Recent literature emphasizes that digital competence has become a fundamental requirement for modern health professionals, and this necessity increasingly extends to those in engineering fields involved with healthcare technology. The European Union explicitly identifies digital competence as one of the key competencies for lifelong learning, highlighting its critical importance in professional life, particularly in healthcare (Mainz et al., 2024).

A 2024 scoping review by Mainz et al. (2024) found broad consensus in the healthcare sector regarding the necessity of digital literacy and competence across various health professions. This competency is crucial for effective performance in contemporary healthcare settings, where digital tools and data play integral roles in clinical decision-making, patient monitoring, and health information management. Furthermore, engineering students involved in developing and implementing health technologies must achieve a high level of digital competence, akin to that of students in health-related programmes. They are likely to engage in creating innovations such as patient-monitoring devices, health informatics software, or AI-driven diagnostic tools. Mastery of digital skills allows these future engineers to design solutions that meet clinical needs and smoothly integrate into healthcare workflows, ultimately benefiting patient care.

International frameworks now position digital competence at the core of health-profession education. The European Commission’s Digital Competence Framework for Educators (DigCompEdu) encourages educators, including those in health, to continuously enhance both their digital skills and pedagogical approaches. It emphasizes areas such as professional collaboration (e.g., sharing and co-creating knowledge) and ongoing professional learning with digital resources (Redecker and Punie, 2017; The Joint Research Centre: EU Science Hub, 2017).

International frameworks now place digital competence at the centre of health-profession education. The European Commission’s DigCompEdu framework continues to encourage educators to develop both their digital practices and pedagogy (Redecker and Punie, 2017; The Joint Research Centre: EU Science Hub, 2017). In parallel, the citizen framework has been updated to DigComp 3.0 (27 Nov 2025), which keeps the overall structure while updating competence wording, revising proficiency levels (now four), adding learning outcomes, and integrating AI competence transversally across all 21 competencies (The Joint Research Centre: EU Science Hub, 2025). Its development was guided by content themes: AI (including generative AI), cybersecurity, digital rights/choice/responsibilities, wellbeing in digital environments, and tackling mis/disinformation.

It also sets out application themes: digital competence as part of lifelong learning; recognition of prerequisites for reaching a basic level; recognition of differences in needs across individuals and over time; and the need for flexible, agile application of the framework. Together, these frameworks underline the need for continuous upskilling in health education, with AI embedded across domains rather than treated as a stand-alone topic.

Despite progressive policy mandates, implementation lags: many health programmes struggle to keep pace with AI and related technologies, so students often graduate with limited hands-on experience, an implementation gap confirmed by recent reviews showing non-standardised AI training and inconsistent coverage of ethics, data literacy, and practical tool use (Rincón et al., 2025; Shishehgar et al., 2025). Authoritative stakeholders echo this concern: a 2024 National Academy of Medicine convening noted minimal formal guidance in medical and nursing curricula despite growing AI investment in health systems, and similar gaps persist in public health education where machine-learning and large-language-model training remains scarce (National Academy of Medicine, 2024). In response, bodies such as the WHO call for deliberate curriculum redesign, embedding AI literacy, data science/informatics, and AI ethics; expanding experiential learning; and investing in faculty development, to prepare an “AI-ready” workforce. The present study contributes to this agenda by showing how structured, hands-on AI experiences, interpreted through DigCompEdu, can surface concrete design targets (e.g., strengthening A4-oriented assessment) to close the policy–practice gap.

Emerging literature links an innovation mindset with digital competence in healthcare. Disruptive technologies can augment (not replace) professional practice by elevating uniquely human skills—problem-solving, critical thinking, and adaptive creativity—provided clinicians are competent and confident in integrating these tools into care (Meskó et al., 2017). As Meskó et al. note, digital tools improve outcomes only when cultural and human factors are addressed; clinicians need creative problem-solving skills and openness to work effectively with AI. Accordingly, we treat innovation competence, particularly at the service/process level, as integral to DigCompEdu A6 (information and media literacy, responsible use, digital problem solving) and A1 (professional engagement).

Innovation in healthcare ecosystems involves the co-creation and orchestration of technological, organisational, and process solutions that enhance service delivery, stakeholder engagement, and value creation across networked actors (Adner, 2017; Pikkarainen et al., 2017). A useful distinction separates service/process innovation (new ways of organising care, workflows, and delivery) from product/technology innovation (new tools, devices, or software) (Greenhalgh et al., 2004; Omachonu and Einspruch, 2010). Both matter, but they engage different stakeholders and competencies.

Service innovation in healthcare is often enabled by digital platforms and data-driven capabilities, addressing evolving patient needs and system-level challenges (Lusch and Nambisan, 2015; Pikkarainen et al., 2022). Such innovation is increasingly central to professional practice and frequently conditions the successful integration of new technologies (Shaw et al., 2018). Creating an innovative culture relies on empowered teams, effective communication, and supportive leadership; recent work describes healthcare innovation as a dynamic interplay between technological advances and evolving delivery systems (Kosiol et al., 2024). By contrast, product/technology innovation (e.g., drugs, devices, AI algorithms) is typically led by industry and engineering teams, with clinicians participating as users, co-designers, or testers (Alami et al., 2020). Given this division of labour, the primary innovation focus for health professionals is often service/process improvement, while high-tech product development proceeds through interdisciplinary, technology-led approaches.

These distinctions have significant implications for education and professional identity. Graduates need to be prepared to innovate in practice by adapting workflows, improving care processes, and critically integrating tools, without necessarily writing code for devices or algorithms. Studies on AI-enabled decision support show identity tensions among clinicians. While some feel threatened by a perceived loss of autonomy and expertise, others view AI as a tool that enhances their capabilities (Ackerhans et al., 2024). Therefore, educators and leaders should foster a professional identity that values continuous learning, interprofessional collaboration, and innovation, equipping future practitioners to lead rather than resist digital transformation.

This study examines how master’s-level students in an interdisciplinary health-technology course envision the role of AI in their future clinical practice and academic studies after structured, hands-on engagement with the technology. We use the DigCompEdu framework to analyse students’ future-oriented perspectives regarding digital-competence areas and the supports they consider necessary for the development of professional identity, innovation in practice, and readiness for the future. To address this interdisciplinary relevance, we designed and conducted a course that brought together students from both health and engineering disciplines., ensuring future clinicians and technology developers alike cultivate these essential skills.

In this study, a pre- and post-survey evaluation of a master course is used. The Interactions in Health and Technology course focuses on interdisciplinary collaboration, technological innovation, and user-centred design in healthcare. Students work in groups (4–6) to develop case proposals addressing health-technology issues, often engaging with external stakeholders. The course includes lectures, hands on workshops, and case-based learning, emphasizing AI integration, ethical considerations, and sustainable healthcare solutions. Key topics include technological research, system design, user engagement, and innovation, with contributions from health services and industry professionals. Through a mix of theory and hands-on work, students gain insights into public and private stakeholder roles while addressing real-world healthcare challenges.

The data were collected electronically via a pre-survey and a post-survey. The pre-survey took place about 2 weeks before the course. Data on the students’ professional background, previous experiences with working interdisciplinary and with technology innovation and experience using AI, was reported.

The post-survey was conducted immediately after the course. This survey was a course evaluation and additionally contained the following open-ended questions related to AI’s role in their future clinical practice and studies:

We analysed open-ended responses from the end-of-course survey (n = 34), using reflexive thematic analysis (RTA) (Braun and Clarke, 2019; Braun and Clarke, 2021) from a constructivist position. Our approach was primarily inductive and semantic, focusing on participants’ explicit meanings. Analysis proceeded iteratively through (i) familiarisation (repeated reading, note-taking), (ii) initial coding of all relevant segments in Excel, (iii) collation of codes into candidate themes, (iv) review of themes against coded data and the full dataset, and (v) defining and naming themes by specifying each theme’s central organising concept. To support transparency and reflexivity, the primary analyst maintained analytic memos, an audit trail, and evolving theme maps. All co-authors were given access to the anonymised raw data and coding materials and were invited to interrogate and, where appropriate, challenge emerging interpretations in regular “critical-friend” discussions; some co-authors independently read the full dataset or subsets to probe alternative explanations and refine theme boundaries. We actively sought variation and contrasting examples and selected illustrative quotations to convey both typicality and range.

The DigCompEdu framework was used as a sensitising framework at the interpretation/reporting stage (not as a deductive codebook during coding) (The Joint Research Centre: EU Science Hub). The DigCompEdu describes 22 competencies, organized in the following six areas (see Figure 1): Professional engagement (A1); Digital resources (A2); Teaching and learning (A3); Assessment (A4); Empowering learners (A5); Facilitating learners’ digital competence (A6). We translated typical educator-oriented areas into learner-facing competence descriptors to help articulate and name patterns of shared meaning without constraining inductive coding. To illustrate themes, we include 1–2 anonymised verbatim quotations per theme in the Results. Quotes were minimally edited for readability (ellipses for omissions; brackets for clarifications) and, where applicable, translated into English from Norwegian. In Table 1, the column “Evidence from data (illustrative)” presents participant-proximal, synthesised data extracts, short paraphrases that preserve participants’ phrasing and meaning. These extracts provide descriptive evidence (not interpretive claims).

For RQ1 (anticipated uses of AI), we coded all segments that described intended or envisioned applications of AI in study/clinical work (e.g., evidence triage, documentation support, scoping/structuring), grouped these into patterns of practice, and then situated them at the interpretation stage within relevant DigCompEdu areas (primarily A6, with contributions from A1–A3). For RQ2 (risks and conditions for responsible use), we coded statements naming downsides (privacy, bias, accuracy/“hallucinations,” over-reliance) alongside enabling conditions (guidance, governance literacy, hands-on training, time/tools for verification) and interpreted them against A6.4 Responsible use (and A1 for organisational supports). For RQ3 (innovation competence), we operationalised a learner-facing, service/process view, the ability to identify opportunities, run small tests of change, verify/mitigate risks, and collaborate across disciplines, and used these facets as interpretive prompts (not a deductive codebook) when defining and naming themes, consistent with DigCompEdu (A6/A1) and PDSA-based quality-improvement approaches (Redecker and Punie, 2017; Taylor et al., 2014; Langley et al., 2009).

Norwegian Agency for Shared Services in Education and Research (SIKT) approval (with ref. number 397521) was applied and obtained for the study. Participants were informed of their right to withdraw at any time without any consequences. Written informed consent was obtained from participants.

Students envisioned AI becoming embedded in everyday clinical work, such as streamlining documentation, triaging information, and supporting decision-making pathways. They expected clinicians to act as informed evaluators, translating algorithmic outputs into clinically responsible actions while preserving professional accountability.

“AI can support monitoring, interpret medical images, assist in surgery, and help track patient flow in the emergency department.”

“Recognizing more pathology on radiologic images and providing faster on-site results… could reduce radiologists’ workload.”

By the end of the course, many reported concrete plans for using AI to scope and structure their learning tasks, complemented by verification routines and citation management. Students framed AI as a study aid that supports their academic reasoning and self-regulation rather than replacing it.

“I’ll use AI more in my studies, but in a critical way.”

“This experience encourages me to integrate AI for research, organisation and idea generation to improve efficiency in my studies.”

The exposure gained during the course led some to reframe AI from being merely a ‘tool’ to being a ‘collaborator’ that facilitates communication across professional boundaries. Students saw themselves as translators between clinical and engineering perspectives, emphasizing collaboration while retaining a patient-centred approach to decision-making.

“Automate hospital processes to free time for demanding tasks and shorten patient wait times.”

“Streamline booking, coordination and paperwork, fewer phone calls and a more efficient workflow.”

Optimism about AI was tempered by caution. Students frequently raised concerns about privacy, bias, accuracy (including ‘hallucinations’), and the risk of over-reliance on AI. They highlighted the need for governance literacy, institutional guidance, and adequate time and resources for verification to enable responsible and well-supervised use of AI.

“AI will have a big influence, but safety questions must be solved before it’s safe to use.”

“I learned how to use it properly and safely, e.g., never share sensitive information with ChatGPT.”

In line with RQ3, students’ adjustments to their workflow using AI, such as templating notes, triaging information, and scaffolding decision paths were seen as micro-innovations. They considered iteration, testing, and collaboration with technical peers essential for ensuring that AI remains useful, safe, and centred on patient care.

“Increase efficiency and free time for patients, AI could manage booking, provide patient-centred information, and help coordination.”

“Automate routine tasks and improve data analysis, enhancing decisions and letting professionals focus on critical and creative work.”

We aligned each theme to the DigCompEdu framework to indicate where competencies cluster and where gaps remain (Table 1). In Table 1, the ‘Evidence from data (illustrative)’ column presents participant-proximal, synthesised extracts (concise paraphrases of respondents’ wording).

As Table 1 shows, evidence concentrated in A6 (Facilitating learners’ digital competence) and A1 (Professional engagement), whereas A4 (Assessment) was sparsely represented, highlighting a need to strengthen evaluative judgement and feedback practices.

Students engaged most with A6 (specifically, A6.1, A6.3, A6.4, A6.5) and A1, while A4 (Assessment) remained under-represented. Rather than a shortcoming, this is a constructive diagnostic that can guide curriculum design. Low-stakes activities, such as utilizing AI-feedback vs. rubric comparison, criteria calibration with exemplars, and conducting brief reflect-revise justifications, can enhance coverage of A4 and develop evaluative judgement without undermining human oversight (Redecker and Punie, 2017; Rincón et al., 2025).

The recently launched DigComp 3.0 redefines AI as a transversal competence that cuts across other digital domains, cybersecurity, rights/responsibility, wellbeing, and mis/disinformation (The Joint Research Centre: EU Science Hub, 2025). This shifts the focus from merely ‘using a tool’ to exercising judgement, governance, and safe practice within each of these domains. In health-profession education, this means embedding AI literacy in these relating areas. For example, pairing the use of models with privacy/security routines; situating prompting and verification within information/media literacy; and aligning assessment with responsibility and critical appraisal rather than treating AI as a stand-alone topic.

Students demonstrated nascent service/process innovation through AI-supported micro-innovations (templating, triaging, decision-pathway scaffolds), paired with explicit verification routines and plans for health–engineering collaboration. Read through an ecosystem lens, these practices reflect co-creation and orchestration of service change that clinicians often lead and map to DigCompEdu A1 (Professional engagement) and A6 (digital problem solving/responsible use) (Lusch and Nambisan, 2015; Pikkarainen et al., 2022).

A useful distinction remains service/process innovation (new ways of organising care, workflows, and delivery) versus product/technology innovation (new tools, devices, software). The former is typically clinician-led within local ecosystems; the latter usually requires engineering/industry partnership and platform capabilities (Greenhalgh et al., 2004; Omachonu and Einspruch, 2010). To cultivate the service-innovation capability, programmes can embed short Plan–Do–Study–Act (PDSA) cycles, use a simple implementation canvas (problem → proposed change → verification plan → risks/oversight), and run brief usability walk-throughs with end users. Each micro-change should include lightweight metrics (e.g., time-on-task, error/omission rate, patient understanding) and A6.4 Responsible use checks (privacy, bias, documentation). This moves beyond “implementation” toward digital-health service design, while reserving product-level development for co-design with engineering/industry (Kosiol et al., 2024).

Students view AI as a collaborator rather than just a tool, which supports their identities as AI-augmented professionals who retain judgement, communication skills, and empathy (Ackerhans et al., 2024; Cruess et al., 2014). To normalise this identity without undermining human agency, programmes can embed brief, structured reflections. Such as error audits, accountability maps, and short memos on “where my professional judgement adds value,” linked to A1 (Professional engagement) and A6.4 (Responsible use) (Braun and Clarke, 2019; Braun and Clarke, 2021).

Given the mixed cohort, students anticipate collaboration between health and engineering regarding usability, integration, and governance. Short co-design sprints mirror real-world deployment, build a common language, and cultivate habits that make AI both usable and safe (Greenhalgh et al., 2004; Guidance WHO, 2021; Rincón et al., 2025). This interprofessional stance aligns with A1 (Collaboration/Professional engagement) while product-level development remains reserved for partnerships with engineering and industry.

Students’ optimism is consistently paired with ethical caution, particularly regarding privacy, bias, and conditions for deployment (A6.4 Responsible use), rather than considering them barriers. In the context of DigCompEdu, these concerns fall under A6.4 (Responsible use) and point to specific supports: guidelines, regulatory/governance literacy, hands-on training, and time/tools for verification (Guidance WHO, 2021; Vuorikari et al., 2022).

A practical programme-level support framework could include: (i) concise data-handling and prompt-safety guidance; (ii) brief model/feature cards that state known limitations; (iii) embedded verification routines (dual-source checks, error logs); (iv) short governance primers (risk categories, documentation, accountability); and (v) targeted faculty development. These measures operationalise responsible use and align with sector recommendations (Rincón et al., 2025; Mainz et al., 2024).

To convert cautious optimism into practical competence, programmes should combine hands-on AI with verification routines, governance literacy, assessment activities focused on A4, such as comparing AI-feedback to rubrics and calibrating criteria, interprofessional co-design, and faculty development. This approach embeds the DigCompEdu competencies A6 (information & media literacy, responsible use, digital problem solving) and A1 into everyday learning. It responds to health-education and workforce bodies’ calls for structured, experiential AI training (Guidance WHO, 2021; Redecker and Punie, 2017). As a result, graduates will not only be more competent with tools but also ethically grounded and prepared for the future.

In interpreting these findings, several limitations of the study should be acknowledged. First, the study relied on self-reported data rather than objective measures of behaviour or performance. While self-reports capture students’ perceptions and intentions, they are subject to biases, such as, social desirability or recall bias that may not accurately reflect actual engagement or learning outcomes (Chan and Hu, 2023). Second, the participants were drawn from a single master’s-level course at one institution, which limits the generalizability of the results beyond this context (Divya et al., 2022). The specific educational setting and student population in this course may not represent other programmes or institutions, so caution is warranted when extrapolating these findings to broader health professions education. Finally, the short timeframe between the baseline and end-of-course surveys (within one academic term) captures only immediate changes in attitudes, not long-term effects. It is unclear whether the observed shifts in perspective would persist or translate into practice over time, underscoring the need for longitudinal studies to track sustained impacts (Chan and Hu, 2023).

Priority next steps are (1) to follow graduates longitudinally to test whether AI-integrated education predicts responsible use (A6.4), evaluative judgement (A4), and effective adoption in practice, triangulating self-report with observed tasks, artefact analysis, and OSCE-style assessments; (2) comparing pedagogical models (dedicated AI courses vs. integrated modules; simulation-based vs. classroom) using common outcomes (AI literacy, information/media practices, verification behaviour, ethical reasoning) (Redecker and Punie, 2017; Rincón et al., 2025); (3) conducting multi-site/international replications to examine curricular and cultural effects; (4) operationalizing service-innovation competence by measuring micro-change cycles (e.g., PDSA), interprofessional co-design, and lightweight metrics (time-on-task, error/omission rates, documentation quality, patient understanding); and (5) evaluating A4-oriented assessment designs (AI-feedback-vs-rubric comparisons, criteria calibration, reflect-revise justifications) for their impact on students’ evaluative judgement (Redecker and Punie, 2017; Rincón et al., 2025).

Instead of creating new frameworks, it is advisable to align curricula and accreditation with DigCompEdu and the recent DigComp 3.0, which treats AI as a transversal competence that intersects cybersecurity, rights/responsibility, wellbeing, and mis/disinformation (The Joint Research Centre: EU Science Hub, 2017, 2025). Specifically, oversight bodies should: (a) specify core AI/digital-health competencies mapped to DigCompEdu areas and embed them in programme learning outcomes and assessment blueprints; (b) require faculty development so educators can confidently supervise AI-assisted learning; and (c) provide resource for governance and ethics primers, time/tools for verification, and interprofessional co-design to reflect real deployment (Bekiaridis and Attwell, 2024; Redecker and Punie, 2017; Vuorikari et al., 2022). At the institutional level, clear AI-use policies, seed funding for curriculum innovation, and platforms for sharing exemplars/rubrics will help programmes keep pace while safeguarding quality and equity (Mainz et al., 2024; Shishehgar et al., 2025).