Exploring Human–AI Interaction in Primary Healthcare: A Qualitative Study ^†

Abstract

While artificial intelligence is rapidly reshaping the healthcare sector, it is important to assess readiness for AI integration to properly prepare healthcare professionals, particularly in countries where clinical AI systems have not yet been implemented in primary healthcare. This qualitative study explores healthcare professionals’ perceptions of the future use of Artificial Intelligence in Greek Primary Healthcare settings. Two focus groups were conducted with 18 Primary Health Care professionals working in a health center and a local primary healthcare unit (TOMY) in Greece. Thematic analysis identified six major themes: potential uses, challenges and risks, ethical concerns, readiness and training needs, trust factors, and AI impact on professional roles and skills. Future research should focus on developing training programs, establishing ethical and regulatory frameworks, and examining the long-term impact of Artificial Intelligence on professional roles, skills, and interprofessional collaboration in Primary Healthcare.

1. Introduction

Workforce shortages, ageing populations, rising service demand, and increasing multimorbidity have placed significant strain on healthcare services [1]. These pressures not only affect the capacity of the system but also the quality and safety of care. In daily practice, health professionals are confronted with uncertain and complex situations that require time-critical decision-making, interpretation, and synthesis of high-volume information, and moral dilemmas. Moreover, increasing working hours, emotionally demanding working environments, and interpersonal workplace relations contribute significantly to burnout among healthcare professionals [2]. Artificial intelligence (AI) is rapidly reshaping the healthcare sector. AI is often framed as a solution to these challenges, promising gains in effectiveness and efficiency. Although technological advancements are accelerating, healthcare professionals often appear unprepared to seize the opportunities and to address the challenges.

Particularly, in Primary Healthcare (PHC), AI has been recognized as a promising tool for improving workflow efficiency, clinical decision-making, and administrative support [3]. However, the crucial question is not how technically advanced AI systems are, but how prepared health professionals are to manage them in ways that ensure the safety, quality, and human character of AI-supported care.

The existing literature underpins concerns regarding trust, ethical accountability, data privacy, and professional autonomy [3,4,5]. Previous studies have shown that healthcare professionals generally perceive AI as a supportive tool rather than a replacement for human clinicians, emphasizing the importance of human oversight and professional control [3,5]. Furthermore, education, digital readiness, and structured governance frameworks have been identified as essential prerequisites for the safe and responsible integration of AI into healthcare practice [4,5].

This study explores healthcare professionals’ perceptions regarding the future use of AI, in Greek PHC settings, focusing particularly on expected benefits, potential risks, ethical concerns, and their readiness for human–AI-assisted care. Despite the growing international literature on AI in healthcare, evidence from Greek PHC remains limited. Moreover, healthcare professionals in Greek PHC have no direct experience with clinical AI systems. Exploring healthcare professionals’ perceptions of human–AI-assisted care before implementation provides meaningful insights for planning, training, and ethical policymaking. Building digital and ethical readiness today will determine how responsibly AI shapes the future of PHC.

2. Materials and Methods

A qualitative design was employed to investigate healthcare professionals’ perceptions regarding the future integration of AI in PHC. Focus groups were selected as an appropriate qualitative method because they promote direct interaction among participants and allow the exploration of shared experiences, attitudes, concerns, and expectations. The study involved healthcare professionals who were working in a Health Center and a local PHC Unit (TOMY) in Greece. A total of 18 healthcare professionals participated in the study, including 3 physicians, 8 nurses, 1 midwife, 3 health visitors, and 3 administrative staff members. A purposive sampling strategy was used to ensure an interprofessional approach and to include professionals from diverse backgrounds. This allowed the collection of multiple perspectives regarding the potential implementation of AI technologies in daily healthcare practice.

Data was collected through two focus groups using a semi-structured discussion guide, which was developed based on the existing literature on AI in healthcare. The interview guide included an introductory open-ended question that invited participants to describe their views regarding the use of AI, in Greek PHC settings. Additional questions were used to further explore the participants’ perceptions concerning four main axes: (a) potential uses of AI, (b) perceived challenges, (c) healthcare professionals’ readiness, and (d) potential impact of AI on professional roles and skills. The discussion focused on participants’ perceptions of human–AI-assisted care. Before the start of the focus group, a questionnaire was administered to each participant to collect demographic data, including work experience, age, educational level, and job responsibilities (

Table 1. Demographic and Professional Characteristics of Participants.

Participant ID	Age	Gender	Professional Role	Educational Level	Years of Work Experience in Healthcare	Years of Work Experience in PHC	Previous Experience with Clinical AI	Previous Training in Clinical AI
P1g1	46	Female	Midwife	BSc	17	3	None	None
P2g1	57	Female	Nurse Nurse	BSc	35	32	None	None
P3g1	34	Female	Nurse Assistant	Post-secondary vocational education	10	5	None	None
P4g1	62	Male	Nurse	BSc	33	33	None	None
P5g1	55	Male	Physician	MD, MSc	25	20	None	None
P6g1	48	Female	Administrative staff	BSc	24	22	None	None
P7g1	35	Female	Nurse	MSc	8	6	None	None
P8g1	43	Female	Health visitor	BSc	18	17	None	None
P9g1	37	Male	Administrative staff	MSc	8	5	None	None
P10g2	51	Female	Administrative staff	BSc	25	6	None	None
P11g2	52	Male	Physician	MD, MSc	23	18	None	None
P12g2	40	Female	Health visitor	MSc	8	7	None	None
P13g2	36	Female	Health visitor	BSc	10	8	None	None
P14g2	37	Female	Nurse	MSc	16	5	None	None
P15g2	33	Female	Nurse	MSc	8	7	None	None
P16g2	36	Female	Nurse	BSc	8	6	None	Limited
P17g2	40	Female	Nurse	MSc	13	5	None	None
P18g2	65	Female	Physician	MD, MSc	40	40	None	None

BSc = Bachelor of Science; MSc = Master of Science; MD = Medical Degree, PHC= Primary Healthcare.

). The focus groups were conducted in-person, audio-recorded with participants’ consent, and transcribed verbatim for analysis. Two researchers were involved in conducting the focus groups: one acted as the moderator, facilitating the discussion, while the second served as an observer and took field notes. The duration of each focus group was 65 min for group 1 and 75 min for group 2.

Thematic analysis was conducted according to Braun and Clarke’s six stages approach [6]. An inductive coding approach was used. Two researchers reviewed and coded the transcripts independently, and any differences were discussed until agreement was reached. Codes were grouped into themes, which were reviewed and refined through discussion among the researchers. The thematic analysis of data revealed six main themes: potential uses, challenges and risks, ethical concerns, readiness and training needs, trust factors, and AI impact on professional roles and skills. Data saturation occurred after the second focus group, at which point information redundancy was reached and no new information was revealed. Therefore, data collection was concluded.

Participation was voluntary, and informed consent was obtained from all participants before data collection. The study was conducted in accordance with the principles of the Declaration of Helsinki. Confidentiality and anonymity were maintained at all stages of the study.

3. Results

At the beginning of the focus groups, two introductory questions were posed to participants about their existing experience with AI applications and the spontaneous association of the term “artificial intelligence” with a word. All participants had no previous experience in the use of any specialized medical application of AI in their work. Their experience was limited to general tools, such as ChatGPT, Google Gemini, and Microsoft Copilot, for information search and managing daily tasks. When participants were asked to spontaneously associate the term AI with a single word or phrase, the responses reflected both positive and negative perceptions, including terms such as “robot”, “machine”, “friend”, “helper”, “easy solution”, and “fast information”, as well as concerns related to “misinformation”, “frisk”, “fimpersonal communication”, “funcertainty”, and “fear”.

3.1. Potential Uses

Participants identified several potential applications of AI in PHC. Particularly, they discussed the support of clinical decision-making, diagnosis, treatment planning, and the organization of community health actions. AI was perceived as a useful tool for improving administrative organization, vaccination scheduling and monitoring, automated appointment management for patients with chronic illnesses, and follow-up of prevention programs. The midwife said (P1g1), “It could help us provide more individualized follow-up for women undergoing mammography and Pap testing”.

Many participants emphasized that AI could help reduce bureaucracy and save valuable time, which will be used to direct patient care. Additional perceived benefits included support in emergency triage, telephone assessment of incidents, organization of rotating work schedules in a faster and fairer manner, and simplification of supply and administrative procedures. Participants also described AI as a possible “safety net” in nursing interventions. Εspecially, in tasks like patient identification and medication management, AI could help nurses reduce clinical errors and promote patient safety. A nurse (P14g2) said, “AI could promote patient safety, checking patients’ details and medication preparation like a second pair of eyes, like a nurse assistant”.

Some participants believed that AI could strengthen nursing autonomy by supporting independent clinical decision-making in situations where nurses currently have limited decision-making capacity. A nurse (P15g2) reported that: “I believe that AI could widen the field of our decisions and responsibilities, empowering nurses’ roles”.

3.2. Challenges and Risks

Potential risks and challenges associated with AI integration in PHC were expressed by the participants. Concerns involved issues of trust, confidentiality, data privacy, and accountability. Participants emphasized the potential risk of sensitive personal data leaks and highlighted uncertainty regarding responsibility in cases of errors associated with AI-assisted care. A physician (P11g2) raised the question, “That’s the serious question. Who will be responsible if AI makes a mistake?”.

Some participants also expressed concerns that increasing automation could negatively affect the human-centered nature of healthcare interactions. Fears were raised that over-reliance on AI could lead to alienation, isolation, and a gradual loss of the human dimension of care, negatively impacting the therapeutic relationship between healthcare professionals and patients. As one nurse (P14g2) noted:“Healthcare is based on human contact. I am concerned that excessive reliance on technology could reduce the personal interaction between healthcare professionals and patients.”

Additionally, several participants expressed concerns about the potential job displacement, particularly among administrative staff, due to the process automation of bureaucratic and administrative processes. A member of the administrative staff (P9g1) expressed concerns about potential job displacement: “I believe that many administrative tasks could be automated in the future, which may reduce the need for administrative staff.” Furthermore, all participants argue that AI should serve as a supporting tool rather than a substitute for healthcare professionals.

Major concerns included misinformation or insufficient information provided to both patients and professionals, and excessive dependence on AI systems. A health visitor (P12g2) stated that “A patient may come with wrong information from an AI application, and this can negatively affect trust and communication with healthcare professionals”.

3.3. Ethical Concerns

Participants expressed strong ethical concerns about the use of AI in the field of healthcare. One of the key issues that emerged was the lack of emotion, empathy, and moral judgment by AI systems, as they were deemed to be unable to fully understand human needs and the emotional dimension of care. Several participants reported that AI-generated responses are often generalized and cannot be meaningfully tailored to each patient’s individuality, values, and specific needs. At the same time, it was pointed out that AI has difficulty considering cultural, social, and value parameters that influence health decisions. A nurse (P3g1) noted that “AI has no feelings, no empathy for patients and colleagues. It will not understand nonverbal communication and specific circumstances, as a health professional could understand”.

Participants also raised ethical concerns regarding privacy issues, with several participants expressing concern about how sensitive health information is managed and stored. A health visitor (P13g2) highlighted concerns regarding confidentiality and data protection: “Health information is highly sensitive, and I would be concerned about who has access to these data, how they are stored, and whether patients’ privacy can be fully protected”. A nurse (P15,g2) added, “If sensitive patient data were leaked, healthcare professionals could still be held responsible?”

In addition, several participants reported that the boundaries of the ethical use of AI in healthcare are not yet clearly defined, which creates uncertainty and concern. The midwife (P1g1) shared that “It isn’t clear to me when AI use is ethical and when it is not. There is a need for clear boundaries and guidelines”.

3.4. Readiness and Training Needs

Participants expressed different opinions on the readiness of healthcare professionals to use AI tools in PHC. Younger healthcare professionals seem to feel more prepared and familiar with digital technologies and AI tools, while senior professionals often appear more cautious or less prepared to incorporate them into daily practice. A younger nurse (P7g1) commented: “I already use digital tools in many aspects of our daily lives, so I think adapting to AI would be easy for me.” In contrast, a more experienced nurse (P4g1) stated: “I am open to AI, but I would prefer a gradual introduction and appropriate training before using it in practice.”

Participants emphasized the need for reliable and balanced information about both the positive elements and potential risks of AI in healthcare. A nurse (P2g1) stated, “I’m not ready! For me, it is very important to have comprehensive information about the positive aspects and risks of using AI.”

Comprehensive and continuous training was considered a prerequisite for the safe and effective use of AI tools in clinical practice. Several participants reported that the training should not be limited to the technical use of the tools, but should also include ethics, privacy protection, AI limitations, and critical evaluation of AI-produced results.

3.5. Trust Factors

Participants described various factors that affect their trust in the use of AI in healthcare. Human oversight and professional control emerged as key prerequisites, as participants emphasized that healthcare professionals should retain responsibility for final decisions and oversight of AI-assisted processes. A physician (p18g2) said: “AI may provide useful recommendations, but the final decision should always remain with the healthcare professional.”

The need for a clear institutional and organizational framework for the competent bodies and health administrations to regulate the use of AI tools was highlighted. An administrative employee (P10g2) mentioned that “I need to know that the system I will be using is safe and officially approved by the Ministry of Health”. Some participants also mentioned the need for a license or certification to use such tools.

Overall, trust in AI appears to be linked to maintaining human control, protecting data, and ensuring the human role in healthcare services through comprehensive frameworks and official guidelines.

3.6. AI Impact on Professional Roles and Skills

All participants believed that the use of AI would bring about changes in the roles of health professionals in PHC. However, senior participants appeared more thoughtful, expressing the view that in the next decade, no particularly significant changes in daily practice are expected.

Significant differences were observed depending on the professional specialty of the participants. In particular, nurses more often reported possible changes related to strengthening autonomy and expanding their role, while other professional teams focused more on transforming work processes or automating administrative tasks. Participants expressed different views on AI-related job replacement and job losses. Physicians, nurses, and health visitors felt their role would be strengthened, whereas administrative staff worried they could be replaced or that their role would change. A nurse highlighted the importance of the human element in care: “I think AI could rather strengthen than replace nurses. Our work involves direct patient contact, clinical judgement, and emotional support, which require a human presence.” (P17g2, Nurse) Nurses and physicians seemed more confident that their work would not be negatively affected in the future, mainly due to the nature of their work, which is inextricably linked to accountability and the physical presence of patients, which presupposes the performance of their duties.

Participants also raised concerns regarding reduced professional initiative, impaired critical thinking and decision-making abilities, and ambiguity in professional roles and responsibilities among healthcare professionals and patients. Participants expressed mixed views on the impact of AI on the skills of healthcare professionals. A nurse assistant (P3g1) supported that: “If used properly, AI could function as an educational tool and help healthcare professionals develop their skills and expand their knowledge.” Several mentioned that over-reliance on AI may lead to a decrease in collaboration and interpersonal communication among healthcare professionals. Some participants feared that the continuing use of AI tools might limit active information processing and the development of essential professional skills. A health visitor (P8g1) noted, “Health professionals may become too dependent on technology and stop thinking critically. If we don’t use our skills in everyday practice, it might be a risk to lose important professional abilities and skills.”

However, other participants argued that, when AI is used correctly and with appropriate training, AI can contribute positively to skill development, enhancing access to knowledge, faster information processing, and ongoing professional growth.

4. Discussion

The present study explored healthcare professionals’ perceptions regarding the future integration of AI in PHC. The findings suggest that participants recognize the potential of AI to improve efficiency, support clinical decision-making, and reduce administrative workload. This study contributes to the limited evidence on AI in Greek PHC by providing qualitative insights into healthcare professionals’ perceptions, expectations, and concerns regarding AI integration. It highlights the coexistence of optimism and uncertainty surrounding AI adoption and identifies key factors influencing trust and preparedness, including training, human oversight, and ethical considerations. These findings may help inform future research, educational initiatives, and policy discussions in the Greek PHC context. Participants had no direct experience with clinical AI systems. As a result, their views were largely based on familiarity with general AI tools, and may differ from those of healthcare professionals who use clinical AI applications in practice. For this reason, the findings may also be relevant to healthcare settings where professionals, similarly to the participants in this study, have no previous experience with clinical AI systems and are at a preliminary stage of AI adoption.

These findings are consistent with previous research evidence highlighting the transformative potential of AI in PHC [3,7,8]. At the same time, important concerns emerged regarding ethical issues, professional autonomy, data protection, and the retention of human-centered care. Similar concerns regarding trust, accountability, confidentiality, and professional responsibility have also been identified in previous qualitative studies exploring AI implementation in healthcare settings [4,5,9,10].

Trust in AI systems appeared closely related to transparency, reliability, and the ability of healthcare professionals to remain actively involved in decision-making processes. These findings support the existing literature suggesting that healthcare professionals are more likely to accept AI technologies when they are integrated as assistive rather than autonomous systems [4,9].

Furthermore, participants expressed concerns regarding the possible reduction in empathy and interpersonal communication in healthcare interactions. These findings highlight the importance of preserving the human-centered nature of healthcare during the integration of emerging technologies. Similar ethical concerns have been widely discussed in the literature, particularly regarding the risk of overdependence on automated systems and the possible weakening of interpersonal aspects of care [5,7,10]. The fear that AI may impair communication and collaboration skills also reflects broader concerns regarding the long-term impact of automation on healthcare professions.

The findings additionally revealed varying levels of readiness among healthcare professionals. The study highlights that professional role and scope of responsibilities may shape healthcare professionals’ perceptions, expectations, and concerns regarding AI use in PHC settings. Furthermore, younger participants appeared more familiar and comfortable with digital technologies, while many participants emphasized the need for structured education and continuous professional training. This finding supports previous research suggesting that digital readiness, AI literacy, and organizational preparedness are essential prerequisites for the successful implementation of AI in healthcare settings [4,5,9]. In addition, increasing workload pressures, workforce shortages, and burnout among healthcare professionals may further influence attitudes toward AI-assisted healthcare and the perceived need for supportive technologies [2].

The findings of the present study have practical implications for healthcare policy, professional education, and organizational planning. Healthcare organizations and policymakers should prioritize the development of governance frameworks and educational interventions before the large-scale adoption of AI systems within PHC settings. Emphasis should be placed on maintaining human oversight, strengthening digital literacy, empowering soft skills, and preserving empathy and patient-centered communication during the adoption of AI technologies. Given the exploratory nature of the study and its context-specific sample, these findings should be interpreted with caution. Future research involving larger and more diverse PHC populations is needed to further explore the impact of AI on professional roles, skills, and interprofessional collaboration in healthcare.

Overall, the study highlights that the successful integration of AI in PHC requires not only technological advancement, but also ethical governance, professional education, and strategies that ensure the supportive role of AI and promote human interaction in healthcare practice.

5. Limitations

This study presents several limitations that should be acknowledged. First, the sample is small and comes from two PHC organizations in a specific area of Greece, which may limit the transferability of the findings to other healthcare settings. Another potential limitation of this research was the pre-existing relationships among focus group participants, which may have encouraged groupthink or social desirability bias. However, this familiarity may serve as a strength, fostering a comfortable environment that allows for a more fluid and spontaneous exchange of ideas. To mitigate potential bias, the moderator actively prompted dissenting views to ensure a balanced collection of data, and the participants seemed to feel comfortable expressing their opinions, even if they were opposite to others. Furthermore, the study did not employ methodological triangulation, which may have limited the depth and validation of data interpretation. Although efforts were made to facilitate open discussion, the potential influence of the moderator cannot be completely excluded. Finally, participants discussed anticipated and hypothetical applications of AI rather than direct experience with clinical AI systems, which may have influenced the nature of their perceptions and expectations.

6. Conclusions

Participants in this study identified potential benefits of AI in supporting decision-making through the management and organization of patient data, and time savings that could be leveraged in providing personalized care. However, they also expressed concerns regarding data confidentiality, accountability, professional deskilling, and the possible loss of human control. Training and rigorous oversight of AI systems emerged as important factors for enhancing preparedness and trust. Given the exploratory nature of the study and its context-specific sample, the findings should be interpreted with caution and considered as preliminary insights into AI adoption in PHC. Future research should focus on developing training programs, establishing ethical and regulatory frameworks, and examining the long-term impact of AI on professional roles, skills, and interprofessional collaboration in PHC.