Year: 2026
Type: article
Study design: Systematic Review
Abstract: Generative Artificial Intelligence (GenAI), a technological trend that has swept the globe in recent years, has reshaped entrepreneurial innovation, yet no complete picture of GenAI research within startups exists in innovation studies. Systematic reviews of relevant findings scattered across disciplines remain scarce. Drawing on 39 papers published between 2022 and 2026 from the Web of Science and Scopus databases; this scoping review addresses three research questions: What roles does GenAI play throughout the startup innovation cycle? What benefits does GenAI bring and what problems arise when startups innovate with GenAI? What is the current state of research, and where are its shortcomings? Based on previous findings, a set of research gap hypotheses is proposed to guide future researchers in building a more integrated theoretical foundation for GenAI in entrepreneurship studies. Following Arksey and O'Malley's five-stage scoping review framework combined with PRISMA-ScR guidelines (Cohen's κ = 0.87), 2,470 papers were screened, yielding a final sample of 39. Analysis of these papers produced three main findings. First, under the Technology-Organization-Environment (TOE) framework, GenAI roles can be classified into three categories—automation, augmentation, and collaboration—each mapping onto a distinct phase of the startup innovation cycle. Second, an empowerment–vulnerability paradox emerges: GenAI lowers innovation costs and knowledge barriers for resource-poor startups, yet also introduces risks such as AI hallucinations, intellectual property exposure, data privacy vulnerabilities, and technology dependence. Third, comparison with prior studies reveals five specific knowledge gaps, among them the absence of process models and under-theorized boundary conditions. Given that the field is still evolving, a temporal dimension was incorporated into the gap analysis, yielding a tentative TOE+P framework extension—adding a Process dimension to the original TOE model—as a conceptual proposition requiring further empirical verification.
Institution: University of Technology Malaysia
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
Funder: Universiti Teknologi Malaysia
OpenAlex ID: W7165359522