AI-Based Adaptive Career Guidance for Sustainable Transformation in TVET Education

A.AI AND DIGITAL TRANSFORMATION IN TVET

The emergence of Education 4.0 represents a major transformation in how learning environments operate, particularly within technical and vocational education. Driven by advances in artificial intelligence, big data, and cyber-physical systems, Education 4.0 encourages institutions to adopt adaptive, personalized, and data-driven practices [13]. In the context of TVET, where learners prepare for technology-intensive careers, integrating AI into guidance and instruction is essential for maintaining relevance and competitiveness.

Globally, AI has been increasingly applied to support key educational functions such as automated assessment, real-time feedback, personalized learning pathways, and predictive analytics. Studies show that machine learning and multimodal analytics can identify student learning patterns, diagnose competency gaps, and recommend targeted interventions more efficiently than traditional tools [14]. These capabilities are especially beneficial for vocational learners who require tailored support to acquire industry-specific skills and navigate evolving occupational requirements.

In several countries, AI-Based Adaptive Career Guidance in TVET systems have begun adopting AI-powered platforms to evaluate competency levels, forecast student performance, and streamline program administration. Multimodal analytics capable of integrating numerical scores, text responses, behavioral data, and contextual variables are gaining traction as a means of capturing richer learner profiles and supporting strategic decision-making [15].

Recent educational technology research also discusses multimodality beyond data integration, including immersive and interactive environments such as virtual or metaverse-based platforms. In the present study, however, this dimension is treated cautiously. The empirical needs analysis conducted with students, counselors, and industry representatives primarily supports data-level multimodality, namely the integration of heterogeneous learner data such as psychometric profiles, academic records, behavioral indicators, and narrative responses. By contrast, immersive delivery environments are not positioned as a core requirement of the proposed framework because they were not identified by participants as an immediate priority. Accordingly, experience-level multimodality is framed here as a possible future extension that may enrich career exploration in institutions with sufficient infrastructure, rather than as a central component of the current conceptual model.

In this study, we therefore use multimodality in two complementary senses. Data-level multimodality refers to integrating heterogeneous learner data (e.g., psychometrics, academic performance, behavioral indicators, and narrative text) into a unified analytic profile. Experience-level multimodality refers to the design of interactive learning or guidance experiences that combine visual, spatial, and interactive modalities (e.g., immersive 3D environments, virtual navigation, and guided exploration). This distinction clarifies that our framework’s multimodal architecture primarily addresses data-level integration, while immersive environments represent a feasible extension layer for delivering guidance in more experiential formats.

However, while AI adoption in instruction is progressing steadily, its integration into career guidance remains far less developed.

The global trend suggests a growing recognition that AI can enhance personalization and improve learner engagement, yet many TVET institutions still lack the infrastructure, expertise, or frameworks needed to implement adaptive guidance systems effectively. This gap is particularly notable in developing nations [16], where digital transformation initiatives often focus on classroom technology rather than holistic, career-oriented support systems.

B.LIMITATIONS OF CONVENTIONAL CAREER GUIDANCE

Despite advancements in AI adoption in education, career guidance practices in vocational schools continue to rely heavily on conventional models. Traditional guidance often involves paper-based assessments, generic career information sessions, and counselor-led interpretation of psychometric results. These practices, although valuable, are insufficient for addressing the unique and evolving needs of vocational learners [17].

One major limitation is manual processing, which requires significant counselor time, especially in schools where counselor-to-student ratios are high. Manual interpretation of assessments can also introduce inconsistencies, and results may not be updated as students grow and develop new competencies.

A second limitation is fragmented data management. Student information such as academic performance, psychometric profiles, behavioral observations, and industry preferences is often stored in separate systems or physical files. This fragmentation prevents counselors from forming complete, longitudinally informed student profiles, resulting in recommendations that may be based on partial or outdated information.

Another concern is low personalization. Traditional guidance models frequently provide generalized recommendations that fail to reflect students’ individual strengths, interests, or career aspirations. Such one-size-fits-all guidance can lead to poor alignment between students’ chosen pathways and industry needs, contributing to lower job readiness.

Perhaps the most critical limitation is the absence of adaptive feedback loops. After students complete assessments or meet with counselors, there are typically no mechanisms for the system to track changes in their interests, performance, or goals. Guidance thus becomes static, failing to adjust when new information emerges. This is incompatible with the dynamic nature of vocational career development and the rapidly shifting labor market.

Collectively, these issues highlight the need for intelligent, integrated, and adaptive career guidance systems capable of supporting students continuously rather than episodically.

C.AI-BASED CAREER GUIDANCE AND ADAPTIVE RECOMMENDATION SYSTEMS

AI-based career guidance models have gained scholarly attention in recent years, though their use remains more common in higher education and general education contexts than in TVET. These models typically incorporate recommendation engines, predictive algorithms, and NLP to match learners with potential career pathways, training programs, or academic courses.

Recommendation engines use machine learning to analyze student data including academic records, interests, and psychometric results to generate personalized guidance. In some systems, algorithms predict career fit by comparing student profiles to industry competency requirements or labor market trends. These approaches have shown promise in improving the relevance and accuracy of career recommendations.

AI-powered psychometric and NLP systems extend the functionality of traditional assessments by analyzing narrative responses, reflective writing, or conversational text inputs. NLP can extract patterns that reveal personality attributes, decision-making styles, or emotional tendencies elements that are highly relevant to career planning [18]. When combined with psychometric tests, NLP allows for richer, multidimensional student profiles.

Despite these strengths, existing AI-based guidance systems also present weaknesses. Many rely on single-modal inputs, meaning they only process one type of data (e.g., textual responses or numerical scores). This limits their ability to integrate complex, multimodal information that better represents learners’ diverse characteristics. Other systems provide recommendations that lack adaptivity; they generate suggestions based on initial data but do not adjust as students interact with the platform or as labor market conditions shift.

Furthermore, few AI-guidance systems are grounded in established theoretical frameworks. Without strong theoretical foundations [19], systems risk offering recommendations that are technically efficient but psychologically unsupported or contextually misaligned.

Finally, very few existing AI-based career guidance tools are tailored specifically for TVET learners. As a result, they may not account for vocational program structures, industry certification requirements, or the hands-on nature of vocational learning [20].

D.AI ECOSYSTEMS, PLATFORMS, AND ASSESSMENT TECHNOLOGIES (NEW

The integration of artificial intelligence into educational systems has expanded considerably in recent years, creating new opportunities for personalized learning, intelligent assessment, and data-informed decision-making [21]. Within the broader AI ecosystem, several strands of research highlight how AI technologies are reshaping digital platforms, learning environments, and guidance systems [22].

AI ecosystems refer to systems in which multiple stakeholders, data streams, and digital infrastructures interact to generate intelligent outcomes. In education and creative industries, AI ecosystems have been used to enhance collaboration, streamline information sharing, and automate complex decision-making processes. Prior studies show that AI combined with cloud computing and big data can create integrated digital environments that connect institutions, learners, educators, and external partners [23].

Personalization is among the most significant contributions of AI to education. Intelligent platforms use learner data to tailor instructional pathways, recommend learning materials, and generate individualized feedback. AI-driven systems in vocational and higher education have demonstrated improved learner engagement, greater accuracy in competency identification, and enhanced motivation through adaptive recommendations [24].

AI-supported assessment technologies have proven effective in diagnosing learner needs, analyzing competencies, and predicting performance outcomes. Machine learning models can process multimodal data from test results to textual explanations to identify learning gaps and generate actionable insights [25].

Digital hubs and marketplace systems have also emerged as models for connecting users, resources, and services across educational ecosystems. These platforms illustrate how centralized data management and AI-driven recommendation systems can improve coordination among students, educators, industries, and policymakers [26].

E.THEORETICAL FOUNDATIONS: SCCT, TPB, AND CIPP

Three theoretical models underpin the conceptual framework proposed in this study: SCCT, the TPB, and the CIPP evaluation model.

SCCT explains how individuals form career interests, make choices, and perform in educational and occupational settings. Central constructs such as self-efficacy, outcome expectations, and personal goals are critical in shaping vocational learners’ career decisions. Integrating SCCT into an AI-based system ensures that recommendations support not only skill matching but also the psychological dimensions of career readiness. It allows the system to interpret student data in ways that enhance confidence, motivation, and realistic goal setting [27].

TPB contributes a behavioral dimension by emphasizing that attitudes, subjective norms, and perceived behavioral control influence individuals’ intentions and actions. In the context of career guidance, TPB helps ensure that recommendations are not only suitable but also actionable aligned with what students believe they can pursue and what their social environments support. This is especially relevant in vocational settings, where family expectations and cultural norms may significantly shape career choices [28].

CIPP, an evaluation model, provides a holistic framework for analyzing the context of implementation, available inputs, internal processes, and measurable outcomes. For AI-based guidance systems, CIPP helps structure the system architecture so that it is responsive to institutional realities [29], stakeholder needs, and continuous improvement cycles. It also ensures that adaptive learning mechanisms are embedded into the system’s operations, making the model both scalable and sustainable.

Together, these theories create a strong, multidimensional foundation that connects psychological, behavioral, contextual, and technological elements.

F.RESEARCH GAP AND CONCEPTUAL POSITIONING

Although previous research demonstrates the potential of AI to enhance educational decision-making, several gaps remain that justify the need for this study.

First, there is a lack of AI-driven career guidance frameworks that integrate multimodal data combining psychometric results, academic patterns, narrative responses, and contextual variables. Most existing systems rely on single data types, limiting personalization quality.

Second, existing platforms rarely incorporate adaptive algorithms that update recommendations as students gain new experiences, develop new preferences, or interact with the system. This absence of adaptivity undermines the long-term usefulness of digital guidance tools.

Third, virtually no conceptual frameworks are designed specifically for Indonesia’s TVET system, which faces unique challenges such as skill mismatches, uneven school resources, and inconsistent industry collaboration. A localized, culturally relevant model is essential for meaningful implementation.

These gaps highlight the need for a theoretically grounded, contextually responsive, and technologically advanced conceptual framework for AI-based adaptive career guidance in vocational education.

A.PARTICIPANTS

Participants in this study were intentionally selected to represent the key stakeholders in vocational education: students, counselors, and industry professionals. The inclusion of diverse perspectives allowed for a holistic needs analysis and enhanced the contextual relevance of the developed framework.

Vocational Students: A total of 210 students from vocational schools in Bekasi, Bandung, and Banjarmasin participated in the survey. Students represented multiple study areas, including engineering, information technology, and business management. Schools were selected using purposive sampling based on their willingness to collaborate, their representation of different geographic areas, and their active implementation of career guidance services. Students were invited to participate through school announcements and counselor networks. Inclusion criteria required participants to be actively enrolled in the school and involved in career guidance activities. Participation was voluntary, and informed consent was obtained prior to data collection.

Guidance Counselors: Eight counselors participated in semi-structured interviews and focus group discussions. Counselors were recruited based on their direct involvement in delivering career guidance services and their familiarity with existing challenges in vocational counseling. Their roles allowed them to provide professional insight into institutional constraints, student needs, and practical considerations for implementing AI-based guidance tools. Counselors were approached through school administrators, who granted permission for their participation.

Industry Representatives: Six industry professionals from fields aligned with vocational school programs contributed their perspectives through interviews. These representatives were selected based on their experience in hiring vocational graduates, involvement in training or internship partnerships, and familiarity with workforce skill demands. Their insights helped identify industry expectations, skill mismatches, and opportunities for aligning guidance with labor market trends.

Collectively, these participants provided rich and varied perspectives that strengthened the ecological validity of the findings and ensured that the resulting framework addressed both educational and industrial needs.

B.INSTRUMENTS

Three primary instruments were used to collect data during the needs analysis phase: a needs assessment survey, interview protocols, and a document analysis guide.

Needs Assessment Survey: The 30-item survey measured students’ perceptions of existing career guidance services, their readiness to adopt AI-supported systems, and their expectations regarding personalization and adaptivity. Survey items were developed through a systematic process: (1) identifying constructs from SCCT, TPB, and digital guidance literature; (2) drafting items based on prior studies and contextual challenges identified in Indonesia; and (3) consulting two educational technology experts and one counseling specialist to refine item clarity and content validity. A small pilot test involving 20 students was conducted to ensure clarity and reliability. Internal consistency was high, with Cronbach’s alpha measured at 0.89.

Interview and FGD Protocols: Semi-structured protocols guided interviews with counselors and industry representatives. The items in these protocols were derived from the literature on digital transformation, AI adoption, and vocational career guidance challenges. Experts reviewed the interview guides for clarity, relevance, and appropriateness for each stakeholder group. The protocols provided flexibility for participants to expand on their experiences while ensuring consistency across sessions.

Document Analysis Guide: The document analysis guide directed the review of national education documents, institutional reports, and vocational policy frameworks. The guide included criteria for identifying themes related to digital transformation, guidance challenges, and strategic priorities for TVET. This ensured that the final conceptual framework was aligned with national educational objectives.

C.DATA COLLECTION PROCEDURES

Data collection occurred between January and April 2025, following a clearly defined sequence to ensure systematic coverage of all stakeholder perspectives.

Survey Administration: Surveys were administered during the first month of data collection. School counselors distributed surveys both electronically and in printed form to accommodate students with varying levels of digital literacy. Students were given clear instructions regarding the voluntary nature of participation and the anonymity of responses.

Interviews and FGDs: Interviews with counselors and industry representatives took place in the second and third months. Sessions were conducted both online and in person depending on participant availability. Counselors participated in an additional focus group discussion, which allowed for collective reflection and deeper exploration of common challenges, such as fragmented data records, heavy caseloads, and limited digital tools.

Document Review: Document analysis was carried out throughout the four-month period. Relevant policy documents and institutional guidelines were reviewed to triangulate findings and confirm alignment between the proposed framework and national educational priorities, including Indonesia’s Medium-Term Development Plan.

Ethical Procedures: Ethical considerations were integrated throughout the study. Institutional permission was obtained from participating schools. All participants were informed about the study’s objectives, confidentiality measures, and their right to withdraw at any stage. No identifying information was recorded, and all data were stored securely. The study adhered to established research ethics standards for educational and social science research.

D.DATA ANALYSIS

Data analysis employed both quantitative and qualitative techniques, allowing for a comprehensive understanding of stakeholder needs and expectations.

Quantitative Analysis: Survey data were analyzed using descriptive statistics to identify patterns in student perceptions. Means, frequencies, and standard deviations were calculated to measure interest in AI-based tools, satisfaction with current guidance services, and expectations for adaptivity and integration. This analysis provided a baseline quantitative representation of student needs.

Qualitative Analysis: Interviews and FGDs were analyzed using thematic analysis, following the six-step approach of Braun and Clarke: familiarization, generating initial codes, searching for themes, reviewing themes, defining and naming themes, and producing the final report. Coding was conducted inductively to allow themes to emerge organically from the data. Two researchers independently coded the transcripts, and discrepancies were resolved through discussion to enhance coder reliability. NVivo software (or a similar qualitative tool) was used to organize coding and support theme generation.

Triangulation: Triangulation across surveys, interviews, and document reviews strengthened interpretive validity. Findings from one source were cross-checked with others to ensure consistency. For instance, student reports of fragmented data systems were corroborated by counselor interviews and institutional document analysis.

Expert Validation: The initial conceptual framework underwent expert validation by three specialists in educational technology and vocational counseling. Experts were selected based on their publication records, domain expertise, and familiarity with AI applications in education. A structured validation rubric was used to evaluate conceptual clarity, theoretical alignment, contextual feasibility, and practical relevance. Two iterative review cycles were conducted, with revisions incorporated after each cycle to enhance model coherence and applicability within TVET environments.

A.STUDENT PERSPECTIVES

Students consistently described the guidance services available to them as overly general, infrequent, and disconnected from their personal aspirations. Many reported that career guidance activities were largely administrative limited to distributing forms, completing required documentation, or preparing students for routine assessments rather than providing meaningful developmental support. Survey findings indicated that more than 75% of respondents desired individualized guidance aligned with their interests, abilities, and long-term goals. Approximately 70% expressed limited access to digital tools and voiced a clear preference for online platforms that could facilitate continuous exploration and self-assessment. Students also emphasized the absence of adaptive feedback in current systems: once psychometric or interest assessments were completed, little follow-up occurred, and no personalized plan was offered. These perspectives underscore the need for a guidance system capable of integrating multiple dimensions of students’ profiles including interests, psychometric characteristics, academic history, and contextual factors and offering continuous, evolving support rather than one-time recommendations.

Students’ open-ended responses further revealed (a) low self-understanding, (b) limited exposure to data-driven tools, and (c) minimal awareness of real labor market demands. A summary of the key findings from the student needs assessment is presented in Table I.

B.COUNSELOR CHALLENGES

Interviews and focus group discussions with counselors revealed structural and procedural challenges that further limit the effectiveness of current guidance practices. Counselors reported managing exceptionally large caseloads, often supporting hundreds of students with limited technological or administrative assistance. Data fragmentation emerged as a pervasive issue: student records, psychometric test scores, academic data, and behavioral notes were stored across different documents or platforms, making it difficult to assemble a holistic profile for each learner. Counselors also noted varying levels of digital literacy among staff and insufficient institutional infrastructure for integrating technology into guidance workflows. These conditions hindered their ability to provide individualized attention, monitor student progress over time, or incorporate real-time labor market information into career discussions. Counselors expressed strong interest in an integrated, AI-supported platform that could consolidate student data, provide accessible analytics, and support more accurate and efficient decision-making.

C.INDUSTRY EXPECTATIONS

Industry representatives highlighted ongoing misalignment between vocational school outputs and workplace needs. They observed that many students lacked awareness of the specific technical and soft skills required for entry-level positions and often struggled to understand how their academic programs translated into real job opportunities. Industry stakeholders emphasized the need for a system capable of mapping student competencies to current labor market trends, enabling more accurate, context-informed recommendations. They also stressed the importance of adaptability, communication skills, and career self-efficacy factors that traditional guidance systems frequently overlook. For industries, an AI-driven platform with real-time analytics would help ensure that career pathways recommended to students are closely aligned with evolving workforce demands, reducing skill mismatches and improving employability outcomes.

Across all stakeholder groups, three consistent themes emerged. First, there is a strong demand for personalized and adaptive guidance that reflects students’ unique backgrounds and aspirations. Second, stakeholders recognized the need for integrated digital ecosystems capable of organizing and analyzing multimodal student data to support more holistic and continuous guidance processes. Third, all groups underscored the importance of aligning guidance outputs with industry trends, ensuring that students receive relevant, timely information about careers and competencies valued in the labor market. Together, these findings informed the development of the system requirements and guided the construction of the conceptual framework described in the subsequent sections, paralleling the earlier empirical tables presented in the manuscript. The key themes identified from counselor and industry feedback are summarized in Table II.

D.IDENTIFIED SYSTEM REQUIREMENTS

The findings from Phase 1 were systematically analyzed and translated into a set of core requirements for developing an AI-based adaptive career guidance framework. These requirements reflect the collective expectations of students, counselors, and industry professionals, and they establish the functional and structural foundations necessary for an effective and sustainable system. Four major categories of system needs emerged from the analysis: multimodal data integration, adaptive intelligence, real-time analytics, and alignment with labor market dynamics.

E.MULTIMODAL DATA INTAKE

Across stakeholder groups, there was strong agreement on the necessity of integrating diverse forms of student data into a unified platform. Effective guidance requires more than academic records or psychometric assessments alone; it must draw upon a comprehensive picture that reflects students’ interests, personality traits, competencies, learning behaviors, and contextual preferences. Stakeholders emphasized that narrative responses such as reflections, open-ended questions, or counselor notes should be incorporated and analyzed using NLP to capture affective and motivational dimensions that traditional numerical data cannot reveal. A multimodal intake mechanism allows the system to synthesize cognitive, behavioral, and contextual inputs, enabling more nuanced and accurate career recommendations.

F.AI-DRIVEN ADAPTIVITY

The second requirement centers on adaptivity. Career development is dynamic, and system recommendations must evolve as students acquire new skills, refine their interests, or encounter different educational experiences. Participants expressed dissatisfaction with current systems that offer static, one-time assessments without ongoing follow-up. To address this gap, the framework must incorporate adaptive algorithms capable of updating student profiles, recalibrating recommendations, and generating new insights in response to user interactions and changing conditions. This adaptivity transforms guidance from an episodic event into a continuous developmental process one that aligns more closely with the needs of vocational learners who must navigate shifting labor market landscapes.

G.REAL-TIME ANALYTICS

Industry representatives consistently stressed the importance of real-time insights that reflect current workforce demands. They noted that job roles, required competencies, and sectoral trends evolve rapidly, particularly in technology-driven industries. As such, the system must include the capacity to analyze labor market information, detect emerging trends, and provide up-to-date recommendations that remain relevant as conditions change. Real-time analytics also support counselors by offering timely indicators of student readiness, potential skill gaps, and opportunities for targeted interventions. By integrating live data streams or regularly updated repositories, the system can help ensure that students’ career decisions are informed by the most current and accurate information available. In operational terms, the framework assumes that labor-market intelligence will be drawn from multiple structured sources, including national employment statistics, government workforce and training information systems, occupational standards and competency documents, vacancy trend data from authorized digital job platforms where accessible, and validated inputs from school–industry partners. These data sources are intended to complement one another rather than function as a single stream. The framework further assumes a tiered update cadence: institutional student data may be synchronized continuously or at regular school reporting intervals; labor-market indicators may be refreshed monthly or quarterly depending on source availability; and school–industry validation may be conducted periodically through partnership review meetings. This clarification is important because “real-time” in the present conceptual framework refers to decision support based on the most recently available verified data, rather than uninterrupted live streaming from all sources.

H.INTEGRATION WITH LABOR MARKET TRENDS

Closely linked to real-time analytics is the need for systematic integration with national, regional, and sectoral labor market trends. Stakeholders emphasized that guidance must extend beyond personal attributes and academic data to reflect the realities of the employment landscape. This requirement ensures that career suggestions align with occupations that are in demand, emerging job roles, or sectors experiencing rapid growth. By linking student profiles with labor market intelligence whether derived from government databases, industry reports, or digital job platforms, the framework can help reduce skill mismatches and enhance job readiness. This alignment is particularly important in the vocational context, where employability is a central educational outcome.

I.CONCEPTUAL FRAMEWORK OUTPUT

The conceptual framework developed in Phase 2 brings together multimodal data processing, adaptive artificial intelligence, and theoretically informed decision pathways to support individualized and context-responsive career guidance within vocational education. The framework is designed as a layered architecture composed of interconnected components that work together to analyze student data, interpret developmental needs, and generate tailored recommendations. These components are organized into three primary layers input, processing, and output each serving a distinct but interdependent function in the guidance process, as illustrated in Figure 1.

Fig. 1. Conceptual framework of AI-based adaptive career guidance input layer.

The input layer functions as the foundation of the system by collecting and organizing diverse forms of learner data. This includes psychometric assessments, interest inventories, academic performance records, attendance and behavioral indicators, and contextual preferences such as preferred work environments, geographic mobility, and industry sector interests. In addition, the system captures narrative inputs from students’ reflections, open-ended responses, and self-descriptions which are essential for understanding affective and motivational aspects of career development. Through this multimodal intake, the framework constructs a multidimensional representation of each learner, enabling the system to move beyond static, single-source records and toward a holistic understanding of the individual’s potential, challenges, and aspirations [30].

In addition to learner-specific data, the input layer also accommodates structured reference data relevant to the Indonesian TVET context. These include qualification levels, occupational clusters, competency standards, certification pathways, and region-specific industry profiles. In the processing layer, such reference data are operationalized through rule-based matching and adaptive recommendation logic, allowing the system to relate student characteristics not only to general career categories but also to nationally and regionally recognized pathways. This explicit integration ensures that the framework is localized, policy-aware, and better suited to the institutional realities of Indonesian vocational education.

J.PROCESSING LAYER

The processing layer forms the analytical core of the framework. It integrates several AI-driven mechanisms to interpret and synthesize the data collected in the input layer. NLP tools analyze textual responses to extract themes related to career interests, self-efficacy, perceived barriers, and motivational cues. Adaptive learning algorithms identify evolving patterns in student behavior and preferences, enabling the system to adjust recommendations based on new information or changes in user interaction over time. Rule-based decision engines align student profiles with occupational categories, training pathways, and competency requirements.

Crucially, the processing layer operationalizes principles from SCCT and the TPB. Elements such as self-efficacy, outcome expectations, attitudes toward careers, perceived social norms, and behavioral control are incorporated into the recommendation logic, ensuring that outputs reflect not only technical alignment but also psychological readiness and motivation. As students interact with the platform or as labor market conditions shift, the adaptive engine recalibrates suggestions to maintain relevance and accuracy. This dynamic processing ensures that the system supports continuous career development rather than one-time guidance.

K.OUTPUT LAYER

The output layer presents the interpreted information in clear, actionable forms tailored to different users. For students, the system generates personalized career recommendations, suggested learning pathways, and visual reports that highlight strengths, potential areas for growth, and alignment with various occupational clusters. These outputs help students understand how their interests, competencies, and preferences fit into real-world opportunities. For counselors, the system provides dashboards that consolidate student data, track developmental progress, and display analytics on performance indicators and guidance needs. This enables counselors to offer more targeted, informed interventions.

In addition, the output layer incorporates labor market analytics, displaying current job trends, required competencies, and projected industry demands. By integrating real-time workforce information, the system allows both students and counselors to make decisions grounded in contemporary labor conditions rather than outdated or generalized assumptions. The diagrams included in your manuscript illustrate the flow of data across the three layers, showing how psychological, behavioral, and contextual elements interact to produce adaptive and meaningful guidance outputs.

L.EXPERT VALIDATION

The conceptual framework was subjected to expert validation to ensure its clarity, theoretical robustness, contextual alignment, and practical viability. Three experts were invited to participate in this process, selected based on their established academic credentials, publication records, and professional experience in educational technology, AI integration in learning environments, and vocational counseling. Their diverse expertise allowed for a comprehensive evaluation of both technological and pedagogical dimensions of the proposed model.

The experts reviewed the framework using a structured validation rubric that assessed dimensions such as logical coherence, alignment with theoretical constructs, contextual suitability within Indonesia’s vocational education system, and potential feasibility for school-level implementation. The validation process was conducted in two iterative feedback cycles, during which experts provided written comments, suggested revisions, and evaluated refinements made after each iteration. This iterative approach ensured that the conceptual framework evolved in response to informed critique and achieved a higher degree of internal consistency and relevance.

M.LOGICAL CONSISTENCY

Experts agreed that the framework demonstrated clear and coherent logic, particularly in illustrating the flow of multimodal data across layers, the integration of adaptive algorithms, and the mapping of learner attributes to career recommendations. They noted that the layered structure input, processing, and output reflected a well-organized system architecture that aligns with contemporary AI-enabled educational design. Minor improvements were recommended, such as clarifying terminology within the diagrams and simplifying certain labels to improve interpretability for non-technical users.

N.THEORETICAL ALIGNMENT

The reviewers affirmed that the framework was strongly grounded in the selected theoretical models SCCT, the TPB, and the CIPP evaluation framework. They highlighted that key constructs such as self-efficacy, outcome expectations, behavioral intentions, and contextual evaluation were meaningfully embedded within the processing logic of the system. This theoretical integration, according to the experts, strengthened the conceptual validity of the model and ensured that its adaptive mechanisms aligned with established principles of career development and decision-making.

O.CONTEXTUAL RELEVANCE

All three experts emphasized that the framework appropriately addressed the unique challenges of Indonesia’s vocational education context. They acknowledged the model’s responsiveness to issues such as fragmented data systems, limited technological infrastructure, variability in digital literacy among school staff, and misalignment between school competencies and industry needs. To further enhance contextual fit, the experts suggested incorporating region-specific skill taxonomies and industry profiles, which would increase the precision and cultural relevance of career recommendations across different geographic areas.

P.PRACTICAL FEASIBILITY

Regarding feasibility, experts recognized that successful implementation would require adequate training for counselors, stakeholder collaboration, and institutional support for digital transformation. They viewed the framework as practical and achievable, provided that vocational schools adopt a phased implementation strategy and develop guidelines for integrating AI tools into counseling workflows. The experts also stressed the need to embed ethical safeguards particularly those related to data privacy, informed consent, and the handling of sensitive student information to ensure compliance with educational and legal standards.

Overall, the expert validation process confirmed the framework’s conceptual strength and practical potential. A narrative summary corresponds to the type of validation table presented earlier in the manuscript (Table III), which synthesizes expert evaluations across validation criteria.

Table I. Summary of key findings from student needs assessment

Table II. Key themes identified from counselor and industry feedback

Table III. Summary of expert validation results