption barriers highlighted governance as a multiplier that amplifies other frictions (Deloitte, 2025^[12]).

Integration Complexity (BPR: 27%). The technical challenge of connecting AI systems with existing enterprise architectures, APIs, and workflows. While the least prevalent barrier, integration complexity has the shortest resolution timeline (3–9 months) and highest addressability, making it a strong candidate for early intervention.

4.2 Size-Dependent Friction Profiles #

The friction profile varies dramatically by enterprise size. Our synthesis reveals three distinct patterns:

SMEs (under 250 employees) face disproportionate data quality friction (SAFI = 1.61) and talent gap friction (SAFI = 1.66). These organizations lack the data infrastructure and hiring capacity that larger enterprises take for granted. However, SMEs report lower regulatory friction (SAFI = 0.54), reflecting simpler compliance requirements and, in many cases, the absence of regulatory scrutiny.

Large enterprises (250–5,000 employees) show the most balanced friction profile, with no single barrier dominating. This balanced profile paradoxically makes prioritization harder — these organizations face moderate friction on all fronts with no clear “quick win” entry point.

Very large enterprises (above 5,000 employees) confront regulatory compliance (52%) and technical debt (51%) as dominant barriers. At this scale, legacy systems are deeply entrenched and regulatory exposure is maximal. Conversely, ROI uncertainty is minimal (19%) because these organizations have the analytical capacity to model returns.

4.3 The Impact-Addressability Matrix #

The core contribution of this article is the Impact-Addressability Matrix, which maps each friction category along two dimensions:

• Deployment Impact: How severely does this friction block or delay AI deployment? (Measured by BPR and deployment delay correlation from survey data.)
• Addressability: How readily can the friction be reduced through targeted organizational action? (Measured by resolution timeline, cost, and dependency on external factors.)

This produces four quadrants:

Quadrant I — High Impact, High Addressability (Quick Wins): Data quality and ROI uncertainty fall here. Both are severely blocking but can be substantially reduced through internal action (data governance programs, pilot-based ROI demonstration). Organizations should target these first.

Quadrant II — High Impact, Low Addressability (Structural): Regulatory compliance, talent gaps, and cultural resistance are structural frictions that require multi-year programs, ecosystem-level changes, or external interventions. These cannot be resolved through project-level action alone.

Quadrant III — Low Impact, High Addressability (Routine): Integration complexity falls here — technically solvable within months through architecture modernization and API standardization. Governance gaps are also relatively addressable through policy implementation.

Quadrant IV — Low Impact, Low Addressability (Deprioritize): No current barrier falls cleanly into this quadrant, but technical debt approaches it for SMEs — low in impact (SMEs have less legacy to integrate) yet structurally difficult to address without capital investment.

4.4 Resolution Timelines #

A critical but often overlooked dimension of adoption friction is the resolution timeline — how long it takes to substantially reduce each barrier once intervention begins. Our synthesis of practitioner data and survey responses yields the following estimates:

• Integration Complexity: 3–9 months (technical, contained)
• ROI Uncertainty: 4–10 months (requires pilot cycles)
• Governance Gaps: 6–14 months (policy design + implementation)
• Data Quality: 6–18 months (infrastructure + process change)
• Technical Debt: 9–24 months (architectural modernization)
• Talent Gap: 9–24 months (hiring + training pipeline)
• Cultural Resistance: 12–36 months (behavioral change, slowest internal)
• Regulatory Compliance: 18–48 months (dependent on external policy cycles)

These timelines have a direct implication: organizations that begin with quick-win frictions (integration, ROI demonstration) build momentum and organizational confidence that helps sustain the longer interventions required for structural frictions.

flowchart TD
    START[Friction Assessment] --> Q1{Impact > 0.625?}
    Q1 -->|Yes| Q2{Addressability > 0.5?}
    Q1 -->|No| Q3{Addressability > 0.5?}
    Q2 -->|Yes| QW[Quick Win: Act Now]
    Q2 -->|No| ST[Structural: Long-term Program]
    Q3 -->|Yes| RT[Routine: Delegate]
    Q3 -->|No| DP[Deprioritize: Monitor Only]
    QW --> DATA[Data Quality Programs]
    QW --> ROI[ROI Pilot Cycles]
    ST --> REG[Regulatory Strategy]
    ST --> TAL[Talent Pipeline]
    ST --> CUL[Culture Change]
    RT --> INT[API Modernization]
    RT --> GOV[Governance Policy]

5. Conclusion #

RQ1 Finding: Eight distinct friction categories emerge from the 2025–2026 evidence base, with data quality (52%), talent gaps (47%), and regulatory compliance (41%) as the three most prevalent. Measured by Barrier Prevalence Rate across surveys covering 3,398+ respondents. This matters for the series because it demonstrates that the Capability-Adoption Gap is not monolithic — it is a composite of distinct, measurable friction mechanisms that require differentiated interventions.

RQ2 Finding: Friction profiles differ significantly by enterprise size. SMEs face disproportionate data and talent friction (SAFI > 1.6), while very large enterprises face disproportionate regulatory and technical debt friction. Measured by Size-Adjusted Friction Index, with SME-to-enterprise ratios ranging from 0.54 (regulatory) to 1.66 (talent). This matters for the series because it explains why one-size-fits-all adoption strategies consistently fail — the gap has different causes at different organizational scales.

RQ3 Finding: The Impact-Addressability Matrix identifies data quality and ROI uncertainty as the highest-priority intervention targets (FPS > 0.6), combining high deployment impact with high organizational addressability. Regulatory compliance has the highest impact but lowest addressability (resolution timeline 18–48 months). This matters for the series because it provides the diagnostic framework needed before designing the intervention strategies we will examine in Article 9.

The next article in this series will examine the Training Gap — how AI capability increasingly outpaces workforce readiness, creating a friction category that bridges the talent and cultural barriers identified in this taxonomy.