Stakeholder engagement and strategic innovation in higher education through AI competency

Synthesis: Nunez-Canal, de Obesso Arias, Perez-Rivero, and Alvarez-de-Mon (2026) ask how university faculty competencies and attitudes toward artificial intelligence affect student learning, and test four hypotheses on 105 business and economics professors recruited by snowball sampling. The instrument is an adaptation of the DigCompEdu digital competence framework to AI, reduced by factor analysis from 35 items into four factors, with a single 6-point self-report item as the dependent variable for perceived improvement in student learning. Only two hypotheses survive: professors' ability to leverage AI for feedback, Assessment, and Pedagogies and Teaching Strategies (B = 0.709, Beta = 0.483, p < 0.001) and their attitudes toward AI use, data governance, and ethical use (B = 0.608, Beta = 0.415, p < 0.001) are linked to perceived learning gains, while knowledge of regulatory frameworks (p = 0.113) and efforts to protect students from AI risks (p = 0.262) are not. The authors read the result through human capital theory and stakeholder theory: faculty are simultaneously the institution's most critical investment and its central stakeholder, so AI training is a strategic requirement rather than a discretionary cost.

Key Findings

• Two of four hypotheses are confirmed. Factor 1, teaching and learning, has the strongest influence (B = 0.709, Beta = 0.483, p < 0.001), followed by Factor 2, attitude toward the use of AI, data governance, and ethical use (B = 0.608, Beta = 0.415, p < 0.001). Factor 3, knowledge of the guidelines and regulatory framework for AI (p = 0.113), and Factor 4, protecting students from the impact of AI and its risks (p = 0.262), do not reach significance at the 0.05 level.
• The model fits. The regression reports a moderate multiple correlation and an adjusted R-squared of roughly four tenths, a Durbin-Watson statistic of 1.620, and an ANOVA F-statistic significant at the .001 level, which the authors read as support for independence, normality, linearity, and homoscedasticity.
• A substantial proficiency gap. 61% of professors rated their own AI proficiency as novice to integrator (1 to 3 on the 6-point scale in Q1). The authors contrast this with 37.5% reporting low digital competency in the same population before ChatGPT, and attribute the gap partly to accelerated digitalization during the COVID-19 pandemic.
• Attitudes are optimistic but ambivalent. 55.3% believed AI could enhance their pedagogical practice, and 51.5% were confident AI could coexist with their teaching role without undermining professional responsibilities. Yet 60.8% were concerned that AI might negatively affect student learning.
• Institutional support is widely seen as thin. 74.3% of respondents reported that their institution provided only minimal or moderate support for preparing teachers to use AI tools, while only 3.8% felt fully supported.
• The instrument is reliable. Cronbach's alpha across the 35 items is 0.977. The Kaiser-Meyer-Olkin measure is 0.935 and Bartlett's test of sphericity is significant (p = 0.000), with most item-pair correlations above 0.30.
• Four factors replace six domains. Principal components extraction with varimax rotation, using a 0.535 loading cutoff, groups the 35 items into four factors that explain 73.257% of variance: (1) teaching and learning, (2) attitude toward AI use, data governance, and ethical use, (3) knowledge of guidelines and regulatory frameworks, and (4) protecting students from AI's impact and risks.
• Highest loadings cluster in teaching. The strongest single loadings belong to two Area 5 items and an Area 4 item, with the teaching and learning items also loading on Factor 1.

Study Design

• Design and question. A quantitative survey study framed by the question of how professors' competency and attitudes toward AI contribute to student learning. The authors combine human capital theory (Becker, 1962) with stakeholder theory (Freeman, 1984) as a single analytical lens.
• Instrument. An adapted DigCompEdu questionnaire on AI-related digital competency, inspired by Ng et al. (2023), with 35 items across six areas of AI use in teaching, answered on a 6-point Likert scale from 1 to 6. Four additional items analyze variables of interest, and one dependent item, Q4, asks to what extent the use of AI tools improves student learning.
• The framework. The six areas are professional engagement, digital resources, teaching and learning, assessment, empowering students' AI learning, and developing students' AI literacy. The adaptation builds on the European Commission's DigComp and DigCompEdu, and on Ng et al.'s reformulation, which adds four dimensions of AI literacy: knowledge and understanding, application and use, evaluation and creation, and ethical perspective.
• Sample. 105 anonymous respondents, 62.9% men and 37.1% women, of whom 68.6% have teaching as their main activity. By qualification, 37.1% hold a bachelor's degree or equivalent professional qualification, 21% hold a doctorate, and 20.1% hold a doctorate with teaching accreditation. Teaching experience spans 18.1% with 1 to 5 years through 21% with more than 25 years; online teaching experience is concentrated at 1 to 5 years (67.6%).
• Recruitment. No comprehensive sampling frame existed, so chain-referral (snowball) sampling was used, a non-probability method in which initial respondents recruit others from their networks. Data collection ran from November to December 2023, with invitations distributed by email, LinkedIn, and WhatsApp. The questionnaire was administered in Spanish and translated into English for publication.
• Analysis. Descriptive analysis of Q1 to Q4, then factor analysis on the 35 items, then linear regression of the dependent variable on the four factors. The authors justify the single dependent item by a long tradition of using single items to capture complex constructs such as learning outcomes in the economics of education literature (Kifle, 2025).
• Ethics. The authors state that all relevant ethical considerations were addressed to guarantee data anonymity and compliance with regulations on data use, approval procedures, and transparency requirements of the participating universities.

What this means for practice

• Treat faculty AI training as a strategic investment, not a budget line, and build one program with three components. From a human capital perspective the authors argue that universities' investment in faculty AI competency is not discretionary, and that structured, accessible training tailored to AI competency is critical. Training should integrate technical, pedagogical, and ethical components rather than treating them as separate agendas, and should develop not only how to use AI but why, when, and under what conditions it is appropriate.
• Anchor programs in the framework's own language. The authors recommend aligning institutional programs with international guidance from UNESCO (2023) and the World Economic Forum (2025), and adopting an AI competency framework at institutional level to prioritize investment, connect training to actual AI use in courses, and monitor progress through explicit measurement.
• Use governance to protect without constraining. The non-significant results for regulatory knowledge and risk mitigation should not be read as evidence that governance is unimportant. The authors argue these dimensions are more distal to day-to-day teaching, and that governance should safeguard ethical use while creating enabling conditions for AI to support teaching and learning.
• Address the concern that AI harms learning. With 60.8% of respondents worried about negative effects on student learning, faculty development should engage those concerns directly rather than assume enthusiasm, and should include critical thinking, bias, and data privacy alongside operational mastery of tools.
• Measure and report. The authors call for measurement systems that let institutions demonstrate accountability to external Stakeholders and improve institutional effectiveness, and for policymakers to fund capacity-building and provide flexible regulatory frameworks that safeguard rights while leaving room for pedagogical innovation.

Limitations

• The study covers only business and economics faculty at a specific set of universities, which constrains generalizability to other disciplines, institutional types, and higher education systems.
• All measures are self-reported, so social desirability and perception bias may lead to over- or underestimation of actual proficiency and dispositions.
• The adapted DigCompEdu framework may not fully capture all dimensions of faculty readiness in rapidly evolving technological and regulatory environments.
• The analysis is restricted to educators' perspectives. The views of students, institutional leaders, and employers, central to a stakeholder-based understanding of AI integration, are absent.
• The dependent variable is a single self-report item, which cannot distinguish perceived learning improvement from measured learning gains, and no longitudinal or comparative evidence supports a causal reading.
• The authors' own suggested remedy is multi-stakeholder designs, comparison across regions, disciplines, and institutional profiles, and mixed methods or longitudinal approaches to examine the long-term effects of faculty AI training on student learning, organizational change, and institutional performance.

Citation

Nunez-Canal, M., de Obesso Arias, M. de las M., Perez-Rivero, C. A., & Alvarez-de-Mon, I. (2025). Stakeholder engagement and strategic innovation in higher education through AI competency. Journal of Innovation & Knowledge, 13, 100920.