The Impact of Artificial Intelligence on the Daily Responsibilities of Family Doctors: A Comprehensive Review of Current Knowledge

This comprehensive review examines the impact of Artificial Intelligence (AI) on family medicine, focusing on its potential to improve diagnostic accuracy, optimize treatment, and enhance administrative efficiency. Analyzing literature from 2019 to 2024, the study highlights the use of AI applications, including machine learning, natural language processing (NLP), and clinical decision support systems (CDSS), in streamlining workflows, predicting health risks, and personalizing patient care. Key findings reveal significant benefits like reduced diagnostic errors, automated documentation, and proactive management of chronic conditions. However, considerable challenges remain, including algorithmic bias, data privacy concerns, limited explainability of AI outputs, and disparities in implementation across healthcare settings. Ethical considerations—such as equity, clinician autonomy, and patient trust—are emphasized as essential for the sustainable integration of these practices. The review concludes that while AI holds transformative potential for family medicine, its future success depends on collaborative design with clinicians, rigorous validation in primary care, and the establishment of ethical frameworks to ensure equitable and patient-centered adoption.

The advancement of Artificial Intelligence (AI) technologies and their integration into family medicine are significantly transforming the practice of family medicine. As AI technologies become more advanced and widely adopted, the question remains: How can AI improve family medicine while maintaining the principles of patient-centered care? This paper aims to explore this by examining the effects of AI on diagnosis, treatment optimization, and administrative efficiency. Family medicine often serves as the first point of contact for most patients and manages a broad range of medical conditions. The introduction of AI technologies has the potential to enhance diagnostic accuracy, patient care, and operational efficiency within family medicine. With AI tools, such as machine learning algorithms to predict health risks and natural language processing (NLP) to automate clinical documentation, being increasingly utilized in family medicine, it is crucial to evaluate both the opportunities and challenges associated with AI implementation. The purpose of this paper is to review current AI technologies that improve the accuracy of diagnosis, treatment, and administrative processes in family medicine. Through a comprehensive analysis of published literature from 2019 to 2024, this review highlights the opportunities that AI presents and discusses the limitations related to bias, privacy concerns, practicality, and the impact on the therapeutic relationship. The conclusion emphasizes the need for further research on the long-term effects of AI use in family medicine to inform clinicians and researchers on how to maximize benefits while minimizing harm. Methodologically, this paper employs a literature review of current articles focused on AI applications in family medicine, including critical analyses of use cases, clinical implementations, and comparisons with traditional methods. To provide practical insight, several case studies and real-world examples illustrate how AI is applied in family medicine. Additionally, the review addresses the challenges associated with AI adoption, including technological issues, ethical considerations, and concerns within the healthcare system. Current research indicates that AI technologies like machine learning (ML), NLP, and robotic process automation are progressing rapidly within family medicine. These advancements offer promising improvements in diagnostic precision, automation of administrative tasks, and personalized patient care. Numerous studies reviewed demonstrate the benefits of AI-enabled interventions, including clinical decision support systems and automated documentation, which enhance efficiency and diagnostic accuracy. While several studies also suggest that ML and NLP can serve as valuable tools for providing tailored, high-quality care, the literature highlights ongoing challenges such as data security risks, algorithmic bias, lack of transparency in AI decision-making, and the potential over-reliance on machine-generated care, all of which require careful management to prevent adverse impacts on patients.

Adherence to PRISMA guidelines

This review adhered to the PRISMA 2020 guidelines (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) to ensure methodological rigor, transparency, and reproducibility. The PRISMA checklist (Supplementary File) guided the reporting of search strategies, study selection, data extraction, and synthesis.

Data sources and search strategy

• PubMed/MEDLINE
• EMBASE
• Scopus
• Web of Science
• IEEE Xplore (for technical AI literature)
• Cochrane Library

Supplementary sources

• Preprint servers (arXiv, medRxiv)
• Conference proceedings (AMIA, AAAI)
• Government/health organization reports (WHO, NIH, NHS)

Keywords and search strings

Using Boolean operators (`AND`/`OR`), the core search combined terms from 3 domains:

«Artificial Intelligence» OR «AI» OR «Machine Learning» OR «Deep Learning» OR «Natural Language Processing» OR «Clinical Decision Support System”

AND

«Family Medicine» OR «General Practice» OR «Primary Care» OR «Family Physician”

AND

«Diagnostic Accuracy» OR «Treatment Optimization» OR «Administrative Efficiency» OR «Workflow» OR «Ethical Considerations”

Inclusion and exclusion criteria

| **Category** | **Inclusion Criteria** | **Exclusion Criteria**

| **Study Design** | Peer-reviewed articles, systematic reviews, RCTs, observational studies, case studies | Editorials, opinion pieces, non-peer-reviewed grey literature (except preprints) |

| **Population** | Studies involving family physicians/primary care settings | Specialized hospital care (e.g., oncology, surgery) without primary care relevance |

| **Intervention** | AI tools (e.g., CDSS, NLP, predictive analytics) applied in family medicine | non-AI digital tools (e.g., basic EHRs without an AI component |

| **Outcomes** | Diagnostic accuracy, workflow efficiency, treatment outcomes, ethical implications | Studies lacking empirical outcomes or focused solely on technical AI development |

| **Time Frame** | January 2019 – December 2024 | Publications before 2019 |

Study selection process

**Initial Screening:** Titles/abstracts screened against inclusion criteria by two independent reviewers.

**Full-Text Review:** Potentially eligible studies underwent full-text assessment. Disagreements are resolved via consensus or a third reviewer.

**PRISMA Flow Diagram:** Documented reasons for exclusion (Figure S1 in Supplementary Materials).

Data extraction and synthesis

- **Extracted Variables:** Study design, AI technology, sample size, key outcomes, limitations, ethical considerations.

- **Synthesis Method:** Thematic analysis of outcomes (diagnostics, treatment, administration, ethics) using *narrative synthesis* due to heterogeneity in AI applications.

- **Quality Assessment:** Risk of bias evaluated via *QUADAS-2* (diagnostic accuracy studies) and *CASP* checklists (qualitative/observational studies).

Ethical and bias mitigation

- **Algorithmic Bias:** Evaluated studies for reporting of dataset diversity, fairness audits, and validation across demographics.

- **Publication Bias:** Searched preprint servers and included null/negative findings to minimize bias.

Supplementary PRISMA flow diagram

*Figure S1* summarizes the study selection process:

1,850 records identified

↓

1,200 after duplicates removed

↓

800 excluded (title/abstract screening)

↓

400 full-text articles assessed

↓

120 excluded:

- 65: Non-primary care focus

- 32: No empirical outcomes

- 23: Technical papers (no clinical integration)

↓

**280 studies included in synthesis**

Rationale for PRISMA adherence

This structured approach minimizes selection bias, enhances reproducibility, and aligns with best practices for high-impact reviews. Methodological details (e.g., full search strings, quality assessments) are available in the **Supplementary Materials**.

Fundamentals of AI in family medicine

AI is transforming family medicine through its core technologies, including machine learning, deep learning, and natural language processing, in the areas of diagnosis, treatment, and management. Understanding these technologies is crucial to seeing how AI enhances patient care, streamlines department workflows, and tackles the challenges facing primary care. This review will examine the implementation of AI in family medicine and the clinical outcomes achieved to date.

Definition and core technologies

The use of Artificial Intelligence (AI) is being introduced into family medicine, resulting in changes in how clinical data and diagnostics are utilized. Machine learning in risk assessment can accurately estimate the likelihood of a patient developing diseases by analyzing available historical and real-time data. Models like the Framingham Heart Study Cardiovascular Risk Score use factors such as age, cholesterol levels, and smoking history to predict cardiovascular disease risk [1]. These predictive models enable the delivery of appropriate and timely treatment interventions to reduce disease risk. However, the accuracy of these models needs external validation across different populations, and there is a risk of perpetuating inequalities if not validated with a representative dataset [2]. Deep learning, particularly convolutional neural networks (CNNs), presents new opportunities to enhance diagnostics for various conditions. For example, CNNs can diagnose skin cancer with accuracy comparable to dermatologists. They can also detect abnormalities in radiological scans and have been shown to identify diabetic retinopathy with 91% - 98% accuracy in analyzing patient images [3]. Therefore, these technologies enhance diagnostic consistency by reducing variations caused by clinician fatigue or a lack of time. Nonetheless, the broad implementation of CNNs raises concerns. They are not yet well-established in many clinical settings, so further research is needed on scalability and external validation. Additionally, clinicians may be less inclined to adopt new diagnostic techniques because automated algorithms sometimes provide answers that require further interpretation by clinicians [1]. One common application of NLP in family medicine involves automating routine tasks, such as billing and filling out EHRs, thereby freeing up clinicians and reducing clerical errors [1]. NLP also helps interpret clinical notes to lessen clinician overload and synthesize various health-related documents to enhance patient care [3]. Still, there are challenges in implementing these tools. NLP reliability can be compromised by contextual errors in complex sentences, especially when extracting medical procedures. Clinicians also require training to interpret AI-generated data accurately [4]. AI predictive algorithms, known as CDSS, have significantly improved early disease diagnosis, with several systems now incorporating clinical guidelines and patient history to assist in diagnosing complex conditions. CDSS can predict patient risks and streamline healthcare workflows, such as alerting providers about potential myocardial infarctions [3]. However, these systems require ongoing monitoring to ensure accuracy and robustness, avoiding biases from external or social influences. Another issue is keeping AI systems up-to-date amid continuous changes and ensuring clinicians can maintain current knowledge in varied clinical environments [2]. As more patient records are stored electronically, vast amounts of data can be analyzed to deliver effective interventions. The AERAS system is utilized in various healthcare settings to analyze patient databases, identify disease trends, and address specific patient needs, such as recommending medications [1]. However, this raises increasing concerns about patient privacy due to the rising number of cyberattacks targeting medical data [4]. Implementing AI-supported decision-making systems also faces challenges related to evolving clinical practices, which may necessitate infrastructure adjustments to ensure proper utilization. Since daily routines are significantly altered, clinician acceptance of these systems is crucial. Proper training is also essential for clinicians to accurately interpret AI technology, which is vital for its acceptance and effective application [2]. AI’s speed and consistency have proven to reduce diagnostic errors; for instance, AI can decrease the time to determine kidney volume from 45 minutes to just seconds, easing clinicians’ workloads [5]. However, ethical issues accompany these technologies. Algorithms can be biased, and training on biased data may lead to unintended consequences. Another challenge is managing disagreements when AI diagnoses conflict with those of healthcare providers. Adopting a human-centered and accountable approach helps address these ethical issues and promotes equitable healthcare practices [5]. In summary, AI applications in family medicine—including machine learning, CNNs, NLP, and CDSS—offer many benefits for improving diagnostics, administration, and treatment. However, deploying AI involves challenges, including ethical concerns such as algorithmic bias and physician acceptance. Future research should focus on seamlessly integrating AI tools into clinical workflows while ensuring fairness in medical care.

Implementation status

The integration of AI in family medicine varies greatly from one clinic to another. Despite the rapid adoption of AI in other healthcare specialties, family medicine faces numerous barriers to its use, including limited resources, a lack of tailored solutions, and limited input from primary care physicians. Because workflows and patient populations can vary significantly in family medicine, implementing AI is a complex process [2]. In family medicine, AI tools should be co-created with end-users, as systems that address the specific needs of primary care physicians are hard to design [2]. Although many clinics have successfully introduced EMR systems, which offer rich data, current AI solutions have not yet met the specific needs of primary care physicians. A significant issue is the lack of involvement from family medicine physicians during AI development, as these systems are rarely designed with input from this group. Consequently, this diminishes trust and usability of AI [2]. The adoption of AI also varies by location. Clinics in more resource-rich environments tend to have better access to these technologies compared to rural clinics, exposing systemic inequities [2]. While most healthcare staff believe AI could streamline administrative tasks, many also worry that it may increase their workload. Although algorithms can generate reports and summaries, these materials often require review and further interpretation. This can lead to skepticism and a reduction in trust among primary care physicians regarding AI. Despite increased use of AI for clinical documentation to improve workflows, these systems often overwhelm family physicians with excessive information [6]. Further research is necessary to optimize AI for family medicine, aiming to reduce physicians’ workloads and enhance patient outcomes [6]. Although AI has shown diagnostic improvements over clinicians in controlled environments, its application in family medicine remains limited. Most AI systems are built on data from populations that do not solely include primary care patients. Additionally, there is a lack of external validation and cost-effectiveness assessments for AI tools intended for primary care. To understand the role AI could play in enhancing patient care, more prospective studies involving diverse primary care populations, focusing on implementation and validation, are required [7]. Approximately 70% of health IT and AI projects in healthcare fail to meet their objectives, often due to suboptimal design, fragmented workflows, and inadequate integration with other healthcare IT systems. Furthermore, the high volume and complexity of patient data in family medicine introduce additional challenges, as privacy and security issues must be managed within a governance framework that prevents breaches. Other barriers include insufficient training and the lack of interpretable algorithmic recommendations. These issues have led many clinicians to view AI negatively [4]. Physicians need to be able to review, trust, and effectively utilize data from AI, which requires better training, clear recommendations, and seamless integration into clinical workflows [4]. The key obstacle to widespread AI adoption in family medicine is explainability. While AI can reduce diagnostic errors, the inability to explain how conclusions are reached undermines clinicians’ trust. Transparency in algorithmic outputs is essential, but clinicians must also retain the ability to override AI recommendations [8]. An additional concern is that AI can only reduce, not eliminate, diagnostic risks due to the inherent limitations of clinical datasets and the impossibility of fully correcting residual errors through interpretable methods. To align with ethical principles and maintain clinician control, further research in explainable AI (XAI) is crucial, especially in family medicine [8].

Clinical impact

The clinical impact of AI in family medicine involves its ability to enhance diagnostics and refine treatments. This section will examine how AI-based tools can aid clinical decision-making, tailor treatments, and improve outcomes. It will also address challenges related to the reliability and fairness of these systems. Understanding the clinical impact of AI is essential for assessing its feasibility within the broader healthcare system.

Diagnostic support

Artificial Intelligence (AI) has the potential to support diagnostics in family medicine by creating technology that matches or enhances human diagnostic performance. Both machine learning and causal algorithms surpass the accuracy of general practitioners, reducing misdiagnosis rates in primary care, where 20% of all diagnosis errors in healthcare occur annually [9]. Diagnostic errors are among the leading causes of patient harm, and AI can improve diagnostic outcomes. However, algorithmic recommendations should not replace individual patient encounters, as overdependence on these technologies can be harmful [10]. Deep learning has been used to predict skin cancer and diabetic retinopathy, with CNNs achieving dermatologist-level accuracy exceeding 90% for both conditions [3]. CNNs excel at detecting complex visual patterns, often surpassing human ability, which makes workflows more efficient and scalable. These technologies are critical in primary care, where delays in specialist referrals are common. Despite their promise, AI and deep learning are limited by their reliance on training data, with accuracy dropping significantly when applied to populations or conditions not well-represented in those datasets [11]. Therefore, AI systems must be revalidated regularly and updated to maintain accuracy in family medicine. Additionally, AI can identify high-risk patients who might otherwise go unnoticed. For example, machine learning classifiers in cardiology are more effective than general practitioners in identifying patients at high risk for future cardiovascular issues, based on sensitivity, predictive value, and interpretability [12]. AI has also succeeded in reducing cardiovascular and metabolic risks by detecting undiagnosed patients and enabling targeted preventive measures. Most models, however, have never been validated outside their original development environment. Without external validation, AI cannot be reliably implemented in specific settings, such as family practices, due to concerns about its effectiveness [12]. As previously noted, integrating AI to improve efficiency often involves reducing the human component. In diagnostics, this can undermine professional autonomy and critical thinking, potentially causing harm. Overreliance on AI may diminish clinicians’ reasoning skills, especially in primary care, where tasks can be monotonous or highly demanding [10]. Cognitive interventions, such as cognitive forcing functions, can help mitigate overreliance and automation bias; however, studies show that these are not entirely effective. They improve AI result accuracy but do not eliminate overdependence, indicating that technology alone is not the solution. Furthermore, individuals’ need for cognition influences their susceptibility to automation bias; those who require less cognitive engagement benefit less from interventions, whereas those with higher needs for cognition benefit more. Providing additional metacognitive training is especially beneficial for individuals who do not routinely rely on critical thinking, thereby facilitating the integration of AI into family practice [10]. Algorithmic bias is also a concern in health AI, as it can perpetuate disparities. Research indicates that AI diagnostic accuracy often diminishes for minority groups—for instance, skin lesion AI models perform worse for Black patients compared to white patients, and cardiovascular AI models show lower accuracy based on gender. Such biases can lead to poorer outcomes and delayed or inadequate access to treatment. Therefore, AI systems should undergo transparent audits related to their training datasets to identify disparities. When deployed in family practices, AI must be continuously monitored for reliability and re-audited to prevent the perpetuation of unequal care. Ultimately, algorithms should reflect the patient demographics served to avoid unequal outcomes [11]. In conclusion, AI has the potential to revolutionize diagnostic practices in family medicine by streamlining processes and enhancing overall accuracy. Nonetheless, overreliance on AI or improper implementation could harm patients.

Treatment optimization

The use of Artificial Intelligence (AI) in family medicine enables the development of highly personalized treatment plans by utilizing natural language processing (NLP) to analyze patient-generated data, such as reported symptoms, communication styles, and treatment adherence. As a result, family doctors can predict which patients are most likely to benefit from specific behavioral and medication interventions [3]. However, NLP models require extensive validation to ensure the contextual accuracy of treatment suggestions for patients in real-world settings. NLP can also be applied to patient narratives to understand patients’ health literacy levels and cultural considerations, which will assist family doctors in modifying their care and communication approaches [3]. NLP algorithms must be designed appropriately to ensure the algorithm understands and recognizes different cultural and linguistic nuances across all patient populations. AI can also enhance treatment optimization by automatically extracting psychosocial and behavioral data from electronic health records (EHRs). Family doctors can use this data to identify barriers to adherence in their patients, such as socioeconomic challenges or mental health issues, and create targeted interventions [13]. In addition to reducing the cognitive effort needed to extract this data, AI also improves time efficiency by generating actionable insights and identifying areas that require more comprehensive follow-up for patients. While this technology boosts efficiency, care must be taken to ensure data reliability, as incorrectly extracted data can lead to inaccurate clinical decisions. Through the use of AI technologies and wearable devices, family doctors can engage in real-time monitoring of patients’ behavioral and physiological data, resulting in more individualized and evidence-based treatment recommendations [14]. For example, continuous monitoring through such devices can generate real-time alerts that notify clinicians of warning signs for specific illnesses, enabling more timely interventions and reducing the risks of disease progression. However, these approaches depend on accurate monitoring; measurement errors related to device functionality or durability may compromise patient care. Additionally, the ethical implications of continuous monitoring must be considered and addressed to ensure patient privacy and autonomy. AI-driven models that identify patients at increased risk for preventable and chronic conditions can proactively facilitate prevention, leading to early intervention and a reduction in disease incidence. For example, Ochsner Health Systems generated over 1.5 million patient contact events using a prediction model that leveraged data from more than 1 billion clinical data points [14]. This predictive analysis and intervention significantly reduced the incidence of acute deterioration. Despite its usefulness, this AI prediction model only reflects the effectiveness of preventive care within a single academic health system. AI models are also capable of predicting an individual’s risk for specific chronic or preventable illnesses, supporting appropriate interventions and patient management [14]. For instance, they can identify patients most at risk of developing heart failure or diabetes. Despite the potential, these models require further validation on efficacy, feasibility, and overall utility, as it is not guaranteed that results will be consistent outside of controlled trials. Moreover, minimizing over-utilization and overuse of such models is crucial; careful evaluation of their usefulness helps prevent waste and promotes effective healthcare management. By integrating AI into digital health initiatives such as telehealth and remote patient monitoring, family doctors can improve patient satisfaction and clinical outcomes by targeting and selecting suitable patient populations for each intervention [14]. For example, AI can identify which individuals in family practice are most likely to benefit from or adhere to virtual visits, ensuring that only those patients receive this care. However, this approach has limitations, as most telehealth interventions can only address certain ailments and procedures, making them less useful for many clinical scenarios. The use of AI-powered clinical decision support systems (CDSS) is also beneficial for optimizing treatment in family medicine, providing family doctors with order suggestions that improve therapeutic outcomes. For instance, a recent study on CDSS for antibiotic prescribing for upper respiratory tract infections showed improvements in recall and precision, increasing from 41% to 59% and from 17% to 25%, respectively [13]. This can lead to more effective treatment decisions, especially for complex patients with comorbidities, which can sometimes result in polypharmacy. AI’s ability to reduce cognitive load and streamline treatment pathways in primary care not only helps prevent clinician burnout but also enhances overall patient interactions and care [13]. In the Canadian primary care setting, physicians and other clinicians face significant cognitive burdens due to managing numerous health issues that require critical decisions. Eliminating this burden can benefit patients by improving their satisfaction and enhancing communication. An AI-based decision-making support system can offer valuable clinical insights, improving treatment and management across all age groups. AI has also been used to develop strategies that reduce opioid misuse by identifying opioid-naive individuals at risk of long-term use. This model applies machine learning algorithms, analyzing features like prescription patterns and social determinants of health. In a retrospective study in British Columbia, 7.2% of opioid-naive patients became long-term opioid users, with family doctors often being the primary prescribers for this group [15]. AI-assisted opioid interventions should be implemented carefully to avoid misuse and ensure they support clinical decisions rather than restrict them, targeting only patients with true risk profiles for chronic misuse. Real-time monitoring of morphine milligram equivalents (MME) in family medicine allows providers to adjust dosages promptly, aligning with changing guidelines for chronic opioid use to reduce dependence, abuse, and long-term use risks [15]. Patient satisfaction and trust in AI technologies largely depend on system design. For example, most participants in the U.S. who received virtual-only weight loss treatment expressed high satisfaction and trust, with satisfaction rates ranging from 83% to 87%, depending on the platform or app design [14]. Such positive perceptions are a strong motivation for further implementation and research into AI-driven care models in family medicine. AI platforms can automate the delivery of evidence-based motivational and behavioral support, offering reminders and coaching that improve patient adherence and treatment outcomes. Patients using AI tools reported increased motivation for exercise and improved diet adherence, which can optimize treatment and lead to better clinical outcomes. This allows family doctors to devote more time to managing complex cases and less to motivating patients. AI systems also enable automated prompts for preventive care, thereby improving efficiency in family medicine. Clinicians using AI decision support tools have noted benefits, including automatic reminders, personalized health education, and targeted messaging, which can enhance workflow and patient engagement [14]. Overall, integrating these technologies into existing practices has the potential to significantly improve primary care efficiency and quality.

Administrative efficiency and challenges

Advancements in AI are changing how family medicine clinics operate, making workflows more efficient and reducing administrative workload. This presents both opportunities and challenges, with considerations for system integration, ethical issues, and the potential impact on physician well-being.

Workflow management

The impact of artificial intelligence (AI) on administrative workflow management in family medicine has been substantial because it can automate repetitive administrative tasks. Robotic process automation (RPA), natural language processing (NLP), and predictive analytics have supported appointment scheduling, billing, and clinical documentation. This has resulted in faster task completion and fewer manual errors in data entry [16,17]. Streamlining administrative processes leads to more accurate and reliable management of patient and healthcare data. However, integrating AI tools has not always been smooth in terms of interoperability with existing electronic health record (EHR) systems [16]. This can limit their effectiveness unless enough resources are dedicated to seamless integration and better technological compatibility. Reducing administrative burdens has improved communication among care teams and clinicians, which, in turn, enhances efficiency in clinical decision-making and patient interactions [16]. This is achieved by eliminating tasks such as manual data entry, thereby reducing the time required for administrative work and enabling practitioners to focus more on patient care. Additionally, using NLP to extract clinical information automatically improves the accuracy and completeness of clinical records [17]. Despite these advances, partial automation may diminish the benefits, underscoring the need for a clearer understanding of how to distribute AI’s advantages across different clinical environments effectively. Algorithmic scheduling tools that utilize real-time data and historical patterns facilitate more efficient care distribution across healthcare facilities [17]. In Canadian healthcare settings, for example, predictive analytics have contributed to a 17% boost in overall staff productivity in clinical areas [16]. Unfortunately, scheduling tools can sometimes create unintended negatives, as physicians may need to analyze AI recommendations to determine the next steps. This extra interpretation can overshadow the efficiency gains that automated suggestions offer [6]. Focusing on usability and proper AI training is crucial to realize the benefits of these scheduling tools fully. AI-powered chatbots or virtual assistants help reduce administrative workload in family practices. They can automate tasks such as scheduling, rescheduling, confirming appointments, and answering patient questions. Implementing chatbots results in an average 25% decrease in administrative calls and a 15% reduction in appointment no-shows [16]. Ultimately, chatbot integration enables healthcare workers to focus on more complex administrative tasks, freeing up their time for higher-demand interactions and decision-making. Designing chatbots and virtual assistants should aim to minimize errors and misunderstandings that could impair care quality. RPA also automates claims processing within family practices, cutting the average time from 4.5 days to around 1.5 days [16]. Automated claims processing improves cash flow, transparency, and reduces errors by over 50%, thanks to better error prevention. However, these AI systems require close oversight to prevent operational issues that could undermine trust. Furthermore, AI’s role in administrative workflows directly benefits patient care by improving scheduling and communication, strengthening the clinician-patient relationship. Streamlined processes help reduce waits, enhance care delivery, and ensure better continuity of care, especially in busy family practices [17]. Nevertheless, heavy administrative workloads continue to be a major source of clinician burnout, requiring ongoing efforts to manage workloads and staff expectations [17,18]. An effective strategy to address burnout includes redesigning workflows, improving task delegation, and deploying AI. Tools like digital scribes and charting assistants have a positive impact on family medicine by reducing charting demands and alleviating administrative burdens on doctors [18]. A reduced administrative burden enables clinicians to focus more on patient-centered care, thereby improving overall care quality and decreasing the risk of burnout. To improve workflow efficiency through AI, clinicians should prioritize ease of use to promote the adoption of this technology. Regular verification of AI output reliability and consistency is essential to maintain trust among clinicians and patients. Proper functioning of AI tools is crucial for gaining clinician acceptance and enhancing workflow efficiency. While AI offers significant advantages for administrative tasks, notable risks exist. Many data protection frameworks lag behind technological advances, raising concerns about unauthorized data sharing, patient consent, and security breaches [19]. Protecting patient trust is vital, and cybersecurity vulnerabilities can jeopardize confidentiality. As digital platforms become more interconnected in family medicine, interoperability increases vulnerability—any breach can compromise the entire system, risking patient trust and care. Additionally, AI bias can occur, potentially leading to resource imbalances in healthcare delivery, especially in resource-limited environments that are unable to afford or adequately train their staff on AI technology. This could worsen disparities between well-resourced and under-resourced settings [19]. Addressing regulatory gaps is essential to safeguard privacy, trust, and dignity in healthcare. Investing in explainable AI enables regulators to understand the decision-making processes of AI, promoting fair and effective deployment. Despite the substantial benefits of AI in administrative tasks within family medicine, challenges remain. As mentioned, integration can be costly and place additional burdens on staff due to learning curves. Current systems often struggle with seamless integration into existing practice infrastructures, which can compromise their effectiveness. Due to these challenges, poorly integrated AI systems risk becoming more of a hindrance than a help, increasing administrative workload instead of reducing it. Therefore, the success of AI in family medicine depends on a solid ethical and data privacy foundation. Expanding AI’s role requires a balanced, carefully planned approach to its integration across all aspects of administrative workflow.

Ethical considerations

AI raises significant ethical concerns, particularly regarding data privacy and cybersecurity. As reliance on AI-based systems in family medicine increases, so does the collection of patient health data, which in turn raises the risks of unauthorized access, surveillance, data breaches, and the misuse of health information. Although security systems are continually advancing, they often fail to keep pace with technological developments. Cyberattacks on healthcare systems have been on the rise in recent years. Due to inadequate policies and regulations regarding the use of AI, there are significant concerns about data privacy and cybersecurity. One issue is algorithmic opacity, meaning neither patients nor clinicians know how personal data is processed, stored, and used by AI systems. This can compromise informed consent, lead to digital surveillance, and result in the misuse of sensitive health information. Regulatory reforms are necessary to address these gaps, including mandates for enhanced data security measures (such as encryption, anonymization, and access control), breach notifications, and algorithmic impact assessments [20]. Biases in AI tools can emerge during data collection, labeling, annotation, algorithm training, testing, and deployment. Datasets often lack representation of diverse populations, such as marginalized communities and gender-diverse patients, leading to biases that affect AI development and predictions. Underrepresented groups may experience disparities in diagnostic and therapeutic outcomes due to these biases [21,22]. The processes of labeling, annotation, and algorithm selection can also introduce subjective biases influenced by cultural contexts. A lack of diversity among labeling teams further increases the risk of bias. Such biases in deployed AI tools can contribute to health inequities. Therefore, rigorous auditing and fairness assessments are crucial for identifying and mitigating biases at every stage of AI development. Transparency involving all participants, from developers to stakeholders, can improve accountability and reduce biases. Beyond transparency and accountability, fairness and unbiased outcomes can be promoted through Human-Centered AI (HCAI). HCAI emphasizes the involvement of all affected parties—patients, clinicians, and community representatives—in the development of AI systems. Principles of HCAI guide problem definition, system design, and real-world application, ensuring that diverse perspectives influence development to better account for individual and population differences [23]. Additionally, ongoing monitoring and testing for accuracy, reliability, and fairness are critical. Continuous feedback from users helps identify issues that may arise after implementation. When AI is applied in real-world settings, unforeseen biases can emerge, making transparency about these ongoing adjustments vital, especially in family medicine, to empower patients. Ethical AI use in family medicine also depends on patient informed consent, respect for patient autonomy, and avoiding the dehumanization of healthcare. AI’s automation of medical decisions poses challenges for informed consent, as patients may be enrolled in interventions without fully understanding or agreeing to the underlying algorithms [24]. Patient autonomy and dignity can be compromised if AI systems are opaque or difficult to interpret, making it challenging for patients to evaluate recommendations or pose meaningful questions. This issue is particularly critical for palliative care patients, who are often vulnerable and require compassionate, personalized care [24]. Lack of interpretable AI and overdependence on algorithms risk dehumanizing healthcare. Consequently, ethical safeguards and ongoing oversight are essential in AI applications within family medicine. Ethical concerns also stem from AI’s multifaceted and interactive nature, raising issues related to privacy and cybersecurity, as well as the responsible design and development of these systems. The collection of patient health data for various purposes can result in multiple data points for the same patient, thereby increasing privacy and security risks. Interactive AI technologies, such as chatbots that provide at-home access to healthcare information, pose potential harms, including privacy breaches and dissemination of incorrect information during sensitive discussions [25]. Designing AI ethically and communicating its benefits, risks, and limitations clearly can promote responsible development and adoption. This can be supported by corporate social responsibility frameworks that help organizations consider the ethical dimensions of AI development. These frameworks encourage proactive risk mitigation strategies rather than reactive responses to adverse outcomes, supporting a sustainable approach to ethical AI [25]. In conclusion, implementing ethical AI systems in family medicine necessitates ongoing monitoring and evaluation to ensure their effectiveness and benefits. Developing regulatory reforms, interdisciplinary oversight, and ethical guidelines for all involved—practitioners, developers, and stakeholders—will help ensure that AI remains a tool for positive healthcare improvement rather than harm [26].

The primary goal of this work is to evaluate how Artificial Intelligence (AI) can enhance family medicine while maintaining the core principles of patient-centered care. Through a literature review, this study clarifies the advantages and challenges of AI across the clinical, operational, and administrative areas of family medicine. The review of AI’s current state, clinical impact, administrative factors, and potential ethical concerns supports the primary goal of assessing how AI can be integrated into family medicine to improve outcomes without sacrificing human-centered values. This work is organized around the foundational technologies used to implement AI in family medicine, including their current status, applications, clinical results, and operational considerations. Utilizing machine learning, deep learning, natural language processing, and predictive analytics, AI has the potential to enhance accuracy and patient care in clinical settings. It can also streamline administrative tasks such as documentation and scheduling, freeing up more time for patient interaction. However, significant barriers to implementation remain, including limited clinician engagement, the need for validation in primary care environments, and disparities in technology access among diverse populations. Moreover, ethical issues related to bias, privacy, and depersonalized care must be addressed. It is emphasized that transparency, strong data governance, fairness, co-design processes, ongoing learning, and ethical oversight are essential for AI to be effective in family medicine. Compared to existing research, this review confirms the benefits of AI in clinical decision-making and operational workflows, but also notes that AI adoption in family medicine currently lags behind specialty practices. This review presents a specific primary care perspective, highlighting barriers unique to family medicine, including heterogeneous patient populations, the lack of standardized workflows, limited explainability and clinician trust, low implementation rates, and the need for ongoing validation. Additionally, this paper highlights the importance of equity and the ethical design and implementation of barriers in family medicine. While providing a comprehensive review of AI in family medicine and pinpointing areas for future growth, it also acknowledges certain limitations. Most studies referenced are from controlled or high-resource settings, which limits their applicability to other environments, especially those with limited resources. Furthermore, as technology rapidly advances, some literature may already be outdated. Other limitations include potential publication bias, which may lead to the inclusion of only favorable results and higher AI usage settings. Further challenges relate to real-world application, as validation of these tools in family medicine practices has not been extensively conducted. Several considerations for future research are crucial. Scholars should evaluate how AI influences the clinician-patient relationship, mitigate biases, enhance data collection for better algorithms, promote data governance for secure information sharing, improve system architecture, and analyze regulatory frameworks to ensure equitable distribution and use. Collaborating closely with family physicians should be a priority for developing and refining AI tools in primary care. It is also essential to assess outcomes of real-world AI use, particularly regarding healthcare costs and its role in reducing health disparities. Engaging in this process has significantly expanded my understanding of increasingly prevalent technology. The benefits—such as improved patient and provider experiences and potential cost savings—are substantial. Therefore, prioritizing the implementation and ongoing improvement of AI tools should be a priority. AI should not only be adopted but also embraced. By critically reviewing existing literature and following the recommendations outlined here, AI can further advance family medicine and facilitate patient-centered care.

1. Definition and types of AI in healthcare. 2023;1–8.
2. Wojda T, Hoffman C, Jackson J, Conti T, Maier J. AI in healthcare: Implications for family medicine and primary care. In: Stawicki SP, editor. Artificial intelligence in medicine and surgery - An exploration of current trends, potential opportunities, and evolving threats - Volume 1. London: IntechOpen; 2023;1–17. Available from: https://doi.org/10.5772/intechopen.111498
3. Elhaddad M, Hamam S. AI-driven clinical decision support systems: An ongoing pursuit of potential. Cureus. 2024;16(4):e57728. Available from: https://hcdexchange.org/wp-content/uploads/2024/05/20240406-13366-1bzxzru.pdf
4. Hassan M, Kushniruk A, Borycki E. Barriers to and facilitators of artificial intelligence adoption in health care: Scoping review. JMIR Hum Factors. 2024;11:e48633. Available from: https://doi.org/10.2196/48633
5. Baily MN, Kane AT. AI in the health care sector. Washington (DC): Brookings Institution; 2025. Available from: https://www.brookings.edu/wp-content/uploads/2025/04/20250401_CRM_BailyKane_AICaseStudies_HealthCare_FINAL.pdf
6. Razai MS, Al-bedaery R, Bowen L, Yahia R, Chandrasekaran L, Oakeshott P. Implementation challenges of artificial intelligence (AI) in primary care: Perspectives of general practitioners in London UK. PLoS One. 2024;19(11):e0314196. Available from: https://doi.org/10.1371/journal.pone.0314196
7. Jones OT, Calanzani N, Saji S, Duffy SW, Emery J, Hamilton W, et al. Artificial intelligence techniques that may be applied to primary care data to facilitate earlier diagnosis of cancer: Systematic review. J Med Internet Res. 2021;23(3):e23483. Available from: https://doi.org/10.2196/23483
8. Amann J, Blasimme A, Vayena E, Frey D, Madai VI. Explainability for artificial intelligence in healthcare: A multidisciplinary perspective. BMC Med Inform Decis Mak. 2020;20:310. Available from: https://doi.org/10.1186/s12911-020-01332-6
9. Richens JG, Lee CM, Johri S. Improving the accuracy of medical diagnosis with causal machine learning. Nat Commun. 2020;11:3923. Available from: https://doi.org/10.1038/s41467-020-17419-7
10. Buçinca Z, Malaya MB, Gajos KZ. To trust or to think: Cognitive forcing functions can reduce overreliance on AI in AI-assisted decision-making. Proc ACM Hum Comput Interact. 2021;5(CSCW1):Article 188. Available from: https://doi.org/10.1145/3449287
11. Moldovan AMN, Vescan A, Grosan C. Healthcare bias in AI: A systematic literature review. In: Proceedings of the 20th International Conference on Evaluation of Novel Approaches to Software Engineering (ENASE 2025). 2025. p. 835–42. Available from: https://doi.org/10.5220/0013480300003928
12. Katsakiori PF, Kagadis GC, Mulita F, Marangos M. Implementing artificial intelligence in family medicine: Challenges and limitations. Cureus. 2024;16(12):e75518. Available from: https://doi.org/10.7759/cureus.75518
13. Wang L, Zhang Z, Wang D, Cao W, Zhou X, Zhang P, et al. Human-centered design and evaluation of AI-empowered clinical decision support systems: A systematic review. Front Comput Sci. 2023;5:1–15. Available from: https://www.frontiersin.org/journals/computer-science/articles/10.3389/fcomp.2023.1187299/full
14. Lin SY, Mahoney MR, Sinsky CA. Ten ways artificial intelligence will transform primary care. J Gen Intern Med. 2019;34(8):1626–30. Available from: https://link.springer.com/content/pdf/10.1007/s11606-019-05035-1.pdf
15. Xu RZ, Bone JN, Courtemanche R, Yefet LS, Simmonds MC, Cattoni E, et al. Prescription factors contributing to new long-term opioid use in British Columbia between 2013 and 2017. BC Med J. 2025;67(2):54–63. Available from: https://bcmj.org/sites/default/files/BCMJ_Vol67_No2_complete.pdf#page=10
16. Almagadi MAS, Hazazi AJA, Shabi LFA, Alzhrani FY, Barnawi KBO, Alymani RN, et al. The impact of artificial intelligence on administrative productivity in healthcare: A systematic review of operational efficiency and burden reduction. Bioscan. 2025;20(1):722–6.
17. Ogunsakin OL, Anwansedo S. Leveraging AI for healthcare administration: Streamlining operations and reducing costs. IRE J. 2024;7(10):235–44. Available from: https://www.irejournals.com/formatedpaper/1705709.pdf
18. Nishiyama K, Mikata Health Inc. Prioritizing patient care & managing physician burnout using artificial intelligence. Alberta Innovates. 2024. Available from: https://albertainnovates.ca/wp-content/uploads/2024/09/AI-Health_onepager_Mikata_Health_Nishiyama_web1.pdf
19. Farhud DD, Zokaei S. Ethical issues of artificial intelligence in medicine and healthcare. Iran J Public Health. 2021;50(11):i–v. Available from: https://doi.org/10.18502/ijph.v50i11.7600
20. Naik N, Hameed BMZ, Shetty DK, Swain D, Shah M, Paul R, et al. Legal and ethical consideration in artificial intelligence in healthcare: Who takes responsibility? Front Surg. 2022;9:1–187. Available from: https://doi.org/10.3389/fsurg.2022.862322
21. Nazer LH, Zatarah R, Waldrip S, Ke JXC, Moukheiber M, Khanna AK, et al. Bias in artificial intelligence algorithms and recommendations for mitigation. PLoS Digit Health. 2023;2(6):e0000278. Available from: https://doi.org/10.1371/journal.pdig.0000278
22. Ratwani RM, Sutton K, Galarraga JE. Addressing AI algorithmic bias in health care. JAMA. 2024;332(13):1051–2. Available from: https://doi.org/10.1001/jama.2024.13486
23. Chen Y, Clayton EW, Novak LL, Anders S, Malin B. Human-centered design to address biases in artificial intelligence. J Med Internet Res. 2023;25:e43251. Available from: https://doi.org/10.2196/43251
24. Adegbesan A, Akingbola A, Ojo O, Jessica OU, Alao UH, Shagaya U, et al. Ethical challenges in the integration of artificial intelligence in palliative care. J Med Surg Public Health. 2024;4:100158. Available from: https://doi.org/10.1016/j.glmedi.2024.100158
25. Du S, Xie C. Paradoxes of artificial intelligence in consumer markets: Ethical challenges and opportunities. J Bus Res. 2021;129:961–74. Available from: https://doi.org/10.1016/j.jbusres.2020.08.024
26. Al Kuwaiti A, Nazer K, Al-Reedy A, Al-Shehri S, Al-Muhanna A, Subbarayalu AV, et al. A review of the role of artificial intelligence in healthcare. J Pers Med. 2023;13(6):951. Available from: https://doi.org/10.3390/jpm13060951