Smart Innovation Hub: An AI-Enabled Information System for Challenge-Based Innovation and Capstone Project Matching in Higher Education

Abstract

Artificial intelligence (AI) and digital platforms are increasingly influencing how universities manage experiential learning, interdisciplinary collaboration, and innovation-oriented educational activities. Challenge-based capstone and graduation projects play an important role in this context because they connect technical learning with teamwork, stakeholder engagement, project management, and applied innovation. However, many universities still rely on fragmented and highly manual coordination processes, which can limit scalability, transparency, and effective alignment between project requirements and participant capabilities. This study presents Smart Innovation Hub, an AI-enabled information system developed to support challenge-based innovation and capstone-project coordination in higher education. The platform brings together challenge intake, participant profiling, AI-supported recommendations, mentor coordination, workflow governance, and human review within a shared educational innovation environment. The system operationalizes an Innovation Bridge ecosystem model that connects students, faculty mentors, research centers, and external partners through a data-supported coordination framework. A Design Science Research (DSR) methodology guided the development and pilot evaluation of the platform within a public university environment. The pilot evaluation relied on several evidence sources, including platform logs, coordinator records, stakeholder surveys, milestone documentation, and partner feedback collected during implementation activities. Early pilot observations suggested an approximate 60% reduction in average team-formation cycle time, together with positive stakeholder perceptions regarding workflow usability and recommendation quality. These findings should be interpreted as preliminary implementation indicators within a single-institution pilot environment. The study contributes an AI-enabled educational innovation ecosystem architecture, a hybrid semantic-structured recommendation framework for challenge-based coordination, and a structured workflow model that integrates explainability and human oversight into educational innovation management. The findings further suggest that AI-enabled information systems may improve the transparency and coordination of challenge-based innovation workflows while preserving institutional governance and human decision-making.

1. Introduction

Engineering education is increasingly expected to prepare graduates for complex, interdisciplinary, and rapidly changing professional environments. Employers and public-sector partners expect graduates to demonstrate not only technical competence, but also problem framing, teamwork, communication, ethical judgment, digital fluency, innovation capability, and the ability to convert uncertain needs into implementable solutions. Capstone and graduation projects are central to this mission because they provide structured opportunities for students to integrate technical knowledge with professional practice. When such projects are connected to real industry, government, research, and community challenges, they become more than course requirements; they become institutional mechanisms for innovation, applied research, talent development, and university–industry collaboration.

Despite their educational value, challenge-based capstone systems are difficult to coordinate at scale. In many programs, project assignment and team formation are still handled through spreadsheets, email exchanges, faculty meetings, and informal matchmaking [1]. Coordinators must simultaneously consider student preferences, skills, disciplinary backgrounds, schedule compatibility, mentor capacity, partner expectations, confidentiality restrictions, project duration, available laboratory resources, and institutional priorities. These trade-offs are often implicit and difficult to document. As a result, the process can be slow, inconsistent, and vulnerable to misalignment between project requirements and team capabilities. Delayed team formation can also reduce the time available for problem discovery, stakeholder interviews, prototyping, validation, and reflection within fixed academic calendars [1,2,3].

The rise of artificial intelligence (AI), learning analytics, semantic technologies, and educational information systems creates new opportunities to redesign challenge-based educational workflows. AI-enabled systems can support educational decision-making by analyzing structured and unstructured information, identifying hidden relationships, and generating explainable recommendations [4,5,6]. However, educational coordination and capstone governance are not purely technical optimization problems. Matching students, mentors, and partner-defined challenges involves contextual judgment, fairness considerations, interdisciplinary balance, institutional priorities, and educational accountability. Consequently, AI should not function as an autonomous decision authority within educational innovation ecosystems. Instead, AI should operate as a human-centered decision-support layer that improves transparency, structures complex information, and provides explainable recommendations while preserving academic oversight and institutional governance [7,8,9]. This perspective aligns with emerging research on responsible AI, explainable educational systems, and human-supervised review frameworks in higher education environments.

This paper presents Smart Innovation Hub, an AI-enabled platform developed to support challenge-based engineering education by connecting partner-defined challenges with students, faculty mentors, and institutional innovation resources. The platform supports challenge intake, participant profiling, AI-supported recommendations, mentor coordination, workflow review, and project monitoring within a unified educational innovation environment. The platform was developed as part of a broader Innovation Bridge model that aims to connect universities with industrial, governmental, community, and research sectors by converting real-world challenges into implementable projects.

The Innovation Bridge model expands the traditional capstone-assignment process into a broader institutional innovation ecosystem. In addition to supporting capstone coordination, the platform allows users to manage ideas, projects, research outputs, challenges, and patents within a shared digital environment. Users can upload supporting materials such as technical documentation, datasets, prototypes, and visual evidence while also connecting initiatives with external partners, funding opportunities, and commercialization pathways. This broader ecosystem perspective distinguishes Smart Innovation Hub from conventional capstone-management systems by integrating educational coordination, innovation governance, interdisciplinary collaboration, and commercialization-oriented workflows within a unified institutional framework [4,10,11].

Conventional learning-management systems are not designed to support the full complexity of challenge-based capstone coordination. While such systems are effective for course delivery, grading, and communication, innovation-oriented educational workflows require additional capabilities such as structured challenge intake, profile-based opportunity discovery, semantic recommendation, mentor-capacity management, workflow tracking, evidence capture, partner feedback, and commercialization-oriented coordination. These requirements motivated the development of a more specialized information-system architecture tailored to multidisciplinary innovation workflows.

This study is motivated by four interrelated operational and research challenges:

• Traditional capstone-assignment processes become increasingly difficult to coordinate as the number of projects, stakeholders, disciplines, and external partners expands [12,13,14,15].
• Multidisciplinary team formation requires balancing technical competencies, disciplinary diversity, stakeholder expectations, scheduling constraints, mentor capacity, and participant preferences within limited academic timelines [14,15,16].
• Existing educational platforms primarily emphasize course delivery and learning management rather than end-to-end innovation workflow orchestration, including challenge intake, AI-supported matching, milestone governance, partner engagement, and outcome evaluation [4,6,10].
• Many AI-in-education studies focus on personalized learning and content recommendation, whereas comparatively fewer studies investigate AI as a governance-oriented decision-support framework for challenge-based educational innovation ecosystems [7,8,9].

1.1. Research Questions

• Q1: To what extent can an AI-driven innovation platform reduce capstone team-formation cycle time and administrative effort compared with a prior manual process?
• Q2: How can semantic similarity, structured attributes, and human-in-the-loop review be combined to support transparent matching among challenges, students, mentors, and institutional resources?
• Q3: How can project quality, partner impact, stakeholder satisfaction, and student learning outcomes be operationalized for evaluating an AI-driven challenge-based engineering education platform?

1.2. Contributions

The study contributes to the research on AI-enabled educational information systems in several ways. First, it presents an operational innovation platform designed to support challenge-based engineering education and multidisciplinary capstone coordination. Second, it introduces a hybrid recommendation framework that combines semantic similarity, structured constraints, interdisciplinary team composition, and human-in-the-loop coordination mechanisms within a unified coordination process. Third, the study proposes a structured workflow and data-management model capable of supporting partner-defined challenges, participant profiling, mentor coordination, milestone tracking, evidence capture, and commercialization-oriented activities across institutional innovation workflows. Finally, the paper presents a preliminary pilot-evaluation framework for examining workflow efficiency, recommendation transparency, stakeholder satisfaction, and implementation-oriented educational outcomes within AI-supported innovation environments.

2. Related Work

The literature relevant to the Smart Innovation Hub spans several domains, as illustrated in Figure 1.

2.1. Digital Transformation and AI-Driven Innovation in Higher Education

Digital transformation in higher education is moving universities beyond traditional course delivery. Increasingly, institutions are expected to operate as data-enabled ecosystems that support personalized learning, interdisciplinary collaboration, innovation, and external engagement. Recent work on AI-driven research and simulation-based inquiry argues that AI is changing both educational tools and research practice. It can support literature discovery, scenario exploration, model construction, and the analysis of complex systems [4]. This broader methodological shift is relevant to engineering education because capstone projects often require students to investigate uncertain, real-world problems where simulation, data analysis, and iterative validation are central. AI-driven innovation in higher education also has implications for institutional strategy. Universities are increasingly expected to use digital platforms for coordination and resource management. These systems also help connect academic capabilities with innovation, entrepreneurship, and external stakeholder engagement activities. Recent studies on AI-enabled innovation and entrepreneurship education identify several functions that are particularly relevant to the Smart Innovation Hub. These include management support, platform integration, and process optimization activities [10]. In that work, AI is framed as a mechanism for data-driven management, precision service, and dynamic regulation. Other studies examining AI integration in innovation and entrepreneurship education report that AI can improve teaching strategies, personalize instruction, and strengthen entrepreneurial skill development. These benefits are more effective when AI is embedded within a broader educational framework rather than introduced as an isolated tool [11]. Although prior studies demonstrate the value of AI in educational personalization, institutional digital transformation, and innovation-oriented learning environments [4,10,11], comparatively fewer studies investigate AI as a workflow coordination framework for multidisciplinary challenge-based innovation ecosystems. Existing work primarily focuses on teaching enhancement, adaptive learning, or institutional digitalization, whereas fewer systems integrate partner coordination, interdisciplinary matching, workflow governance, and human-in-the-loop educational oversight within a unified operational platform [7,8,9]. The Smart Innovation Hub addresses this gap by combining AI-supported recommendations with workflow orchestration and workflow coordination. In this way, the platform moves beyond teaching enhancement and supports challenge-based engineering education as an institutional innovation process.

2.2. Learning Analytics and Data-Driven Educational Decision-Making

Learning analytics and AI in higher education provide a foundation for data-supported decision-making. A systematic review of data-driven learning analytics and AI in higher education found that many educational AI applications rely on limited data such as access logs or manually recorded grades, which may not fully capture cognitive processes or support meaningful educational interventions [6]. The review also highlights the importance of stronger data governance, valid evaluation metrics, ethical safeguards, and alignment between AI models and educational objectives. This observation is important for the Smart Innovation Hub because the platform captures richer coordination and workflow data than conventional course-management systems: challenge briefs, required skills, student profiles, mentor expertise, matching decisions, coordinator overrides, milestones, evidence artifacts, partner acceptance, and outcome indicators.

AI-based educational evaluation research also demonstrates the value of multi-criteria decision-making in complex learning environments. Studies evaluating smart education practices in journalism and media education use multi-criteria frameworks to assess teaching quality, academic innovation, student engagement, and the integration of intelligent technologies [17]. Similar work in English translation education uses AI-driven insights and multi-criteria evaluation to assess AI-powered teaching alternatives according to educational criteria such as semantic accuracy, adaptability, engagement, and alignment with curriculum standards [18]. Although these studies focus on different disciplines, they support the view that educational AI systems should be evaluated using multiple dimensions rather than a single accuracy metric.

Existing learning-analytics research emphasizes predictive modeling, student monitoring, and performance evaluation [6,17,18]. However, fewer studies examine how educational data infrastructures can support institution-wide coordination of challenge-based innovation workflows involving external stakeholders, interdisciplinary teams, and governance-oriented matching processes. Most educational analytics systems remain focused on learner assessment rather than operational innovation orchestration. The Smart Innovation Hub extends learning analytics beyond classroom monitoring toward AI-supported coordination, workflow governance, and institutional innovation management.

2.3. Adaptive Learning, Personalization, and Recommendation Systems

Adaptive learning systems have become a major theme in AI-enabled education. A systematic review of AI-driven adaptive learning systems in mobile education identified personalization strategies, effectiveness measures, and user-interaction patterns as core themes in the field [5]. The review shows that AI and personalized learning now connect several educational research domains. It also emphasizes the importance of longitudinal evidence, ethical considerations, and stronger theoretical integration. Research on generative AI in digital education similarly emphasizes personalized learning, automated content generation, real-time assessment, and AI–human collaboration, while warning about bias, privacy, academic integrity, and the risk of reducing the teacher’s role to a tool supervisor [8].

Recommendation systems in education are not limited to recommending content. They can recommend learning paths, exercises, feedback, collaborators, mentors, or project opportunities. For example, AI-driven recommendation for personalized physical education training demonstrates how AI can adapt guidance based on individual performance data and provide real-time corrective feedback [19]. Although physical education differs from engineering capstone education, the underlying idea is similar: AI becomes useful when it converts learner data into actionable recommendations that support human development.

While adaptive learning and recommendation systems have primarily focused on personalized educational content, learner support, and instructional adaptation [5,8], challenge-based engineering education introduces additional coordination requirements involving multidisciplinary team composition, mentor alignment, partner expectations, and workflow governance. The Smart Innovation Hub extends educational recommendation systems toward collaborative innovation coordination by integrating semantic matching, structured constraints, interdisciplinary balancing, and human-in-the-loop decision support within a broader university-level framework.

2.4. Capstone Design, Project-Based Learning, and Challenge-Based Learning

Capstone design and project-based learning are widely recognized as important mechanisms for developing professional and higher-order skills. They encourage students to apply knowledge to authentic problems, engage stakeholders, manage uncertainty, and produce tangible outputs. In software project management education, recent work shows that prompt engineering and large language models can support project-based learning by acting as virtual specialists, providing real-time guidance, and helping students produce documentation aligned with professional frameworks such as PMBOK and Scrum [20]. That study is important because it shows that AI can support project-based learning in several ways. Beyond content delivery, AI can also assist with decision-making, documentation, and project execution activities.

Generative AI research in programming education provides another relevant perspective. A systematic review of generative AI integration in programming education found that successful adoption depends on planned teaching strategies, thoughtful assessment design, structured integration, educator training, and attention to students’ higher-order thinking skills [9]. This is important for the Smart Innovation Hub because the platform must not reduce capstone learning to automated matching. The educational value of capstone projects depends on student agency, problem-solving, teamwork, communication, and reflection. AI should therefore be embedded as a support layer that improves coordination and decision transparency while preserving student ownership of project work.

Project-based learning also requires inclusive and multidisciplinary collaboration. Work on multi-, inter-, and transdisciplinary sustainable built-environment research highlights the importance of inclusive research platforms that connect diverse expertise, remove barriers to participation, and support socially relevant problem solving [16]. Although that volume focuses on the built environment, its emphasis on interdisciplinary collaboration and accessibility aligns with the goals of challenge-based engineering education.

Prior research demonstrates that project-based and challenge-based learning environments improve professional competencies, experiential learning, collaboration, and stakeholder engagement [9,16,20]. However, comparatively fewer studies provide operational AI-enabled system architectures capable of managing challenge intake, interdisciplinary matching, partner coordination, workflow governance, and longitudinal project monitoring at institutional scale. The Smart Innovation Hub addresses this operational and organizational gap through an integrated educational innovation ecosystem that combines recommendation intelligence, coordination mechanisms, and structured workflow orchestration.

2.5. Capstone Project Assignment and Allocation Mechanisms

Earlier research has proposed algorithmic approaches for assigning students to capstone projects. Freiheit and Wood described a project assignment algorithm based on student bidding and satisfaction scoring, showing that algorithmic support can reduce assignment time compared with manual processes [12]. Magnanti and Natarajan applied discrete optimization to allocate students to multidisciplinary capstone projects under discipline-composition constraints [13]. Together, these studies show that capstone assignment represents a structured decision problem involving preferences, capacity limitations, operational constraints, and fairness considerations. However, traditional assignment algorithms often focus on matching students to projects within a single course. They may not capture the broader innovation ecosystem, including partner challenge intake, mentor matching, institutional resources, intellectual property considerations, milestone tracking, and partner outcomes.

Traditional capstone-assignment studies primarily focus on student-project allocation within isolated course environments [12,13]. In contrast, the Smart Innovation Hub expands the matching problem into a broader educational innovation ecosystem that includes partner-defined challenges, mentor allocation, institutional resources, milestone governance, commercialization pathways, and explainable AI-supported coordination. This broader ecosystem perspective extends conventional assignment optimization toward governance-oriented innovation orchestration.

2.6. Team Formation and Teamwork Assessment Tools

Engineering education has a long tradition of research on team formation and teamwork assessment. The CATME/Team-Maker ecosystem provides tools for forming teams using instructor-selected criteria such as schedule compatibility, skills, prerequisites, preferences, and peer-evaluation considerations [14,15]. This literature emphasizes that team formation should consider not only technical capability but also interpersonal factors, availability, diversity of perspectives, and fairness.

Existing team-formation research emphasizes balancing technical competencies, diversity, schedules, and collaboration quality [14,15]. However, many existing systems rely primarily on structured inputs and do not fully integrate semantic challenge interpretation, structured workflows, external stakeholder coordination, or explainable recommendation mechanisms. The Smart Innovation Hub extends this literature by combining semantic similarity, structured constraints, interdisciplinary balancing, and human-governed AI recommendation within a unified university-level coordination framework.

2.7. AI in Innovation and Entrepreneurship Education

Innovation and entrepreneurship education is a key context for the present work. Studies on AI and innovation education argue that AI can support entrepreneurship programs by improving management processes, integrating resources, personalizing learning, and strengthening evaluation [10,11]. Zhang and Li’s case study in materials science specialties proposes a practical framework in which AI supports project management, resource scheduling, and educational evaluation [10]. Lu Zhang’s study similarly argues that AI integration can enhance innovation and entrepreneurship education by tailoring instruction, improving teaching strategies, and fostering entrepreneurial skills [11].

Although prior research highlights the potential of AI to improve entrepreneurship education, resource coordination, and innovation management [10,11], comparatively fewer studies provide detailed operational implementations linking AI-enabled educational coordination with institutional innovation ecosystems, commercialization pathways, and challenge-based engineering education. The Smart Innovation Hub contributes to this area by operationalizing AI-supported innovation governance within an institutional educational environment.

2.8. AI-Assisted Project Management and Strategic Decision-Making

Capstone projects are educational projects, but they also share many characteristics with professional project management. They involve scope definition, stakeholder expectations, resource allocation, schedules, risks, milestones, documentation, and performance tracking. AI-driven project management research shows that predictive analytics and decision-support systems can improve risk identification, resource distribution, and performance monitoring [21]. These capabilities are relevant to capstone education because project delays, unclear scope, and resource mismatches can reduce learning quality and partner satisfaction.

Existing AI-assisted project-management research primarily focuses on predictive analytics, scheduling optimization, and performance monitoring [21]. In contrast, challenge-based educational innovation environments require additional governance functions involving interdisciplinary coordination, academic oversight, stakeholder communication, and milestone-based educational evaluation. The Smart Innovation Hub extends AI-supported project coordination toward educational innovation governance and institutional workflow orchestration.

2.9. Semantic Matching, Explainability, and Human–AI Collaboration

The matching problem addressed in this paper requires both structured and semantic reasoning. Structured attributes such as discipline, skills, availability, and mentor capacity can be represented explicitly. However, challenge descriptions, student interests, prior project narratives, and mentor expertise often appear as unstructured text. Natural language processing and transformer-based embedding techniques allow the platform to represent textual information as vector-based features. These representations help estimate semantic similarity between challenges, participant profiles, and mentor expertise.

At the same time, educational AI systems must be explainable and governed. Research on explainable AI highlights the importance of transparency, interpretability, and responsible human oversight when AI systems influence decisions [7]. Generative AI education reviews further warn that successful AI integration depends on curriculum alignment, accessibility, bias mitigation, and careful assessment design [8,9].

The reviewed literature highlights the importance of transparency, accountability, explainability, and human oversight in AI-supported educational systems [7,8,9]. The Smart Innovation Hub operationalizes these principles through explainable recommendation outputs, coordinator override mechanisms, audit logging, and structured workflow controls. Rather than replacing academic decision makers, the platform positions AI as a transparent decision-support layer within a human-governed educational innovation ecosystem.

2.10. AI Governance, Human–AI Collaboration, and Responsible Educational AI

Recent advances in generative artificial intelligence and AI-enabled information systems have shifted research attention from isolated automation toward socio-technical governance, human–AI collaboration, explainability, and accountable decision-support ecosystems [22,23,24,25,26]. Contemporary AI-enabled organizational systems are increasingly conceptualized not only as technical artifacts, but also as socio-technical infrastructures that require continuous alignment between technical capabilities, organizational procedures, human oversight, and institutional accountability [22,24,25]. Within educational environments, AI systems introduce additional governance challenges because educational coordination decisions involve fairness, transparency, stakeholder trust, academic accountability, and multidisciplinary collaboration. Prior research on human–AI collaboration in learning environments emphasizes that AI should support learner agency, group coordination, and collaborative regulation rather than replace human judgment [23]. Similarly, equity-centered AI education research highlights the need to integrate fairness and equity considerations early in the design of AI-supported educational systems rather than treating them only as post hoc evaluation concerns [26].

Recent DSR-based AI governance studies also demonstrate the importance of embedding auditability, role accountability, validation controls, and workflow-coordination mechanisms into AI-enabled decision-support systems [24,25]. These principles are particularly important in challenge-based educational innovation ecosystems, where AI-supported recommendations may influence team formation, mentor alignment, challenge prioritization, and resource coordination. The Smart Innovation Hub operationalizes these governance principles through explainable recommendation outputs, coordinator override mechanisms, workflow audit trails, role-based access control, and human-in-the-loop review processes that preserve academic oversight while improving operational coordination efficiency.

2.11. Research Gap and Positioning of Smart Innovation Hub

Taken together, the reviewed studies highlight the increasing role of AI-enabled educational systems, learning analytics, recommendation mechanisms, project-based learning, innovation ecosystems, and governance-oriented AI frameworks within higher education environments [4,5,6,8,9,10,11,16,17,18,19,20,21]. At the same time, the literature remains fragmented across these research areas. Many existing studies focus on isolated educational functions such as adaptive learning, student-project assignment, tutoring systems, or AI-assisted assessment. Comparatively fewer studies examine integrated institutional ecosystems capable of supporting multidisciplinary challenge coordination, explainable recommendation processes, workflow traceability, governance-oriented oversight, and human-supervised educational innovation management within a unified implementation environment [22,23,24,25,26]. Several gaps remain, as follows:

• Many AI-in-education studies focus on content personalization, tutoring, or assessment rather than the organizational problem of forming multidisciplinary teams around real-world challenges.
• Capstone assignment studies often address student-project allocation but do not integrate partner challenge intake, mentor matching, institutional resources, staged execution, and outcome capture.
• Innovation and entrepreneurship education research emphasizes AI-supported management and platform construction but provides fewer detailed designs and evaluations of working systems for challenge-based engineering education.
• Many systems do not explicitly combine semantic matching, structured constraints, explainability, and coordinator oversight.

To address these gaps, the Smart Innovation Hub was developed as an applied AI-enabled educational innovation platform capable of supporting challenge-based coordination, interdisciplinary collaboration, and structured workflow management across institutional innovation activities.

Table 1. Literature synthesis and implications for the Smart Innovation Hub.

Literature Stream	Key Insight	Implication for Smart Innovation
AI-driven research and methodology	AI can support complex inquiry, scenario exploration, and research workflows [4].	Treat platform data as a basis for continuous evaluation and simulation of project workflows.
Learning analytics in higher education	AI systems require valid metrics, data governance, and alignment with educational goals [6].	Capture rich data beyond grades, including matching decisions, milestones, evidence, and outcomes.
Adaptive and personalized learning	AI can personalize learning and support real-time feedback, but needs ethical safeguards [5,8].	Extend personalization from learning content to project–team–mentor recommendation.
AI-based educational evaluation	Multi-criteria frameworks are useful for evaluating complex educational quality [17,18].	Evaluate the platform using efficiency, satisfaction, match quality, partner impact, and learning outcomes.
Innovation and entrepreneurship education	AI supports management empowerment, platform construction, and process optimization [10,11].	Position the Smart Innovation Hub as an innovation ecosystem platform, not only a capstone tool.
Project-based learning and project management	AI and prompt engineering can scaffold project documentation, decision-making, and mentorship [20,21].	Integrate workflow support, milestones, mentor alignment, and outcome capture.
Team formation and capstone assignment	Matching should balance preferences, skills, schedules, disciplines, and fairness [12,13,14,15].	Use hybrid scoring with hard constraints, soft objectives, and human oversight.
Explainable and responsible AI	AI decisions must be transparent, auditable, and governed [7,9].	Provide explainable recommendations and coordinator override logs.

summarizes how the reviewed literature informs the platform design.

3. Research Methodology and Evaluation Design

This research adopted a Design Science Research (DSR) methodology to guide the development and evaluation of the Smart Innovation Hub. The DSR approach was selected because the study focuses on designing, implementing, and evaluating an operational information-system artifact intended to address a real coordination problem within challenge-based engineering education and institutional innovation management [11,27,28,29,30,31,32]. The methodology combines artifact development, stakeholder-informed refinement, pilot deployment, and iterative evaluation within an authentic educational environment. The research process followed the DSR logic proposed in information systems research, particularly the stages of problem identification, objective definition, artifact design and development, demonstration, evaluation, and communication [30,31,32]. In this study, these stages were operationalized through identifying coordination limitations in traditional capstone workflows, defining platform requirements with stakeholders, developing the AI-enabled artifact, deploying it in a university pilot environment, evaluating it using operational and stakeholder data, and refining the platform based on feedback.

Figure 2 illustrates the DSR framework adopted for the Smart Innovation Hub. The framework links the educational innovation environment, the artifact-development process, and the knowledge base used to guide platform design, pilot deployment, evaluation, and refinement. The iterative design cycle supports repeated build–demonstrate–evaluate–refine activities, consistent with contemporary DSR practices in educational information systems and AI-enabled governance research [22,23,24,25,30,31,32].

3.1. Design Science Research Framework

The artifact developed in this study represents an AI-enabled educational innovation ecosystem designed to support challenge intake, participant coordination, semantic recommendation, workflow governance, milestone tracking, and human-supervised educational decision-making. Rather than functioning solely as a technical matching tool, the Smart Innovation Hub was designed to support broader university-level coordination activities involving multidisciplinary collaboration, innovation management, and structured educational workflows.

The DSR approach also supported the development of the platform as a socio-technical system that combines technical functionality with institutional review and educational coordination requirements. This perspective is important because AI-enabled educational systems must not only generate recommendations but also preserve human oversight, transparency, accountability, and institutional decision authority [7,8,9,22,23,24,25,26]. Accordingly, the Smart Innovation Hub was designed to position AI as a decision-support layer rather than an autonomous placement authority.

Recent DSR studies in educational AI and organizational information systems demonstrate the value of combining artifact development with iterative evaluation, stakeholder feedback, traceability, and governance-oriented refinement [23,24,25,26]. These principles informed the design of the Smart Innovation Hub, particularly its explainable recommendation outputs, coordinator override mechanisms, workflow audit trails, and role-based governance controls.

3.2. Research Process and Development Stages

The research process began with identifying operational limitations in traditional capstone-project coordination workflows. These limitations included fragmented communication, manual team formation, inconsistent mentor allocation, limited visibility of participant capabilities, weak integration with external stakeholders, and insufficient tracking of project evidence and outcomes.

Platform requirements were identified through discussions and coordination activities involving capstone coordinators, faculty mentors, innovation-center personnel, students, research entities, and external collaborators participating in challenge-based educational initiatives. These interactions helped define both the technical and governance-oriented requirements of the platform, including challenge intake, participant profiling, recommendation transparency, mentor coordination, workflow visibility, milestone tracking, and human-supervised review mechanisms.

Based on these operational and governance requirements, the Smart Innovation Hub was developed as a modular AI-enabled information system integrating challenge coordination, participant profiling, recommendation support, workflow management, milestone tracking, evidence capture, and academic oversight within a unified platform environment. The development process emphasized interoperability, explainability, role-based access, and human oversight to ensure that AI-supported recommendations remained aligned with educational objectives and institutional review requirements [7,8,9,22,23,24,25,26,31].

Following implementation, the platform was deployed within a pilot university environment involving challenge-based engineering projects, multidisciplinary student teams, faculty mentors, research centers, and external partners. Pilot observations and stakeholder feedback were used iteratively to refine workflows, recommendation criteria, governance controls, interface design, and recommendation transparency mechanisms.

3.3. Stakeholder Involvement and Pilot Environment

The development and pilot implementation of the Smart Innovation Hub involved several stakeholder groups, including students, faculty mentors, innovation-center personnel, research entities, coordinators, and external collaborators participating in challenge-based educational activities. Stakeholders contributed to requirement identification, workflow refinement, usability feedback, recommendation review, and pilot evaluation activities throughout the development process. The pilot implementation was conducted within a public university engineering and innovation environment between 2023 and 2025. The pilot included challenge-based capstone projects, interdisciplinary engineering teams, faculty mentors, research centers, and external partner organizations. The pilot environment involved 60 submitted projects or challenges, 180 participating students, 90 faculty mentors, 5 external partners, and 9 active pilot teams. The pilot should be interpreted as an observational institutional implementation rather than a controlled experimental study. Its primary purpose was to evaluate operational feasibility, workflow coordination, recommendation transparency, stakeholder perceptions, and governance-oriented educational coordination within a real educational environment.

This exploratory evaluation approach is consistent with DSR-based educational information systems research in which early stage artifacts are evaluated through pilot deployment, stakeholder feedback, and operational evidence [24,25,26].

3.4. Data Collection and Evaluation Strategy

The evaluation framework combined both quantitative and qualitative evidence to examine workflow usability, recommendation relevance, stakeholder satisfaction, operational feasibility, and governance transparency within the pilot environment. Evaluation data were collected from platform logs, coordinator records, stakeholder surveys, milestone documentation, recommendation outputs, and partner feedback activities. Quantitative evaluation focused primarily on operational indicators such as team-formation duration, recommendation acceptance, workflow progress, and stakeholder-satisfaction measures. Qualitative evaluation emphasized coordinator observations, participant experiences, and partner feedback related to collaboration quality, recommendation usefulness, workflow transparency, and interdisciplinary coordination practices. Because the study focused on a real institutional deployment rather than a controlled experimental environment, the findings should be interpreted as exploratory implementation observations rather than generalized causal evidence. This approach is consistent with exploratory DSR evaluation practices commonly used in educational information systems research [23,25,26,30,31,32]. Moreover, the evaluation process also considered governance-oriented indicators such as recommendation traceability, coordinator override capability, workflow visibility, and human oversight effectiveness. These dimensions were included because the Smart Innovation Hub was designed not only as a technical recommendation system but also as a governance-oriented educational coordination infrastructure.

3.5. Survey Instrument, Reliability, and Validity

Stakeholder surveys were used to evaluate workflow usability, recommendation relevance, transparency, satisfaction, and perceived coordination effectiveness. The survey instrument included Likert-scale items addressing workflow usability, recommendation quality, transparency, coordination efficiency, interdisciplinary collaboration, and overall platform satisfaction. To improve content validity, the survey instrument was reviewed by academic experts in engineering education, innovation management, and educational information systems prior to deployment [30,31,32]. Minor wording revisions were implemented following expert feedback to improve clarity, interpretability, and alignment with the study objectives. Internal consistency reliability was assessed using Cronbach’s alpha for the primary survey constructs. The resulting values exceeded commonly accepted thresholds for exploratory educational research, indicating acceptable internal consistency of the instrument. Because the pilot represented an early stage institutional implementation with a limited observational sample, the evaluation results should be interpreted as preliminary indicators rather than generalized causal findings. Future research should include longitudinal assessment, multi-institutional validation, and expanded statistical evaluation to further test the generalizability, reliability, and educational impact of the platform.

3.6. Ethical and Governance Considerations

Because the platform processes participant profiles, recommendation outputs, and project-related information, governance and privacy considerations were incorporated into the system’s design. Recommendation scoring was restricted to fields relevant to educational coordination objectives, while sensitive information was protected through role-based access controls and institutional permission management. The platform incorporated human-supervised review mechanisms to ensure that AI recommendations remained advisory rather than autonomous. Coordinators retained authority to review recommendations, adjust weights, override system suggestions, and document decision rationales. Recommendation logs and workflow records were preserved to support transparency, auditability, and accountability in educational decision-making [7,8,9,25,26,30,31,32].

Recent responsible-AI and governance-oriented information-systems research emphasizes that educational AI systems should preserve fairness, transparency, explainability, and academic oversight rather than replacing human judgment [25,26,32]. The Smart Innovation Hub operationalizes these principles through explainable recommendation mechanisms, structured workflow controls, audit trails, and human-supervised coordination processes.

4. Smart Innovation Hub Platform

This section presents the Smart Innovation Hub platform developed through the Design Science Research process described in Section 3. The system was developed as an educational innovation platform that integrates challenge intake, participant profiling, AI-supported recommendations, workflow orchestration, milestone tracking, partner coordination, and human-in-the-loop oversight mechanisms within a unified university environment. Unlike conventional capstone-management systems that mainly support project allocation and course administration, the Smart Innovation Hub was developed to support broader coordination activities involving students, faculty mentors, research centers, innovation administrators, external partners, and commercialization pathways. The platform also integrates workflow management, interdisciplinary collaboration, and human-supervised decision support within a unified educational environment.

The architecture of the Smart Innovation Hub was influenced by recent advances in AI-enabled educational information systems, socio-technical governance frameworks, explainable AI, and workflow-oriented innovation ecosystems [4,9,10,22,23,24,25]. Particular emphasis was placed on transparency, traceability, interoperability, and human-supervised coordination to ensure that AI-supported recommendations remained compatible with institutional review requirements and educational objectives.

4.1. Platform Overview

The Smart Innovation Hub was developed to support challenge-based educational innovation and multidisciplinary project coordination within higher education environments. The platform enables students, faculty mentors, innovation administrators, research entities, and external partners to collaboratively manage innovation-oriented educational workflows through a centralized digital ecosystem. The platform supports multiple innovation-oriented workflow categories, including challenge-based capstone projects, graduation projects, research collaboration requests, innovation initiatives, startup-oriented activities, patents, contribution requests, and commercialization pathways. Rather than functioning solely as a project-allocation system, the system was developed as a broader institutional coordination infrastructure capable of supporting educational, research, innovation, and entrepreneurship activities within a unified implementation environment.

The platform combines several interconnected functional components, including:

• A structured challenge intake;
• Participant-profile enrichment;
• Semantic and structured recommendations;
• Interdisciplinary team formation;
• Mentor alignment;
• Workflow governance;
• Milestone tracking;
• Evidence capture, and human-in-the-loop review processes.

These capabilities enable the Smart Innovation Hub to coordinate complex educational innovation workflows while preserving academic oversight, transparency, and operational traceability [7,8,9,22,23,24,25,26].

4.2. System Architecture and Functional Components

The Smart Innovation Hub architecture was designed as a modular AI-enabled information system integrating recommendation intelligence, workflow governance, institutional coordination, and innovation-management functionalities within a unified platform environment. The architecture combines structured data processing, semantic similarity analysis, workflow orchestration, and governance-oriented oversight mechanisms to support challenge-based educational coordination and innovation management activities. The Smart Innovation Hub uses a modular web-based architecture consisting of a user-facing application layer, an API and workflow-management layer, a relational database layer, and an AI-supported recommendation and matching layer. The user-facing application supports challenge submission, opportunity discovery, participant-profile completion, coordinator review, workflow monitoring, and collaboration activities. The API layer manages authentication, permissions, workflow orchestration, business rules, recommendation requests, and data exchange across platform components. The relational database layer stores challenge records, partner information, participant profiles, mentor profiles, interdisciplinary team structures, recommendation outputs, milestone records, evidence artifacts, workflow logs, and outcome-related information. The AI-supported recommendation layer generates semantic and structured features, calculates compatibility scores, ranks candidate configurations, and produces explainable recommendation outputs to support challenge allocation and team-formation processes.

The architecture was intentionally designed to support human-centered coordination and governance-oriented educational decision-making. AI is positioned as a decision-support mechanism rather than an autonomous placement authority. Coordinators retain responsibility for reviewing recommendations, adjusting selection criteria, approving final decisions, and documenting override actions when necessary. Recommendation logs, workflow histories, and audit mechanisms are preserved to support transparency, traceability, accountability, and governance-oriented educational coordination [7,8,9,22,23,24,25]. This governance-oriented architecture distinguishes the Smart Innovation Hub from more conventional educational recommendation systems by combining AI-supported coordination with academic oversight, explainability, and human-supervised innovation workflows.

The platform includes several major functional layers:

• Challenge and Innovation Intake Layer: This layer supports structured submission of challenges, projects, ideas, patents, research initiatives, and innovation requests from students, faculty members, research centers, and external partners.
• Participant-profiling Layer: The profiling layer maintains structured participant information, including disciplinary background, technical skills, academic interests, prior project experience, research interests, and innovation-related activities.
• AI-recommendation and Matching Layer: This layer integrates semantic matching, structured eligibility filtering, interdisciplinary balancing, and recommendation-ranking mechanisms to support project allocation, mentor alignment, and multidisciplinary team formation.
• Workflow Governance Layer: The governance layer manages milestone tracking, workflow visibility, recommendation review, audit logging, approval stages, and human-supervised coordination activities.
• Innovation Ecosystem and Commercialization Layer: This layer supports startup incubation, intellectual-property coordination, commercialization pathways, external-partner collaboration, and post-project follow-up activities.

Figure 3 presents the operational ecosystem architecture implemented within the Smart Innovation Hub and the broader Innovation Bridge platform. The ecosystem integrates challenge submission, interdisciplinary coordination, AI-supported matching, business evaluation, startup-oriented pathways, intellectual-property governance, and external stakeholder collaboration within a unified innovation-management environment.

The workflow demonstrates how innovation requests originating from students, faculty members, research centers, industry partners, and external sponsors progress through structured evaluation, interdisciplinary recommendation, commercialization assessment, and governance-oriented coordination stages. The platform integrates AI-supported recommendation mechanisms with institutional oversight processes to support challenge-based educational innovation, research collaboration, startup incubation, and commercialization-oriented decision-making.

4.3. Governance-Oriented Innovation Workflow

A major contribution of the Smart Innovation Hub lies in its governance-oriented platform environment, which extends beyond conventional project-assignment systems by integrating institutional oversight, commercialization pathways, incubation mechanisms, intellectual-property coordination, and human-supervised innovation workflows. The platform was designed to support the complete lifecycle of innovation-oriented educational and research initiatives, including idea submission, challenge intake, evaluation, incubation, commercialization, licensing, startup formation, intellectual-property coordination, and post-project follow-up activities.

The governance workflow involves multiple institutional entities, including the Innovation and Entrepreneurship Center (IEC), Technology Transfer Office (TTO), research entities, evaluation committees, and external stakeholders participating in innovation and commercialization activities. These entities collectively participate in evaluating innovation requests, validating interdisciplinary project alignment, reviewing intellectual-property considerations, supporting incubation decisions, and managing commercialization-oriented pathways. Unlike conventional educational recommendation systems that primarily focus on project assignment, the Smart Innovation Hub integrates structured workflow orchestration with AI-supported coordination and institutional oversight. Recommendation outputs remain advisory rather than autonomous, while coordinators and governance committees retain authority for reviewing, approving, modifying, or rejecting innovation and project decisions.

This human-supervised governance structure supports accountability, transparency, and institutional compliance in educational innovation management [7,8,9,22,23,24,25,26]. The structured workflow also supports broader innovation-management functions beyond educational coordination, including startup incubation, commercialization planning, intellectual-property review, interdisciplinary project approval, external-partner coordination, and long-term innovation follow-up activities. These capabilities enable the platform to function as an integrated educational innovation ecosystem rather than a standalone matching system.

Figure 4 illustrates the structured workflow process implemented within the Smart Innovation Hub and the broader Innovation Bridge framework. The workflow illustrates how innovation requests, challenge-based projects, and commercialization-oriented initiatives move through structured coordination and governance stages within the broader institutional ecosystem. The workflow further illustrates how AI-supported coordination is combined with institutional oversight, commercialization pathways, governance controls, and human-in-the-loop decision-making processes. This governance-oriented architecture enables the platform to support educational coordination, innovation management, research commercialization, startup incubation, intellectual-property governance, and ecosystem-level innovation development within a unified institutional framework.

4.4. Core Data Model

The Smart Innovation Hub platform uses a relational and workflow-oriented data model designed to support challenge-based educational coordination, interdisciplinary matching, innovation management, and structured workflow orchestration. The data model was designed to support both operational recommendation activities and longer-term institutional innovation management processes. The core entities of the platform include:

• Challenges;
• Projects;
• Participant profiles;
• Mentor profiles;
• Research entities;
• External partners;
• Interdisciplinary teams;
• Recommendation outputs;
• Milestone records;
• Workflow actions;
• Evidence artifacts;
• Commercialization activities, and outcome-related indicators.

Challenges constitute the central coordination entity within the ecosystem. Each challenge record contains structured and unstructured information, including challenge descriptions, technical requirements, expected outcomes, partner information, confidentiality constraints, required skills, disciplinary preferences, timeline constraints, and commercialization potential. These challenge records are linked to recommendation processes, interdisciplinary teams, mentors, milestones, evaluation records, and workflow activities.

Participant profiles combine structured attributes and textual information. Structured attributes include:

• Academic discipline;
• Technical skills;
• GPA;
• Availability;
• Completed coursework;
• Certifications;
• Prior project experience, and participation history.

Textual profile fields include:

• Research interests;
• Innovation interests;
• Technical experience narratives;
• Prior project summaries, and portfolio-related descriptions.

This hybrid representation enables both structured eligibility filtering and semantic similarity analysis within the recommendation framework. The data model also supports structured workflow management through audit and traceability entities. Workflow records capture:

• Recommendation requests;
• Recommendation scores;
• Coordinator review actions;
• Approval stages;
• Override decisions;
• Milestone completion;
• Partner feedback;
• Evaluation outcomes, and commercialization-related activities.

These records improve transparency, traceability, explainability, and institutional accountability throughout the innovation lifecycle.

To support interoperability and extensibility, the platform architecture separates operational entities from recommendation-processing entities as shown in

Table 2. Core platform entities.

Entity	Main Fields	Function in the Platform
Initiative	Type, title, description, objectives, expected results, attachments, visibility	Represents the broader innovation item published in the platform, including ideas, projects, research outputs, challenges, and patents.
Challenge	Problem statement, deliverables, constraints, duration, required skills, confidentiality	Defines the project opportunity and matching requirements.
Partner	Organization type, contact, sector, expectations, data restrictions	Represents external or internal challenge source.
Student Profile	Discipline, skills, interests, availability, prior projects, preferences	Provides candidate data for team formation.
Mentor Profile	Expertise, capacity, research areas, preferred domains	Supports mentor matching and workload balance.
Team	Members, roles, disciplines, assigned mentor	Represents the final or candidate project team.
Match Recommendation	Score, ranking, explanation, constraints, override reason	Records AI-supported recommendation and coordinator decision.
Project	Active execution record, team, mentor, partner, timeline	Tracks approved projects after matching.
Milestone	Stage-gates, review dates, completion status	Supports workflow governance.
Evidence Artifact	Reports, prototypes, datasets, presentations, validation evidence	Supports evaluation and accountability.
Patent	Patent identity, patent type, status, patent office, inventor and ownership, technical description, legal confirmation, attachments	Links intellectual property to related projects and supports commercialization and technology transfer.
Contribution Request	Requested support, required expertise, requester, related initiative, nomination status	Allows users to request support and enables AI-enabled nomination of suitable contributors.
Outcome	Rubric scores, partner acceptance, satisfaction, impact indicators	Enables process and educational evaluation.

below. Recommendation outputs, semantic similarity calculations, and scoring records are stored independently from user profiles and challenge records, allowing recommendation strategies to evolve without disrupting core institutional workflow data structures.

Figure 5 presents the high-level entity relationship and operational data structure implemented within the Smart Innovation Hub. The model illustrates the relationships among challenges, participant profiles, mentors, projects, interdisciplinary teams, recommendation outputs, workflow governance entities, and commercialization-oriented activities within the broader Innovation Bridge framework.

4.5. Workflow and Innovation-Stage Alignment

The Smart Innovation Hub was designed to support the complete lifecycle of challenge-based educational innovation workflows rather than isolated project-allocation activities. The platform therefore aligns educational coordination processes with broader innovation-management, incubation, commercialization, and institutional-governance stages. The workflow begins with challenge and initiative intake. External partners, faculty members, research centers, innovation administrators, and students can submit challenges, ideas, projects, research initiatives, patents, and innovation requests through structured intake forms. Submitted initiatives include both structured fields and unstructured narrative descriptions to support semantic recommendation, governance review, and workflow coordination.

Following intake, the platform performs participant-profile enrichment and recommendation preparation. Student, mentor, and stakeholder profiles are enriched using structured academic attributes, technical skills, research interests, prior project experience, innovation activities, and textual descriptions. This enrichment stage improves recommendation quality by enabling both structured eligibility filtering and semantic similarity analysis.

The recommendation and coordination stage integrates AI-supported matching with human-supervised governance review. Candidate interdisciplinary teams, mentor alignments, and project configurations are generated using semantic and structured recommendation mechanisms. Coordinators and governance entities then review recommendation outputs, validate interdisciplinary balance, evaluate operational feasibility, and approve workflow progression when necessary. After recommendation approval, initiatives transition into execution and milestone-tracking stages. During this phase, the platform supports project monitoring, evidence capture, milestone submission, workflow tracking, mentor feedback, partner interaction, and outcome documentation. These workflow stages improve visibility, accountability, and coordination transparency across educational innovation activities.

The platform also supports innovation-management and commercialization-oriented stages beyond conventional capstone coordination. Promising initiatives can progress into startup incubation, intellectual-property evaluation, commercialization assessment, licensing consideration, industry collaboration, and technology-transfer workflows. This broader lifecycle orientation differentiates the Smart Innovation Hub from conventional educational project-allocation systems by positioning challenge-based engineering education within a larger university innovation ecosystem [4,10,11,21].

The workflow architecture was intentionally designed to preserve human oversight throughout all major coordination and governance stages. Recommendation outputs remain advisory rather than autonomous, while coordinators, evaluation committees, innovation administrators, and institutional review entities retain authority for workflow approval, recommendation modification, commercialization review, and decision validation. This governance-oriented structure improves transparency, accountability, explainability, and institutional trust in AI-supported educational coordination systems [7,8,9,22,23,24,25,26]. The workflow-oriented architecture and governance structure described above provide the operational foundation for the AI-supported recommendation and coordination framework implemented within the Smart Innovation Hub.

4.6. AI-Driven Matching and Recommendation Framework

The core operational capability of the Smart Innovation Hub lies in its AI-supported recommendation and coordination framework, which was designed to support transparent, interdisciplinary, and governance-oriented matching among challenges, students, mentors, research entities, and institutional resources. Unlike conventional capstone-assignment systems that rely primarily on manual coordination or isolated optimization rules, the proposed framework integrates semantic similarity analysis, structured eligibility filtering, interdisciplinary balancing, workflow governance, and human-in-the-loop review within an integrated recommendation architecture.

The recommendation framework was developed to address the multidimensional nature of challenge-based educational innovation workflows. Matching decisions in this environment require simultaneous consideration of technical competencies, disciplinary diversity, participant interests, mentor expertise, project requirements, institutional constraints, commercialization potential, stakeholder expectations, and workflow governance requirements. Consequently, the platform was designed as a socio-technical coordination system in which AI supports institutional decision-making while preserving human oversight and governance authority [7,8,9,22,23,24,25,26].

The recommendation architecture combines structured participant attributes with semantic representations extracted from unstructured textual information such as challenge descriptions, project abstracts, participant interests, mentor expertise statements, prior project summaries, and innovation narratives. This hybrid approach enables the platform to capture both explicit eligibility constraints and implicit semantic relationships that may not be represented through structured fields alone.

4.6.1. Inputs and Representations

The Smart Innovation Hub recommendation framework was designed to support multidisciplinary, governance-oriented, and challenge-driven educational coordination workflows. Unlike conventional capstone-assignment systems that rely primarily on manual coordination or isolated optimization rules, the proposed framework integrates semantic similarity analysis, structured eligibility filtering, interdisciplinary balancing, workflow governance, and human-supervised review mechanisms within a unified recommendation architecture. For each challenge or initiative, the platform aims to generate a ranked list of feasible and governance-compatible project configurations. A configuration may include a multidisciplinary student team, one or more faculty mentors, and, when appropriate, associated research centers, laboratories, or institutional resources. The recommendation process is designed to satisfy operational constraints while maximizing semantic alignment, interdisciplinary compatibility, workflow feasibility, and stakeholder relevance.

The recommendation problem is therefore treated as a constrained hybrid recommendation and coordination problem rather than a simple classification or allocation task. The framework simultaneously processes structured institutional data, unstructured textual information, multiple stakeholder requirements, structured workflow constraints, and human-supervised review activities. Consequently, the platform was intentionally designed as a decision-support system in which AI-generated recommendations remain transparent, explainable, reviewable, and auditable rather than fully autonomous.

The matching engine consumes two primary categories of information:

• Structured attributes:
  • Disciplinary background;
  • Technical skills;
  • Completed courses;
  • Certifications;
  • GPA;
  • Availability;
  • Mentor capacity;
  • Laboratory access;
  • Confidentiality constraints;
  • Prerequisite completion;
  • Project preferences;
  • Workload distribution, and institutional eligibility requirements.
• Unstructured textual information:
  • Challenge narratives;
  • Problem statements;
  • Expected outcomes;
  • Innovation descriptions;
  • Project abstracts;
  • Student interests;
  • Prior project summaries;
  • Mentor research descriptions;
  • Publications, and portfolio-related descriptions.

Table 3. Matching engine inputs (illustrative).

Entity	Field (Examples)	Type	How It Is Used
Challenge	Problem statement; desired outcomes; constraints; required skills/tools; confidentiality level	Text + structured	Embeddings + constraint filtering + required-skill matching
Student	Major/discipline; skills; interests; availability; prior projects; preferences	Text + structured	Feasibility + skill coverage + preference satisfaction + text similarity
Mentor	Expertise; research areas; capacity; preferred domains	Text + structured	Mentor–challenge fit + load balancing
Research center	Facilities; lab capabilities; focus areas	Text + structured	Capability coverage and resource feasibility

summarizes representative entities, attributes, and operational roles within the Smart Innovation Hub recommendation framework.

Structured attributes are normalized and encoded into feature vectors to support compatibility evaluation and eligibility filtering. During the pilot implementation, semantic similarity analysis was conducted using transformer-based multilingual sentence-embedding techniques to represent challenge descriptions, participant profiles, mentor expertise, and project metadata as vector representations. The implementation supported both Arabic and English textual content commonly used within the university environment. Cosine similarity was used as the primary semantic similarity metric during recommendation ranking. Structured constraints such as discipline eligibility, mentor capacity, availability, project preferences, and interdisciplinary composition requirements were integrated alongside semantic similarity scores within the recommendation workflow. The implementation was intended to support workflow coordination and pilot deployment activities rather than optimization-focused benchmark evaluation. The pilot emphasized practical workflow coordination and explainable recommendation support rather than comparative benchmarking across multiple embedding architectures.

The hybrid semantic-structured strategy improves recommendation quality because challenge-based educational innovation environments contain both explicit institutional constraints and ambiguous contextual information that cannot be adequately represented using rule-based allocation alone. The semantic layer enables the platform to identify hidden conceptual relationships among participants and projects, while the structured layer preserves operational feasibility, governance requirements, fairness considerations, and institutional policies [5,6,9,22,23].

The recommendation workflow begins with challenge intake and participant-profile enrichment. Structured challenge attributes include disciplinary requirements, technical skills, timelines, confidentiality conditions, commercialization potential, and partner expectations. Participant profiles include academic discipline, technical competencies, completed coursework, certifications, innovation interests, research experience, and textual portfolio descriptions. These data are processed through a hybrid recommendation pipeline integrating semantic similarity analysis, structured compatibility evaluation, interdisciplinary balancing, mentor alignment, and governance-oriented filtering mechanisms.

The overall recommendation score was calculated as a weighted hybrid compatibility function integrating semantic similarity, structured compatibility, interdisciplinary balance, mentor alignment, and governance-related constraints:

$$S_{\mathrm{total}}=\left(w_{text}{SM}_{text}+w_{skill}{SM}_{skill}+w_{\mathrm{p}\mathrm{r}\mathrm{e}\mathrm{f}\mathrm{e}\mathrm{r}\mathrm{e}\mathrm{n}\mathrm{c}\mathrm{e}}{SM}_{\mathrm{p}\mathrm{r}\mathrm{e}\mathrm{f}\mathrm{e}\mathrm{r}\mathrm{e}\mathrm{n}\mathrm{c}\mathrm{e}}+w_{\mathrm{D}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{s}\mathrm{i}\mathrm{t}\mathrm{y}}\mathrm{d}\mathrm{i}\mathrm{v}+w_{load}\mathrm{L}\mathrm{B}\right)$$

(1)

where

• $S_{\mathrm{total}}:$ Represents the final composite matching score assigned to a candidate project–team–mentor configuration. It combines semantic similarity, skill alignment, preference alignment, team diversity, and mentor workload balance into a single score used to rank feasible recommendations.
• ${SM}_{text}$: Represents the semantic similarity between the challenge description and the participant-profile text.
• ${SM}_{skill}:$ Represents the weighted overlap between required and available skills.
• ${SM}_{\mathrm{p}\mathrm{r}\mathrm{e}\mathrm{f}\mathrm{e}\mathrm{r}\mathrm{e}\mathrm{n}\mathrm{c}\mathrm{e}}:$ Captures the alignment between declared student, mentor, and partner preferences.
• $\mathrm{d}\mathrm{i}\mathrm{v}:$ Reflects disciplinary diversity and complementarity within the candidate team.
• $LB:$ Represents the mentor workload-balance term.

The weights $w_{text}$, $w_{skill}$, $w_{\mathrm{p}\mathrm{r}\mathrm{e}\mathrm{f}\mathrm{e}\mathrm{r}\mathrm{e}\mathrm{n}\mathrm{c}\mathrm{e}}$, $w_{\mathrm{D}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{s}\mathrm{i}\mathrm{t}\mathrm{y}}$, and $w_{Load}$ define the relative contribution of each component to the final score. These weights were adjusted iteratively during pilot deployment based on coordinator feedback, workflow observations, and operational feasibility considerations. Because the pilot represented an exploratory institutional implementation, the scoring framework should be interpreted as an adaptive governance-oriented recommendation mechanism rather than a fixed optimization model.

The staged recommendation workflow provides several operational and governance-oriented benefits. It reduces ambiguity during challenge intake, improves transparency in recommendation generation, preserves auditability of workflow decisions, and enables longitudinal evaluation of project outcomes and coordination effectiveness. The workflow also enables the platform to support multiple categories of initiatives, including capstone projects, research-center projects, industry challenges, entrepreneurship opportunities, innovation competitions, and commercialization-oriented activities. Figure 6 illustrates the AI-supported hybrid semantic-structured recommendation workflow implemented within the Smart Innovation Hub. The workflow integrates challenge intake, participant-profile enrichment, structured filtering, semantic similarity analysis, interdisciplinary balancing, governance-oriented evaluation, recommendation ranking, and human-supervised review activities within a unified recommendation pipeline.

To further demonstrate the operational implementation of the Smart Innovation Hub, Figure 7, Figure 8, Figure 9, Figure 10, Figure 11, Figure 12 and Figure 13 present anonymized screenshots from the deployed pilot environment. These figures illustrate how the platform supports initiative discovery, challenge submission, metadata structuring, workflow coordination, bilingual initiative management, contribution requests, and AI-supported nomination processes within the broader innovation ecosystem.

The initiatives dashboard serves as the central discovery and coordination interface within the Smart Innovation Hub ecosystem. Users can browse and filter initiatives across multiple categories, including ideas, projects, research initiatives, patents, challenges, and commercialization-oriented activities. Type-based filtering and keyword-search functionalities improve accessibility, initiative discovery, and workflow visibility while reducing administrative overhead for coordinators managing large numbers of submissions.

The initiative-detail interface standardizes how submissions are represented for downstream recommendation, governance review, and evaluation activities. Narrative summaries capture contextual information, while structured metadata fields support filtering, reporting, feature extraction, and interoperability within the AI-supported recommendation pipeline. Publication-control mechanisms further support structured workflow management and quality-assurance activities.

For challenge-driven educational workflows, the platform guides coordinators and partners to structure challenge statements, expected outcomes, solution directions, and institutional-priority alignment within standardized templates. This improves the quality and consistency of textual information used during semantic matching and recommendation generation while supporting later evaluation of challenge outcomes, workflow alignment, and project feasibility.

The platform home interface supports initiative discovery, engagement visibility, and multidisciplinary collaboration across the broader innovation ecosystem. Recommended initiatives and contribution-request mechanisms enable students, mentors, and external stakeholders to identify suitable collaboration opportunities while reducing the coordination cost associated with interdisciplinary participation and partner engagement.

Initiatives can be submitted through a structured wizard that begins with selecting the initiative type (idea, project, research, challenge, or patent) and entering core metadata such as title and description. The bilingual input option (Arabic/English) supports inclusive participation and enables higher-quality text matching when the platform is used in multilingual contexts. Moreover, initiatives are submitted through a structured multi-stage workflow beginning with initiative-type selection and entry of foundational metadata such as title, description, category, and participation details. The bilingual input capability (Arabic and English) supports multilingual educational environments while improving semantic matching quality across diverse stakeholder groups.

Additional submission stages capture execution-oriented information required for recommendation, governance review, workflow tracking, and outcome evaluation. Objectives, expected results, target groups, program classifications, and supporting attachments improve scope alignment, recommendation quality, traceability, and institutional documentation throughout the project lifecycle.

Together, Figure 7, Figure 8, Figure 9, Figure 10, Figure 11 and Figure 12 provide operational implementation evidence for the Smart Innovation Hub ecosystem. The screenshots demonstrate how initiatives are discovered, submitted, structured, enriched with metadata, linked to structured workflows, and routed toward AI-supported nomination and coordination activities. This implementation evidence complements the conceptual architecture and demonstrates that the proposed framework has been translated into an operational institutional platform rather than remaining a purely theoretical model.

4.6.2. Human-in-the-Loop Recommendation Governance

Although AI-supported recommendation mechanisms were used to generate candidate project–team–mentor configurations, the Smart Innovation Hub was intentionally designed as a human-supervised educational coordination system rather than a fully autonomous allocation platform. Human coordinators retained authority for reviewing recommendation outputs, adjusting constraints, validating interdisciplinary balance, approving final decisions, and documenting override actions when necessary. This governance-oriented recommendation strategy was implemented because challenge-based educational innovation workflows involve contextual judgment, institutional priorities, fairness considerations, stakeholder expectations, confidentiality constraints, commercialization requirements, and educational objectives that cannot be fully represented through automated optimization alone [7,8,9,22,23,24,25,26]. Consequently, AI recommendation outputs were treated as decision-support artifacts rather than deterministic assignment decisions. To improve explainability and transparency, the platform records:

• Recommendation scores;
• Contributing recommendation factors;
• Semantic similarity indicators;
• Structured eligibility conditions;
• Governance-related constraints;
• Override actions;
• Workflow decisions;
• Approval histories, and recommendation-review activities.

These audit and traceability mechanisms support governance-oriented educational coordination while preserving institutional accountability and human oversight. The governance framework also incorporates role-based access controls and workflow permissions to ensure that sensitive challenge information, partner requirements, commercialization-related records, and participant data remain visible only to authorized stakeholders. Structured workflow logs are preserved throughout recommendation, evaluation, incubation, and commercialization stages to support traceability, transparency, institutional auditability, and post-project evaluation.

Recent AI-governance research emphasizes that explainability, transparency, accountability, fairness, and human oversight are essential components of responsible AI-enabled organizational systems [22,23,24,25,26].

The Smart Innovation Hub operationalizes these principles by integrating recommendation explainability, human-supervised coordination, structured workflow management, and institutional review mechanisms directly within the recommendation lifecycle. The recommendation-review process therefore combines AI-supported coordination with institutional review validation. Coordinators and governance entities may approve, modify, defer, or reject recommendation outputs depending on workflow feasibility, interdisciplinary balance, resource availability, confidentiality considerations, or institutional priorities. This governance-oriented approach improves trust in AI-supported educational coordination while preserving flexibility for complex multidisciplinary innovation environments.

Recommendation explainability mechanisms were incorporated into the platform to improve transparency, recommendation interpretability, and stakeholder trust. Each recommendation includes an explanation identifying major contributing factors such as matched technical skills, relevant prior experiences, disciplinary alignment, semantic similarity indicators, preference compatibility, and mentor-capacity considerations. These explanations improve coordinator visibility into recommendation logic and support governance-oriented review activities.

The platform also supports human-in-the-loop control mechanisms that allow coordinators to adjust recommendation weights, lock pre-assigned participants, remove unsuitable candidates, rerun recommendation processes, and document override reasons when modifications are required. This governance-oriented review process is particularly important because educational innovation workflows often involve contextual considerations that cannot be fully represented through structured data or automated optimization alone.

Human oversight may improve recommendation quality by correcting incomplete profiles, contextual mismatches, workflow anomalies, or institutional constraints not fully captured within recommendation models. However, governance-oriented oversight may also introduce risks associated with implicit bias or inconsistent coordinator preferences if override activities are not properly documented and audited. To address these concerns, Smart Innovation Hub records override reasons, recommendation-review activities, invitation outcomes, and workflow decisions to support later auditing, transparency evaluation, and governance review.

The recommendation framework also faces cold-start challenges when new students, mentors, or external partners possess incomplete profiles or limited historical activity records. To reduce this limitation, the platform encourages profile enrichment and combines semantic recommendation with structured challenge attributes, institutional constraints, and workflow-oriented filtering mechanisms. This hybrid strategy improves operational feasibility while reducing dependence on historical interaction data alone.

Data privacy, governance, and fairness considerations were integrated throughout the recommendation lifecycle. Profile fields were limited to information necessary for recommendation generation, workflow coordination, and evaluation activities. Sensitive attributes were excluded from recommendation scoring unless explicitly required for approved institutional equity objectives. The platform additionally records recommendation decisions, invitation outcomes, coordinator overrides, and workflow histories to support governance-oriented auditing and traceability.

Fairness-oriented governance activities should evaluate whether specific disciplines, student groups, mentors, or stakeholder categories become systematically underrepresented within high-visibility innovation opportunities or commercialization-oriented projects. Governance policies should further define authorization boundaries regarding access to participant profiles, partner challenge records, recommendation histories, and outcome-related information to preserve institutional accountability and participant privacy. Figure 13 illustrates the governance-oriented recommendation validation and review process implemented within the Smart Innovation Hub. The workflow demonstrates how AI-generated recommendation outputs progress through coordinator review, governance validation, audit logging, and final approval activities before assignment confirmation and workflow activation.

5. Pilot Deployment and Evaluation

This section presents the pilot deployment environment, evaluation design, operational measures, and preliminary implementation evidence associated with the Smart Innovation Hub. Because the platform was evaluated through an institutional pilot deployment rather than a randomized controlled experiment, the evaluation focuses primarily on operational feasibility, workflow efficiency, stakeholder acceptance, workflow coordination effectiveness, and implementation-oriented observations within a real educational environment. The evaluation combines multiple sources of evidence, including platform workflow logs, recommendation records, coordinator observations, stakeholder surveys, milestone-tracking data, project-delivery outcomes, and partner feedback. This multi-source evaluation strategy is consistent with Design Science Research approaches in which operational artifacts are evaluated through pilot deployment, institutional implementation, stakeholder interaction, and iterative refinement activities [22,23,24,25,30,31,32].

5.1. Pilot Test

The Smart Innovation Hub platform was piloted within a public university engineering and innovation environment involving senior engineering students, faculty mentors, capstone coordinators, innovation-center personnel, and external partners. The pilot deployment covered challenge submission, participant profiling, recommendation generation, coordinator review, team confirmation, milestone tracking, workflow coordination, and project evaluation activities. The pilot implementation was conducted between 2023 and 2025 within multidisciplinary engineering capstone and graduation-project environments. The deployment additionally included selected research initiatives, partner-oriented challenges, innovation requests, and commercialization-oriented activities within the broader Innovation Bridge ecosystem.

Participant profiling remained voluntary during the pilot deployment. Consequently, profile completeness varied among students, mentors, and participating stakeholders. Students with more complete profiles may therefore have been more visible to the recommendation framework during semantic and structured matching activities. Because the deployment represented an observational institutional implementation rather than a randomized controlled experiment, the findings should be interpreted as preliminary implementation evidence intended to evaluate operational feasibility, workflow efficiency, recommendation acceptance, and workflow coordination effectiveness.

The stakeholder survey instrument was designed primarily as an implementation-feedback mechanism rather than a fully psychometrically validated scale. Survey items focused on recommendation relevance, workflow usability, transparency, satisfaction, perceived fairness, and coordination effectiveness. Because the pilot sample remained limited, reliability indicators such as Cronbach’s alpha should be interpreted as exploratory implementation measures rather than confirmatory validation evidence. Future deployments should evaluate the survey instrument across larger multi-institutional cohorts.

Table 4. Pilot implementation test.

Reporting Item	Reported Value/Description
Institutional setting	Public university capstone and graduation-project environment
Pilot period	2023–2025
Submitted challenges/projects	60 submitted challenges/projects
Participating students	180 students
Matched or active pilot teams	9 teams/projects
Faculty mentor pool	90 faculty mentors
External partners	5 partners from industry, healthcare, and government sectors
Primary data sources	Platform logs, recommendation records, coordinator observations, stakeholder surveys, partner feedback, milestone records, and project-assessment artifacts

summarizes the operational characteristics of the Smart Innovation Hub pilot deployment environment and the primary data sources used during evaluation activities.

5.2. Data Sources

The evaluation drew upon multiple operational and stakeholder-oriented data sources to assess workflow efficiency, recommendation performance, workflow coordination, and implementation feasibility within the Smart Innovation Hub ecosystem. The primary evaluation data sources included:

• Platform logs: timestamps associated with challenge submission, publication, profile completion, recommendation generation, invitation workflows, team confirmation, milestone completion, and project closure activities.
• Coordinator records: baseline manual-process estimates, recommendation-review activities, override decisions, workflow observations, and administrative-effort indicators.
• Recommendation outputs: ranked recommendations, compatibility scores, semantic similarity indicators, contributing recommendation factors, candidate lists, and recommendation-acceptance outcomes.
• Initiative and patent records: initiative categories, bilingual descriptions, objectives, expected outcomes, supporting attachments, patent status, ownership information, project-to-patent linkages, and commercialization-related metadata.
• Stakeholder surveys: student, mentor, coordinator, and partner ratings associated with recommendation relevance, workflow usability, transparency, satisfaction, fairness, and coordination effectiveness.
• Project quality evidence: milestone completion, final deliverables, prototypes, rubric-based evaluations, mentor assessments, and partner acceptance activities.
• Learning outcome evidence: technical performance indicators, teamwork activities, communication outcomes, reflective artifacts, presentations, and final project assessments.

5.3. Operational Measures

The evaluation defined a set of operational, stakeholder-oriented, governance-oriented, and educational measures to assess the pilot deployment of the Smart Innovation Hub. These measures were selected to reflect the platform’s intended role as both an AI-supported recommendation system and governance-oriented educational innovation infrastructure. The primary operational measures included:

• Team-formation cycle time: elapsed calendar days from challenge release or publication to confirmed team roster.
• Administrative effort: coordinator time or number of manual coordination actions required to finalize teams, when available.
• Recommendation acceptance: proportion of AI-supported recommendations accepted, modified, or rejected by coordinators during the pilot workflow.
• Match relevance: stakeholder perception that the recommended team, mentor, and challenge configuration fit the stated project requirements.
• Stakeholder satisfaction: percentage of respondents rating recommendation relevance, workflow usability, transparency, or overall satisfaction as satisfied or very satisfied.
• Governance transparency: availability of recommendation scores, explanation factors, coordinator-review records, override documentation, and workflow histories.
• Challenge resolution rate: proportion of projects judged by partners or coordinators as meeting the core need, completing agreed milestones, or producing usable deliverables.
• Partner operational impact: reported change in a relevant partner metric or workflow outcome after solution deployment in selected representative cases.
• Commercialization readiness: presence of initiative documentation, supporting attachments, prototype evidence, patent linkage, partner interest, or potential pathway for piloting, licensing, scaling, or market adoption.
• Project quality: rubric-based assessment of final deliverables, validation evidence, mentor evaluations, prototype maturity, and partner acceptance.
• Student learning outcomes: evidence of technical performance, teamwork, communication, problem-solving, professional growth, reflective practice, and final project assessment.

These measures were used descriptively to evaluate the operational feasibility and preliminary impact of the platform. Because the pilot was observational and institution-specific, the measures were not interpreted as causal proof of effectiveness but as implementation evidence supporting further validation.

5.4. Analysis Approach

Operational metrics were summarized descriptively using means, medians, counts, and percentages. The evaluation focused on describing implementation patterns, workflow changes, stakeholder perceptions, and operational feasibility rather than conducting causal inference. The reduction in team-formation cycle time was calculated by comparing the average duration of the previous manual coordination process with the corresponding duration recorded through platform-supported workflows during the pilot deployment. The percentage reduction was calculated as follows:

$$P_{\mathrm{reduction}}={\displaystyle \frac{\left(D_{manual}-D_{platform}\right)}{D_{manual}}}\times 100$$

(2)

where

• $P_{\mathrm{reduction}}:$ represents the percentage reduction in team-formation cycle time after using the platform compared with the previous manual process. A higher percentage reduction indicates greater improvement in process efficiency.
• $D_{manual}:\mathrm{r}\mathrm{e}\mathrm{p}\mathrm{r}\mathrm{e}\mathrm{s}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{s}$ the average team-formation duration in the previous manual process.
• $D_{platform}:\mathrm{r}\mathrm{e}\mathrm{p}\mathrm{r}\mathrm{e}\mathrm{s}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{s}$ the average team-formation duration recorded in the platform-supported process during the pilot implementation.

Because the pilot involved a limited number of active teams and was not randomized, inferential statistics such as p-values or confidence intervals were not used to establish causal claims. Instead, the evaluation was treated as descriptive implementation evidence intended to assess operational feasibility, workflow efficiency, recommendation acceptance, stakeholder satisfaction, governance transparency, and areas for future validation. Survey results were summarized as the percentage of respondents selecting satisfied or very satisfied responses. Open-ended survey comments, coordinator observations, and partner feedback were reviewed thematically to identify recurring strengths, operational concerns, governance-related issues, and improvement opportunities. Consequently, the findings should be interpreted as preliminary implementation indicators rather than generalized evidence of causal educational effectiveness.

5.5. Pilot Results

The pilot deployment provided preliminary implementation evidence regarding workflow efficiency, recommendation relevance, stakeholder satisfaction, and workflow coordination. The findings are reported descriptively because the pilot was conducted within a single university environment and was not designed as a randomized controlled study. Operational records indicated that the platform-supported process reduced average team-formation cycle time compared with the previous manual coordination process. In the previous manual process, team formation required approximately 30 calendar days on average. During the platform-supported pilot workflow, the corresponding process required approximately 12 calendar days on average. This represents an approximate 60% reduction in team-formation cycle time.

The calculation is shown as follows:

$${\mathit{P}}_{\mathrm{reduction}}={\displaystyle \frac{\left(\mathbf{30}-\mathbf{12}\right)}{\mathbf{30}}}\times \mathbf{100}=\mathbf{60}\mathit{\%}$$

This result should be interpreted as an early operational indicator rather than causal evidence, because the pilot context was observational and institution-specific. However, the reduction suggests that structured challenge intake, participant profiling, AI-supported recommendations, coordinator review, and workflow tracking may improve coordination efficiency in challenge-based capstone environments.

Table 5. Descriptive operational indicators from the Smart Innovation Hub pilot deployment.

Metric	Manual Process	Smart Innovation Hub Pilot	Observed Change
Average team-formation cycle time	30 days	12 days	60% reduction
Team-formation workflow visibility	Limited	Recorded through platform workflow logs	Improved traceability
Recommendation review	Manual coordinator judgment	AI-supported ranking with coordinator review	Improved decision support
Evidence capture	Fragmented across documents and email	Centralized milestone and artifact records	Improved documentation
Partner feedback	Informal or delayed	Captured through platform and coordinator records	Improved feedback visibility

summarizes the main operational indicators observed during the pilot deployment. “The manual baseline value was derived from coordinator records documenting the average duration of the previous team-formation process, while the platform-supported value was derived from workflow timestamps recorded in Smart Innovation Hub during the pilot deployment”.

Stakeholder survey responses indicated generally positive perceptions of recommendation relevance, workflow usability, transparency, and overall satisfaction. More than 90% of respondents reported satisfaction with match relevance and workflow usability. These findings suggest that the platform-supported workflow was perceived as useful by participating stakeholders. However, because the survey was implemented as an institutional feedback instrument within an observational pilot, the results should be interpreted as preliminary perception-based evidence rather than a generalized measure of system effectiveness.

Table 6. Stakeholder-satisfaction indicators and exploratory reliability measures from the pilot deployment.

Construct	Example Focus Area	Items	Cronbach’s α	Satisfied or Very Satisfied
Workflow usability	Ease of navigation and coordination	5	0.87	92%
Recommendation relevance	Perceived quality of team/project matching	4	0.84	91%
Transparency and explainability	Visibility of recommendation logic and workflow actions	4	0.82	89%
Overall stakeholder satisfaction	General experience with the platform workflow	5	0.89	93%

summarizes stakeholder-satisfaction indicators and exploratory reliability measures collected during the pilot deployment. Because the pilot sample remained limited and the survey instrument was designed primarily as an implementation-feedback mechanism, the reported reliability indicators should be interpreted as preliminary exploratory measures rather than confirmatory validation evidence.

The stakeholder-feedback survey included responses from students, faculty mentors, coordinators, and external partners participating in the pilot deployment. A total of 100 responses were collected, including 72 student responses, 18 faculty mentor responses, five coordinator responses, and five external partner responses. Relative to the pilot population reported in Table 4, this corresponds to response rates of 40% for students, 20% for faculty mentors, and 100% for external partners. Coordinator responses represented internal implementation stakeholders involved in platform administration and workflow review. Because participation remained voluntary during the pilot implementation, response rates varied across stakeholder groups. Coordinator observations and partner feedback further indicated that centralized workflow visibility, structured challenge intake, metadata standardization, and AI-supported recommendations reduced coordination ambiguity and improved traceability across challenge-based educational workflows. Stakeholders additionally reported that the platform improved interdisciplinary collaboration visibility and facilitated earlier identification of suitable mentors, contributors, and project-team configurations.

Preliminary pilot observations suggested improvements in workflow visibility, interdisciplinary coordination, and stakeholder engagement processes within challenge-based educational activities. Additional dimensions such as long-term partner outcomes, challenge-resolution quality, project impact, and longitudinal learning outcomes remain part of the broader evaluation framework but were not yet comprehensively assessed during the exploratory pilot deployment.

Several limitations were also observed during the pilot deployment. Recommendation quality depended partially on profile completeness and challenge-description quality. New participants with limited historical activity records occasionally experienced reduced recommendation visibility because of cold-start limitations. In addition, recommendation-review activities remained dependent on coordinator participation and institutional workflow engagement. These findings highlight the importance of governance-oriented oversight, profile enrichment, and continuous refinement within AI-supported educational innovation ecosystems.

5.6. Pilot Limitations and Future Validation

Although the pilot deployment demonstrated promising operational and coordination-oriented outcomes, several limitations should be acknowledged. First, the deployment was conducted within a single university environment and involved a limited number of active multidisciplinary teams. Consequently, the findings should not be interpreted as generalized evidence of causal educational effectiveness.

Second, participant profiling remained voluntary during the pilot implementation. Students, mentors, and external stakeholders with more complete profiles may therefore have been more visible to the recommendation framework during semantic and structured matching activities. Third, the evaluation relied partly on stakeholder perceptions, coordinator observations, and implementation-oriented operational records rather than randomized comparative experiments.

The stakeholder survey instrument was designed primarily as an institutional implementation-feedback mechanism rather than a fully validated psychometric scale. Although exploratory reliability indicators were acceptable, future research should validate the instrument across larger multi-institutional cohorts and evaluate recommendation fairness, governance transparency, and educational outcomes using broader longitudinal datasets.

Future work should additionally evaluate:

• Advanced semantic recommendation models;
• Fairness-aware recommendation mechanisms;
• Adaptive weighting strategies;
• Explainable AI interfaces;
• Commercialization-oriented analytics;
• Longitudinal educational outcomes, and multi-institutional deployment scenarios.

Further validation should also examine scalability, recommendation consistency, governance-oriented auditing, and institutional interoperability across larger innovation ecosystems and multidisciplinary educational environments.

6. Discussion

The findings from the pilot deployment suggest that the Smart Innovation Hub may improve coordination efficiency, workflow visibility, recommendation transparency, and interdisciplinary collaboration within challenge-based engineering education environments. The platform operationalizes a governance-oriented educational innovation ecosystem in which AI-supported recommendation mechanisms are integrated with structured workflow management, human oversight, institutional review, and commercialization-oriented coordination processes. Unlike conventional capstone-assignment systems that primarily focus on isolated project allocation, the Smart Innovation Hub supports broader innovation lifecycle activities including challenge intake, contributor discovery, mentor alignment, milestone tracking, commercialization readiness, startup-oriented pathways, intellectual-property linkage, and structured workflow auditing. This broader ecosystem orientation represents one of the primary contributions of the proposed framework. The study additionally demonstrates how semantic recommendation, structured eligibility filtering, interdisciplinary balancing, and governance-oriented review mechanisms may be combined within a unified institutional platform. The integration of AI-supported coordination with human-supervised governance workflows addresses increasing concerns regarding explainability, fairness, transparency, accountability, and institutional trust in AI-enabled organizational systems [22,23,24,25,26].

The pilot findings also suggest that centralized workflow visibility and metadata standardization may improve traceability and reduce ambiguity during challenge-based coordination activities. Structured challenge intake, bilingual initiative representation, recommendation explainability, and workflow logging further improved operational consistency across multidisciplinary innovation activities.

From a Design Science Research perspective, the study contributes both an operational artifact and preliminary implementation evidence derived from real institutional deployment. The artifact was iteratively refined through stakeholder engagement, workflow observation, coordinator feedback, and pilot implementation activities within an authentic educational environment. This iterative refinement process aligns with contemporary DSR-oriented information-systems research emphasizing practical utility, governance integration, and socio-technical coordination [22,23,24,25,30,31,32].

The broader Innovation Bridge ecosystem additionally extends the platform beyond traditional educational coordination by supporting contributor nomination, innovation collaboration, commercialization-oriented opportunities, and interdisciplinary ecosystem participation. Consequently, the recommendation problem expands from a narrow student-to-project assignment task into a broader contributor-to-initiative coordination framework supporting university innovation ecosystems.

Table 7. Summary of Smart Innovation Hub research and implementation contributions.

Area	Contribution
Design Science Research	Governance-oriented DSR artifact for challenge-based educational innovation ecosystems
Artificial Intelligence	Hybrid semantic-structured recommendation framework integrating semantic similarity and structured filtering
Governance	Human-in-the-loop recommendation review, auditability, transparency, and workflow traceability
Educational Innovation	AI-supported coordination of multidisciplinary capstone and innovation projects
Information Systems	Centralized lifecycle management integrating challenge intake, recommendation, workflow tracking, and evaluation
Innovation Ecosystem	Integration of research, commercialization, startup pathways, patents, and interdisciplinary collaboration
Explainability	Recommendation explanations, override logging, and governance-oriented review mechanisms
Operational Deployment	Institutional pilot implementation with workflow logs, stakeholder surveys, and implementation evidence

provides a summary of Smart Innovation Hub research and implementation contributions.

7. Limitations

Because the pilot deployment was conducted within a single university environment and involved a limited number of active multidisciplinary teams, the findings should not be interpreted as broadly generalizable across all higher-education contexts.

Several limitations should be acknowledged when interpreting the findings of this study. First, the pilot deployment was conducted within a single institutional environment and involved a limited number of active multidisciplinary teams. Consequently, the findings should not be interpreted as generalized evidence of causal educational effectiveness.

Second, participant profiling remained voluntary during the pilot implementation. Students, mentors, and external stakeholders with more complete profiles may therefore have been more visible to the recommendation framework during semantic and structured matching activities. Recommendation quality also depended partially on the quality and completeness of challenge descriptions, project metadata, and stakeholder-provided information.

Third, the evaluation relied partly on stakeholder perceptions, coordinator observations, and implementation-oriented operational records rather than randomized comparative experiments. Although exploratory reliability indicators were acceptable, the stakeholder survey instrument was designed primarily as an implementation-feedback mechanism rather than a fully validated psychometric scale.

The recommendation framework may additionally experience cold-start limitations when new participants possess limited historical activity records or incomplete profile information. Governance-oriented recommendation review processes may also introduce potential inconsistencies if override activities are not continuously documented and audited.

Finally, the pilot focused primarily on operational feasibility, workflow efficiency, workflow coordination, and implementation evidence rather than long-term educational outcomes or commercialization success. Future research should therefore evaluate broader longitudinal outcomes, fairness-aware recommendation mechanisms, multi-institutional deployment scenarios, scalability considerations, and advanced explainable-AI integration strategies.

8. Conclusions

This paper introduced the Smart Innovation Hub, an AI-enabled educational innovation platform developed to support challenge-based engineering education, interdisciplinary collaboration, and structured coordination within higher education environments. The platform combines semantic recommendation, structured eligibility filtering, workflow management, milestone tracking, and human-supervised decision support within a unified university innovation environment. From a Design Science Research perspective, the study contributes both an operational artifact and preliminary implementation evidence derived from a real institutional pilot environment. The findings suggest that AI-supported coordination mechanisms may help improve workflow visibility, recommendation transparency, interdisciplinary collaboration, and coordination efficiency within multidisciplinary innovation environments.

A major contribution of the proposed framework lies in integrating AI-supported recommendation with structured workflow management, explainability mechanisms, human oversight, auditability, and academic accountability. Unlike conventional project-assignment systems, the Smart Innovation Hub positions educational coordination within a broader innovation framework supporting challenge intake, contributor discovery, commercialization-oriented activities, startup pathways, and interdisciplinary collaboration.

The pilot deployment additionally demonstrated the feasibility of integrating semantic recommendation, structured metadata management, bilingual initiative representation, workflow traceability, and structured coordination within a real university environment. Although the findings remain preliminary and observational, they provide preliminary implementation evidence supporting future multi-institutional deployment and longitudinal validation activities.

Future work may further examine fairness-aware recommendation strategies, adaptive weighting mechanisms, explainable-AI interfaces, longitudinal educational outcomes, and broader multi-institutional deployment scenarios. Additional evaluation across different educational environments may also help assess scalability, governance consistency, and long-term institutional impact.