AI and Big Data lift productivity mainly in data-heavy retail firms while decision-making speeds up in finance and parts of IT; benefits vary sharply by industry and company size, and some IT adopters see little or negative gains.

Main Finding

Integration of AI and Big Data into IT management improves organizational productivity and decision-making on average, but effects are heterogeneous: data-rich sectors (notably retail) see the largest gains, while IT and finance show mixed outcomes. Firm size and depth of technological integration materially moderate benefits; weak or partial adoption can produce little to no improvement and sometimes slower progress.

Key Points

• Research questions: How do AI and Big Data affect (1) productivity in IT management, (2) decision-making speed and quality, and (3) do effects vary by industry and firm size?
• Heterogeneous impacts:
  • Retail and other data-dense industries experienced the clearest productivity and decision-quality improvements.
  • IT and financial firms showed more variable results—some large gains, some negligible or negative changes—depending on adoption level and implementation depth.
• Moderators identified: firm size (small/medium/large), industry type, and level of technology investment/implementation.
• Non-technical issues matter: data-privacy, fairness, ethical governance, and workforce skills/training affect realization of benefits.
• Practical audience: IT managers, strategic planners, and data-driven organizations seeking to improve performance and decision quality.

Data & Methods

• Sample: 150 companies targeted across IT, retail, education, and finance; 143 firms provided valid responses. Total survey responses used: 346.
• Firm-size categories: small (≤50 employees), medium (51–250), large (>250).
• Outcomes (dependent variables):
  • Productivity: number of tasks completed in a 6-month period.
  • Decision-making efficiency: speed (hours to decide) and quality (rated 1–5), supplemented by employee perception surveys.
• Key independent variables and controls:
  • Treatment dummy for AI/Big Data adoption (1 = adopted, 0 = not).
  • Time dummy (pre/post intervention).
  • Controls: firm size, industry type, technology investment.
• Identification strategy: Difference-in-Difference (DiD) design comparing treated vs. control firms before and after integration to estimate causal effects and reduce time-invariant confounding.
• Complementary methods: bibliometric/performance analysis and questionnaires to capture thematic relationships and perceptions.

Limitations noted by authors: - Reliance on survey/self-reported KPIs and perception measures. - Potential selection/implementation heterogeneity across firms. - Short-term pre/post observation and possible unobserved time-varying confounders.

Implications for AI Economics

• Productivity and returns to AI/Big Data are sector-dependent:
  • Expect higher marginal productivity gains in data-rich sectors (e.g., retail, customer-facing services).
  • Returns in IT and finance can be ambiguous without deep, well-designed integration.
• Complementarities matter: firm size, managerial practices, human capital, and investment in infrastructure determine realized gains—consistent with models where technology adoption alone is insufficient without complementary inputs.
• Policy and firm strategy:
  • For policymakers: foster data governance, privacy frameworks, and standards to reduce implementation frictions and ethical risks; support SMEs with subsidies or shared infrastructure to capture scale-dependent gains.
  • For firms: invest in training, robust data governance, and deeper integration (not just pilot deployments) to capture productivity returns.
• Labor and distributional considerations: heterogeneous effects imply uneven productivity and possibly unequal labor impacts across sectors and firm sizes; targeted retraining and transition policies may be needed.
• Measurement and evaluation: the study underscores the value of quasi-experimental methods (DiD) for causal assessment of AI interventions; ongoing evaluation and standardized KPIs are important to compare outcomes across contexts.
• Research directions: need for longitudinal, objective performance data (beyond self-reports), exploration of long-run equilibrium effects on productivity and employment, and deeper study of implementation quality as a mediator of returns.