In 47 Tangerang batik SMEs, AI adoption appears to raise output mainly by improving HR management efficiency rather than through a direct technological boost; however, the finding rests on a small cross-sectional survey and cannot establish causality.

Main Finding

AI implementation in Tangerang batik SMEs has a significant positive effect on production output primarily through improving human resource management (HRM) efficiency. HRM efficiency functions as a key mediating mechanism: AI improves HRM efficiency, and that improved HRM efficiency then increases production output. The study finds AI adoption is currently modest (mean ≈ 2.84/5) while HRM efficiency and production output are at moderate levels (means ≈ 3.21 and 3.47).

Key Points

• Research gap filled: first quantitative, mediation-focused study of AI → HRM efficiency → production output in traditional batik SMEs in Tangerang, Indonesia.
• Hypotheses tested: H1 (AI → HRM efficiency), H2 (HRM efficiency → production output), H3 (AI → production output), H4 (HRM efficiency mediates AI → production output). Authors report support for the mediation model (AI → HRM → output).
• Sample and response: census of all registered batik SMEs in Tangerang (n = 47), 100% response rate via face-to-face survey.
• Descriptive results: AI implementation mean = 2.84 (SD 0.67); HRM efficiency mean = 3.21 (SD 0.58); production output mean = 3.47 (SD 0.62).
• Contextual emphasis: this is a craft- and labor-intensive sector where AI is positioned as complementary to human capital (enhances rather than replaces artisans).

Data & Methods

• Design: quantitative, explanatory, cross-sectional survey.
• Population/sample: all 47 batik MSMEs registered with local authorities in Tangerang (census/total sampling).
• Instruments:
  • AI Implementation: 8 indicators (automated data processing, decision support, predictive analytics, ML, NLP, computer vision, RPA, AI customer service).
  • HRM Efficiency: 9 indicators (recruitment efficiency, performance evaluation accuracy, training effectiveness, scheduling optimization, payroll speed, admin automation, data-driven decisions, workforce planning, employee satisfaction).
  • Production Output: 7 indicators (daily volume, quality consistency, production time efficiency, waste reduction, capacity utilization, delivery accuracy, customer satisfaction).
  • All items on 5‑point Likert scales; instrument content-validated and pilot-tested.
• Data collection: face-to-face visits over 8 weeks; demographic modules captured firm age, employees, revenue, product types.
• Analysis: PLS-SEM (SmartPLS 4.0) with two-stage evaluation (measurement model: AVE, loadings, Fornell-Larcker, HTMT; reliability: Cronbach's α & composite reliability; structural model: path testing and mediation analysis). PLS-SEM chosen due to small sample size, predictive focus, and non-normality tolerance.
• Key sample descriptors: majority owner-respondents (59.6%); firm sizes mostly micro/small (5–19 employees); revenue concentrated in IDR 300M–2.5B band; product types varied (hand-drawn, stamped, printed, combinations).

Implications for AI Economics

• Mechanism-focused insight: The primary productivity gains from AI in these SMEs operate indirectly via HRM improvement rather than solely through direct process automation. Economic models of AI adoption should incorporate organizational/HR channels as central mediators of productivity effects in labor-intensive, craft-based sectors.
• Policy targeting: Subsidies, grants, or tax incentives that aim to raise SME productivity via AI should be coupled with funding for HRM systems, workforce training, and change-management programs—otherwise the direct technology investments may yield limited output gains.
• Labor complementarity: Findings support a complementarity interpretation—AI augments human capital (better task allocation, recruitment, training effectiveness), which can mitigate fears of displacement in cultural-heritage industries and justify inclusive digitalization strategies.
• Heterogeneity across firms: Given varying firm sizes and product types, economic interventions should be tiered (micro vs small/medium) and sensitive to artisanal practices—scalable, low-cost AI modules (e.g., scheduling/predictive maintenance, decision support) may deliver highest ROI for smaller operations.
• Measurement and evaluation recommendations for economists: future cost–benefit analyses and macro scaling estimates (e.g., sectoral GDP gains) should account for:
  • The mediating role of HRM (include variables for HRM uptake/quality).
  • Implementation intensity (current low-to-moderate AI adoption suggests large upside but also barriers).
  • Non-financial values (cultural preservation, artisan skill retention) which affect welfare assessments.
• Research & policy priorities: larger-scale, longitudinal, and experimental studies (randomized rollouts, pre/post objective production measures) are needed to quantify effect sizes, causal durations, and generalizeability beyond one region. Cost-effectiveness and financing mechanisms for bundled AI + HRM interventions are high-priority topics for AI economics.

Limitations to keep in mind (relevant for economic interpretation): small single-region sample (n=47), cross-sectional self-reported measures (Likert scales) rather than objective production metrics, potential endogeneity and omitted-variable bias, and lack of reported path coefficients/standard errors in the excerpt. Replication with objective productivity data and larger, multi-region samples is necessary for robust economic scaling estimates.