Organizations report strong perceived gains from AI decision-support systems—decision quality, perceived usefulness and efficiency score highest—especially when technical capability is matched by implementation readiness and reliable systems. These results reflect respondents' perceptions from a cross-sectional survey and do not establish causality.

Main Finding

A cross-sectional, case-based survey of 210 respondents found that an AI-based decision support model is associated with improvements in decision quality, operational efficiency, and perceived organizational performance. AI model capability, implementation readiness, and decision quality are key predictors: AI capability predicts decision quality (β = 0.31, p < .001); implementation readiness predicts operational efficiency (β = 0.29, p = .001); and decision quality predicts organizational performance (β = 0.34, p < .001). Correlations between core outcomes are strong (e.g., decision quality — organizational performance r = .73). The authors conclude AI decision support generates the most value when technical capability is paired with organizational readiness, reliable systems, and human oversight — acting mainly as augmentation of human judgment rather than replacement.

Key Points

• Sample and context: 210 respondents from cloud-enabled and enterprise-oriented organizations (private, public, hybrid); 108 business-system cases and 102 infrastructure-system cases.
• Measurement: structured five‑point Likert scales on constructs including AI Model Capability, Implementation Readiness, Trust in AI Recommendations, Perceived Usefulness, System Reliability, Decision Quality, Operational Efficiency, Organizational Performance.
• Average respondent ratings were positive across constructs; highest means:
  • Decision Quality: M = 4.25 (SD = 0.58)
  • Perceived Usefulness: M = 4.22 (SD = 0.57)
  • AI Model Capability: M = 4.18 (SD = 0.61)
  • Operational Efficiency: M = 4.11 (SD = 0.64)
  • Organizational Performance: M = 4.07 (SD = 0.66)
• Strong positive correlations among outcomes (notably Decision Quality ↔ Organizational Performance r = .73; Perceived Usefulness ↔ Organizational Performance r = .71; Decision Quality ↔ Operational Efficiency r = .69).
• Socio-technical framing emphasized: technology alone is insufficient; absorptive capacity, data quality, explainability/trust, process integration, and human oversight matter for real-world value.
• Sector relevance: findings framed as applicable to both business and infrastructure systems (e.g., predictive maintenance, service continuity, operations), though the published extract does not report subgroup heterogeneity estimates.

Data & Methods

• Design: Quantitative, cross-sectional, case-based survey study.
• Sample: N = 210 respondents across cloud-enabled and enterprise-oriented organizational cases (mix of private, public, hybrid); split roughly evenly between business (108) and infrastructure (102) cases.
• Instrument: Multi-item five‑point Likert scales for multiple constructs (AI capability, readiness, trust, usefulness, reliability, decision quality, efficiency, performance).
• Analysis: Descriptive statistics, Pearson correlations, and multiple regression models to test predictive relationships among constructs.
• Key statistical results:
  • Means and SDs for primary constructs (see Key Points).
  • Correlations: Decision Quality—Organizational Performance r = .73; Perceived Usefulness—Organizational Performance r = .71; Decision Quality—Operational Efficiency r = .69.
  • Regression coefficients: AI Model Capability → Decision Quality β = .31 (p < .001); Implementation Readiness → Operational Efficiency β = .29 (p = .001); Decision Quality → Organizational Performance β = .34 (p < .001).
• Limitations implied by design: cross-sectional survey (perceptual measures), limiting causal claims; potential selection/self-report biases; subgroup heterogeneity and objective outcome linkage not fully detailed in the provided excerpt.

Implications for AI Economics

• Complementarities matter: Economic returns to AI investments are likely contingent on complementary organizational inputs (implementation readiness, data governance, user training). Models of AI-driven productivity should incorporate complementarities between capital (AI systems) and organizational/skill capital.
• Augmentation vs. substitution: Evidence supports AI as an augmenting technology that raises decision quality and efficiency rather than outright replacing human judgment. Labor market models should emphasize task reallocation and skill-biased complementarities rather than pure displacement.
• Valuation and ROI: Positive perceived impacts on decision quality and organizational performance suggest potential value creation, but economic evaluations should seek objective performance and longitudinal evidence to quantify returns and payback periods.
• Adoption and diffusion: Trust, explainability, and system reliability are economic frictions affecting adoption rates and effective use. Policies or firm-level investments that lower these frictions (standards for explainability, training subsidies, interoperable data infrastructure) can increase realized returns.
• Policy and measurement: Regulators and policymakers aiming to capture social returns from AI in infrastructure (safety, service continuity) should incentivize socio-technical investments (data quality, governance) alongside deployment. Economic impact assessments should incorporate non‑market values (reliability, safety) especially in infrastructure domains.
• Research gaps for economics: Need for causal identification (randomized/quasi-experimental designs), objective performance metrics (productivity, costs, downtime), heterogeneous treatment effects by sector and firm size, and dynamic analyses of adoption trajectories and labor reallocation. These are necessary to move from perceived-effect associations to quantified economic impact estimates.