Generative artificial intelligence and innovation performance in emerging-economy SMEs

Abstract

Generative artificial intelligence is reaching small firms in emerging economies with unusual speed, because it is rented, not built, and is operated through ordinary language instead of code. This paper develops a conceptual framework that explains when and how generative AI use improves innovation performance in emerging-economy small and medium-sized enterprises. Drawing on absorptive capacity, dynamic capabilities and the institution-based view, and on the resource-based view for the diffusion argument, the framework treats generative AI as a low-cost conduit to codified external knowledge whose innovation value depends on the firm's capacity to verify, transform and exploit what the technology produces. Eight propositions, two of them stated as pairs and so ten statements in all, specify the mechanisms and boundary conditions. Generative AI is expected to widen knowledge access most where external knowledge networks are weakest, to strengthen sensing and the design elements of seizing more than transforming, and to favour incremental product and service innovation, process innovation within existing operations, business model innovation and innovation speed over novel product innovation. Its effect is moderated at the firm level by the evaluative expertise of the owner-manager and core staff, and bounded by the type of institutional void the firm faces, by an infrastructure and affordability threshold, by the fit between the model's training data and the firm's language and market context, and by the stage of diffusion among competing firms. As use diffuses, advantage shifts from access to the combination of generative AI with idiosyncratic local knowledge. The paper treats generative AI as a substitute for the channels through which these firms have obtained external knowledge and a complement to the firm's own evaluative expertise, and sets out a research agenda for testing the framework.