UK firms that adopt generative AI report better business performance, largely because AI fuels product innovation rather than just efficiency gains; adoption is concentrated where technological competence and top-management backing exist, and is dampened by regulatory uncertainty.

Main Finding

Generative AI (GenAI) adoption is positively associated with firm-level business performance in UK firms. This effect is partially mediated by both operational efficiency and product innovation, with product innovation showing the stronger mediating role. Technological competence, top management support, and competitive pressure are key adoption drivers, while regulatory uncertainty is an important barrier.

Key Points

• Direct effect: Statistically significant positive relationship between GenAI adoption and business performance.
• Mediators: Operational efficiency and product innovation are both partial mediators; product innovation contributes more to the adoption → performance link than efficiency.
• Adoption drivers (TOE-aligned): Technological competence, top management support (organizational), and competitive pressure (environmental).
• Barrier: Regulatory uncertainty significantly reduces adoption likelihood.
• Theoretical grounding: Model developed from Resource-Based View (RBV) and Technology–Organization–Environment (TOE) frameworks; RBV explains GenAI as a strategic resource enabling competitive advantage, TOE identifies multi-dimensional adoption determinants.
• Originality: One of the first large-scale quantitative studies in the UK to empirically test mediating pathways from GenAI to firm performance.
• Limitations (implicit in design): Cross-sectional survey limits causal claims and is vulnerable to common-method bias; results are UK-specific and based on manager self-report, so external validation (longitudinal/objective metrics) would strengthen inference.

Data & Methods

• Sample: Cross-sectional survey of 312 senior managers across diverse industries in the United Kingdom.
• Measurement: Self-reported measures of GenAI adoption, business performance, operational efficiency, product innovation, and TOE factors (technological, organizational, environmental).
• Analytical approach: Partial least squares structural equation modeling (PLS-SEM) used to estimate direct and mediating effects and to test the conceptual RBV+TOE model.
• Inference: Mediation assessed within the PLS-SEM framework (report interprets product innovation as the stronger pathway).

Implications for AI Economics

• Firm-level productivity and growth: Evidence that GenAI elevates performance primarily through innovation suggests AI’s macroeconomic impact may run strongly through intangible capital accumulation and product-market improvements rather than cost-driven efficiency alone.
• Complementarity and skills: Stronger innovation effects imply complementarity between GenAI and skilled labor; policies should emphasize workforce upskilling to capture innovation gains and avoid skill-biased displacement.
• Measurement: Results reinforce the need for economics research to incorporate innovation-related outcomes (new products, business models) and intangible investment when estimating AI’s economic impact.
• Policy design: Reducing regulatory uncertainty and investing in digital infrastructure and AI-related human capital would likely increase beneficial adoption; targeted support may accelerate innovation-led productivity growth.
• Future research directions: Longitudinal and multi-source (administrative or financial) studies are needed to establish causal pathways, quantify magnitude of productivity gains, and assess distributional effects across firm sizes and sectors.