Adaptive AI Governance Framework in the Territorial Context of a University Business Incubator: PRE–POST Perception Evidence, Stakeholder Heterogeneity, and Provisional Routing Rules

Abstract

Artificial intelligence (AI) is spreading across educational, productive, and institutional settings, creating a need for governance approaches that can adapt to heterogeneous stakeholders. This study proposes an adaptive AI governance framework for the territorial context of a university business incubator, in which the university acts as an articulating rather than a regulatory actor. The evidence comes from a PRE–POST questionnaire study of all 446 incubator affiliates in seven stakeholder groups, with illustrative records from 35 Metaplan working groups; between the two measurements, participants attended an AI training programme and discussed the questionnaire items in these groups. Perceived opportunities of AI governance (dz = 3.45) and the perceived feasibility of mitigating governance challenges (dz = 2.55) increased, but the single-group design, group deliberation, and possible response-shift, demand, and ceiling effects prevent attributing these changes to the programme. Baseline and post-assessment opportunity scores were strongly associated (β = 0.693), indicating relative stability; a positive association with baseline challenge salience is reported as exploratory. Stakeholder groups differed markedly at baseline and converged after the programme. The framework translates perception profiles into provisional routes—controlled scaling, regulatory strengthening, capacity building, and pilot redesign—whose thresholds were defined by the authors after the POST measurement, relative to the study population; controlled scaling additionally requires an independent risk assessment of the activity. The routes are a design proposition that requires validation against independent outcomes.