Administrative burden is widespread in family medicine and primary care, and is a major issue for practitioners,1-3 and a contributing factor for burnout.4 Certain technology-based tools, such as electronic medical records (EMRs), have been viewed in the past as a potential support for primary care practitioners (PCPs), yet evidence shows their use can increase administrative burden.5-7 Why has this happened? Although the causes are multifaceted and include the nature of the technology itself, as well as sociotechnical factors influencing uptake and use,8-10 we suggest 2 additional key reasons: First, practitioners have not had much opportunity to engage in designing or creating these tools, and second, many of the technocentric solutions that have been applied are focused on efficiency rather than practitioner and patient experience. These points are related—the lack of input from PCPs often results in a lack of appreciation of, and accounting for, the time that is required for efficacy-based encounters versus those that are efficiency-driven. Yet, to clinicians, efficacy has a higher value than efficiency.11
Another wave of change associated with technology is afoot, as artificial intelligence (AI) tools are becoming more pervasive, and those designed for routine task support, such as AI scribes (which can record, transcribe, and summarize a patient visit; and, depending on the individual tool’s functionality, extract content and apply it to the EMR) are being rapidly taken up in practice. With an estimated 44% of administrative tasks that could be automated in primary care practice,12 there is great potential for AI tools to relieve administrative burden. The emerging evidence is promising; for example, the implementation of an AI-enabled tool to support documentation in primary care resulted in an almost 30% reduction in documentation time per encounter and a reduction of documentation time outside of regular hours by approximately 11%.13 Most recently, an evaluation of the use of AI scribes in Ontario found that practitioners reported a reduction of 3 hours per week in completing administrative tasks when using these tools.14 The use of AI scribes is also associated with experiences of reduced administrative burden and stress or burnout.13,14
Additional examples of AI tools include AI administrative task assistants, which can perform routine billing, scheduling, and communication tasks, such as patient appointment reminders. Evidence about the impact of using this type of AI tool is limited, yet practitioners see its potential for reducing administrative task burden.14 However, many barriers and concerns exist about the use of AI tools more broadly, including those associated with AI technology (eg, potential for algorithmic bias, explainability),15,16 implementation and use (eg, impact on patient-practitioner relationship, patient safety, need for integration into EMRs, medicolegal concerns, costs),14-16 and lack of evidence to support the use of these tools.
The implications of these issues for AI administrative task tools are varied. How does the information resulting from an AI scribe recording the words uttered during the encounter differ from what is recorded by a practitioner who also observes and conveys nonverbal cues? How might the scribe interpret a patient’s accent? What might be the implications of the AI scribe recording sensitive information and the data being either stored or erased after an encounter? It is possible these and other problems could undermine overall health care quality and outcomes.
Stepwise approach to adoption of AI in the office
Some of the issues with AI will be solved through evolving processes; for example, the College of Family Physicians of Canada’s statement on AI for family medicine provides direction to support their resolution.17 Nevertheless, some concerns will remain and practitioners will need to weigh the potential benefits against the risk of using AI tools in practice. Although guidance and toolkits exist regarding AI implementation in health care,18-20 they are heavily focused on clinical tools and on hospitals or large organizations. Hence, many PCPs seeking guidance regarding the adoption of AI tools are left with guidelines that focus on analyzing research literature21-23 or understanding the proper use of an already procured tool.24-26
Herein we outline a stepwise approach on key factors practitioners should consider in the adoption of administrative AI tools in the primary care setting. The considerations and resources presented in Table 114,27-30 and discussed below are intended to support PCPs in independently determining what AI administrative task tools will work best for their practices.
Considerations and resources to support adoption of AI administrative task tools in primary care practice
Once a practitioner has conducted a preliminary assessment and decided to move ahead with the use of AI tools for administrative task support, we recommend the following be considered before adopting these types of tools.
Relevance and alignment in primary care. Are the AI administrative task tools being considered for use relevant to, and specifically designed for, the primary care setting? More specifically, are the developers of these AI tools attuned to the needs of primary care practitioners? Have PCPs been involved in, or led, the design or development of these tools?
The usefulness of AI tools developed to relieve administrative burden is highly dependent on the end user (ie, the practitioner) having been consulted or involved in, or having led the design and development of these tools. Currently, there is a lack of practitioner engagement in the development and assessment of these tools,31 with few examples of this type of co-design in existence.32,33 Two key approaches might encourage AI developers to collaborate with primary care practitioners. Legislation can mandate such collaboration. Alternatively, PCPs can purchase AI products based on recommendations from independent assurance laboratories or practitioner-oriented organizations. Table 2 summarizes examples of these options and their attendant elements.34,35 Beyond relevance, what about alignment? Is there a match between what is needed in clinical practice and what these tools can deliver?
Key approaches to encourage AI developer-vendor collaboration with PCPs
AI tools can be misaligned if they fail to reflect the values and needs of the end users.36 This problem of alignment is perhaps even more critical in the clinical setting because patient care is at stake. It is important for practitioners to be critical of not just the evidence around AI tools, but also the level of alignment between what is needed in clinical practice and the actual capability or functioning of the tools.
Evidence. What is the evidence that using AI tools reduces administrative burden in the primary care setting? Practitioners must demand both proper real-world testing and rigorous evaluation of AI administrative task tools. These evaluations need to be completed by vendors and also by independent organizations and researchers in academic institutions to ensure that user preferences are embedded in the design of AI tools and that relevant outcomes are assessed. Recent examples of studies evaluating AI administrative task tools represent promising and important developments,14,27 but the evidence base is still limited.31
Patient-centred care. What impact might AI administrative task tools have on the delivery of patient-centred care? The principles of patient-centred care need to remain in place regardless of the mode of care delivery.37 The use of technology can impede practitioners in many ways, so caution must be exercised so technology does not get in the way of the clinician’s judgment and intuition when caring for patients.38 Care must be taken to ensure that the development and implementation of AI tools do not contribute to the distancing of patients and practitioners. This distancing can be either physical or a failure of empathy, as the use of these technologies demands time and attention during the clinical encounter.39
As the development of AI tools and subsequent use cases in primary care are rapidly evolving, and as the evidence base continues to grow, the practitioner might find convincing and informative answers to the above questions and decide in favour of adoption.
We then recommend attending to the following additional considerations outlined below.
Implementation. What is known about how to best implement AI administrative task tools? Preliminary evidence has identified factors that can impede the uptake and use of AI tools, including costs, integration with existing technology, workflow issues, technical difficulties, and need for training.14,27 These factors are commonly encountered in the implementation of technology-based innovations in health care; accordingly, it is important to know what works for whom and in what circumstances. Ideally, practitioners should be able to experience these tools in test mode or sandbox training to both determine real-world performance and assess their own digital literacy and technical support needs.
Readiness. What is the practitioner’s or the organization’s level of readiness for the adoption of AI administrative task tools? Given the complex nature of technology implementation and use in health care, it is important to pay attention to readiness or preparedness for the adoption of AI tools. Although not specific to primary care, readiness for uptake of AI in health care at individual40 and organizational28 levels has been explored.
After the adoption and implementation of AI administrative task tools, we recommend evaluative considerations.
Evaluation. Do the tools work as intended in practice? Is their use worth the time, expense, and effort in terms of benefits? Were the anticipated outcomes achieved? Would continued use of these tools be beneficial, and what resources might be necessary in the long-term to support this use? Emerging research offers guidance on how AI tools that are specific to relieving administrative burden can be assessed and evaluated.14,27
Conclusion
The current crisis in primary care is being compounded by administrative burden, which might be partly remedied by the use of AI tools for administrative tasks. However, the successful implementation of these tools will depend on practitioners being engaged in helping design these tools and acquiring additional skills and knowledge to use them. While these AI tools might reduce administrative burden and save time, how can that returned time be best spent? In a system demanding increased efficiency versus efficacy, there is a danger that the addition of these tools will diminish patient-centred care as emphasis is placed on the implementation of technology. The stepwise considerations outlined in this commentary highlight areas for PCPs to assess when adopting AI tools in practice. These considerations might also prompt further discussion and research to improve guidance for practitioners in preparing for the transformation of primary care via the introduction of disruptive technologies.
Footnotes
Competing interests
Dr Keith Thompson has financial gains to declare as Chief Medical Officer for NuraLogix Corporation, research funding from the Institute for Earth and Space Exploration at Western University, and honoraria from OntarioMD and various pharmaceutical companies. Dr Sian Tsuei is a member of the Artificial Intelligence Advisory Group of the College of Family Physicians of Canada. No contribution or support was provided by any of these organizations.
The opinions expressed in this article are those of the authors. Publication does not imply endorsement by the College of Family Physicians of Canada.
This article has been peer reviewed.
Cet article a fait l’objet d’une révision par des pairs.
- Copyright © 2025 the College of Family Physicians of Canada
