Among Indian women entrepreneurs, favourable attitudes to AI consistently raise intentions to use AI and move nascent founders to action, but only established firms translate AI-driven behaviour into stronger investment and sales; perceived risk shows little explanatory power.

Main Finding

AI adoption operates as a stage-contingent capability for women entrepreneurs: positive AI attitudes reliably increase entrepreneurial intention in both new and established ventures, but the pathway from attitude→behaviour→performance differs by stage. AI attitude more directly drives entrepreneurial behaviour in new ventures, while AI perspective (strategic framing) and enacted AI-related behaviour have stronger effects on investment, purchasing, and sales in established ventures. Perceived risk has limited explanatory power.

Key Points

• Conceptualization: AI adoption orientation measured as three cognitive antecedents — AI attitude, AI perspective (strategic framing), and perceived risk — which influence entrepreneurial intention, then entrepreneurial behaviour, and finally business performance (investment, purchases, sales).
• Stage-contingency: Effects differ by venture maturity (GEM classification):
  • New women entrepreneurs (3–42 months): AI attitude has a stronger direct effect on entrepreneurial behaviour; intention → behaviour link is present.
  • Established women entrepreneurs (>42 months): AI perspective becomes more important (strategic implementation); entrepreneurial behaviour more strongly drives business performance.
• Mediating mechanisms: Entrepreneurial intention mediates attitude→behaviour links; entrepreneurial behaviour mediates the impact of AI orientation on business performance (supporting an attitude → intention → behaviour → performance chain).
• Perceived risk: Showed limited and inconsistent influence on intention, behaviour, and performance in this sample.
• Practical recommendation: Design stage-specific AI support — emphasize attitudinal interventions and experimentation support for early-stage women entrepreneurs; emphasize strategic integration, process redesign and capability scaling in established firms.

Data & Methods

• Sample: Cross-sectional survey of 194 women entrepreneurs in India (2023–24):
  • 108 New Women Entrepreneurs (NWEs; 3–42 months)
  • 86 Established Women Entrepreneurs (EWEs; >42 months)
  • Sectors included retail, fashion, digital services, manufacturing, healthcare, consultancy.
• Measures: 23-item questionnaire adapted from established entrepreneurial and technology-acceptance scales; 7-point Likert responses covering AI attitude (3 items), perspective (3), perceived risk (3), entrepreneurial intention (6), behaviour, and business performance (investment, purchases, sales).
• Sampling & power: Purposive sampling via networks; a priori G*Power calculation (f² = 0.15, α = 0.05, power = 0.80) supported group sizes.
• Analysis:
  • Measurement checks: reliability, validity, and measurement invariance (MICOM).
  • Structural modelling: Partial Least Squares Structural Equation Modelling (PLS-SEM).
  • Group comparison: Multi-Group Analysis (MGA) in SmartPLS to test differences between NWEs and EWEs.
• Limitations noted by authors: cross-sectional design (limits causal claims), purposive sampling, self-reported performance measures, and sample restricted to women entrepreneurs in India (affects external generalizability).

Implications for AI Economics

• Heterogeneous returns to AI: The value of AI adoption depends on venture stage. Empirical models of AI-driven productivity should incorporate firm maturity or organizational readiness as moderators rather than assuming homogeneous treatment effects.
• Targeting human capital interventions: Policies and programs that build positive AI attitudes (awareness, perceived usefulness, trial experiences) may yield larger marginal returns among early-stage firms, whereas investments in organizational processes, strategic integration, and scaling capabilities are likely higher-return for established SMEs.
• Measurement and microdata: Surveys and administrative datasets intended to quantify AI adoption and productivity impacts should collect measures of cognitive orientation (attitude, strategic perspective) and stage indicators to improve identification of mechanisms.
• Gendered diffusion and policy design: For reducing the gendered digital/AI gap, interventions should be stage-tailored (attitude and experimentation support for nascent women entrepreneurs; strategic advisory, procurement support, and implementation grants for established women firms).
• Cost-effectiveness of risk-reduction programs: Since perceived risk explained limited variance in this sample, blanket risk-reduction subsidies may be less cost-effective than programs that change attitudes and practical capacity to experiment with AI.
• Modelling implications: Structural and reduced-form econometric studies of AI adoption effects should consider mediation (intention → behaviour) and potential multi-step causal chains. Heterogeneous treatment effect estimation (by firm age/size/stage) will better capture the real-world distribution of AI benefits.
• Research priorities: Longitudinal and experimental work is needed to establish causal links, quantify productivity elasticities by stage, and test which policy levers (training, subsidized pilots, platform access) deliver the highest welfare-adjusted returns for women-led SMEs.

If you want, I can convert these points into a short policy brief or produce suggested survey items to measure AI attitude/perspective for use in microdata collection.