Artificial intelligence (AI) has swiftly shifted from futuristic hype to everyday reality. Family physicians are uniquely positioned to benefit from this shift because our work requires managing broad medical knowledge as well as time-consuming administrative tasks, all while nurturing the doctor-patient relationship. AI isn't here to sideline us; it's here to support us, like a reliable assistant who handles some of the work, freeing us to focus on the human elements that drew us into this field.¹

This article describes what family physicians should consider when assembling their AI toolkit. I've curated this list through hands-on use in busy clinics and teaching rounds, and I've tested these tools in real-world scenarios, prioritizing those that are free or low-cost, user-friendly, and adaptable. Of course, the AI landscape evolves faster than influenza strains in winter. New tools pop up weekly, and what works today might be outdated tomorrow. This list of AI tools isn't exhaustive, but it's a solid starting point for family physicians looking to dip their toes in the AI waters without drowning in options.

To make this toolkit actionable, I've organized it around five key categories and use cases:

1. Clinical knowledge assistants, which excel at rapid, evidence-based information retrieval — such as pulling up the latest clinical guidelines mid-visit,

2. Decision support tools, which go further, aiding in diagnostic reasoning, differentials, and management plans by synthesizing patient data into structured insights,

3. Documentation support, which can save you time,

4. Administrative support, which can save you frustration,

5. Learning support, which can help you stay current amid information overload.

1. CLINICAL KNOWLEDGE ASSISTANTS

Case: A 45-year-old patient with type 2 diabetes asks about the newest glucagon-like peptide-1 (GLP-1) agonist options during a routine follow-up visit, and you need to quickly access up-to-date, citable evidence without derailing the visit.

The key message to remember across all categories is that while AI augments our expertise, the buck still stops with us. Always verify outputs, especially for clinical or high-stakes decisions.

In moments like this, when a new medication is the subject of TV commercials before CME courses or when memory recall fails or guidelines shift, clinical knowledge assistants act as your on-demand librarian. These tools are experts at fetching reliable information from trusted sources in seconds. OpenEvidence stands out here, purpose-built for clinicians by querying PubMed, Food and Drug Administration labels, and peer-reviewed literature (including the New England Journal of Medicine, JAMA Network, and most recently American Family Physician and FPM) to deliver concise, evidence-based summaries with citations.² It's free with professional verification, awards American Medical Association PRA Category 1 CME credits per query, and is HIPAA-compliant for organizations that enter into a business associate agreement with the company. In the case above, querying OpenEvidence for “current GLP-1 agonists for type 2 diabetes, including cost and efficacy comparisons” would yield a synthesized response with links to trials such as Semaglutide Treatment Effect in People with Obesity (STEP) and Semaglutide Unabated Sustainability in Treatment of Type 2 Diabetes (SUSTAIN). This not only informs the patient conversation but models evidence-based practice for learners, which is important in residency practice settings. I've used OpenEvidence on rounds to confirm heart failure therapies, turning potential “I'll look it up later” conversations into teachable moments on the spot.

A tool that pairs nicely with OpenEvidence is Perplexity, which offers a broader, real-time search and answer engine, ideal for nonclinical scenarios such as state-specific prescribing rules or emerging trends. It's not limited to medical resources, so double-check sources, but it's handy for gray areas overlapping with things outside of medicine — say, integrating wearable data into diabetes management.

Tip: Start with OpenEvidence for pure clinical queries to avoid web noise. It is easy to use. Then layer in other tools for contextual breadth. Together, they can save precious minutes during visits, letting you pivot back to the patient with confidence.

2. DECISION SUPPORT TOOLS

Case: A 32-year-old woman presents with persistent fatigue, intermittent rash, and joint pain, and you're building a differential diagnosis without immediate specialist input.

Decision support tools go beyond information retrieval and offer active reasoning, helping craft differentials and plans while encouraging thoroughness. Glass Health AI is a gem in this area. You can input symptoms, history, and labs, and it generates a structured differential with rationales, often flagging overlooked rarities. In the above case, it might highlight autoimmune conditions such as lupus or Sjögren's, prompting targeted tests. I love using it in residency clinics to spark discussions (e.g., “Why prioritize this over that?”) and to foster critical thinking without spoon-feeding answers. It's subscription-based but offers trials, and its transparency builds trust.

Doximity GPT (which recently acquired Pathway) takes a guideline-centric approach, answering queries like “Algorithm for incidental pulmonary nodule in a smoker?” with step-by-step summaries and CME credits. It's HIPAA-compliant and free for verified users, making it seamless for quick consultations. For instance, I've queried it on nuanced lipid management following advanced panels, ensuring my care plans align with American Heart Association/American College of Cardiology (AHA/ACC) consensus.

Other tools such as VisualDx shine for visual diagnostics. For example, you can upload a photo of a patient's rash, and the tool compares the image against a vast library, which is great for dermatology-heavy family medicine practice. You could use any large language model (LLM) AI program to do this, but most of them would first ask you a series of tailored questions that help with diagnosing a lesion or rash. Another tool, Isabel, now integrated with DynaMed, lists differentials from symptoms, adding depth for complex cases as well.

Tip: Use these tools early in ambiguous cases to broaden your net, but edit queries and outputs to fit the patient's unique story. When used properly, decision support tools can help reduce the cognitive load, allowing us to focus more on the “why” behind decisions.

3. DOCUMENTATION SUPPORT

Case: After a hectic day of back-to-back visits, you're staring at a pile of unfinished charts, dreading “pajama time” at home.

Documentation burden leading to physician burnout is real. It's the thief that steals our evenings. AI scribes tackle this by ambiently transcribing visits and drafting notes, letting you stay present with patients. AI scribe options abound, including Doximity Scribe (currently free for verified clinicians), Nabla Copilot, DAX, and Heidi.ai. Some vendors offer freemium models (i.e., basic services are free, but advanced features require a fee), while others charge hefty prices per user or per encounter, so do your homework on pricing. You'll also want to choose a product that can integrate with your EHR, if possible, for seamless workflows. I foresee a continued race to better pricing with time, until companies can roll out advanced features like auto-pended orders integrated with your EHR.

To start using an AI scribe,³ make sure the tool is approved by your organization if you are employed. Then, simply obtain patient consent, record the encounter using the AI scribe app on your smartphone (audio typically isn't stored, for HIPAA compliance reasons), and voila — a SOAP note appears in seconds, often 90% complete with billing codes suggested. If you haven't tried using an AI scribe yet, you should. Most vendors offer free trials on their websites, where you can use your voice to give a complicated patient history and see how the tool summarizes the history of present illness.

Studies have shown that ambient AI scribes have reduced documentation time, including after-hours charting,⁴ with a 21% drop in burnout rates at large systems such as Mass General Brigham.⁵ In my clinic, trialing Dax during clinic meant more eye contact and richer histories, because patients opened up more when I wasn't typing furiously.

Of course, you will need to verify and edit AI-produced notes for accuracy, especially in multi-speaker or complex cases. I find that social histories often need more context, which I add manually. AI-produced social histories tend to leave out family, travel, pet, and hobby updates, which I value greatly for continuity.

Tip: Start using an AI scribe with simple acute visits to build your comfort, and then scale. This isn't magic, but it's close, and reclaiming time at home is priceless.

4. ADMINISTRATIVE SUPPORT

Case: A patient's insurer denies coverage for a necessary mobility aid, and you need to write a compelling prior authorization appeal without reinventing the wheel.

Administrative tasks are the bane of primary care, consuming hours on forms, letters, and faxes. (Side note: Why do we still fax in 2025?) AI drafting tools lighten the load by generating templates you can easily tweak. One analysis estimated that AI could eventually automate 30% of family physicians' tasks,⁶ potentially enhancing patient access.

Doximity GPT excels here, with HIPAA-compliant prompts for appeals, disability notes, or patient instructions — and, yes, you fax directly from the app. To address the case above involving a prior authorization appeal, you would simply input key clinical details (anonymized if needed), and the tool would craft a criteria-citing letter in seconds, saving 15–20 minutes per task. While AI won't argue with insurers' decisions (yet), it does help us communicate a strong stance with minimal mental effort.

General LLMs such as ChatGPT, Gemini, or Grok can handle generic drafts, but you should avoid inputting a patient's protected health information into these general tools. I've used them for school excuse notes, referral letters, requests for work, patient education, and more, multiplying my time savings across these examples.

Tip: Customize the tone of AI outputs for your intended audience by providing prompts like “keep it empathetic and professional.” This can produce a higher quality response.

5. LEARNING SUPPORT

Case: You are prepping a presentation on long COVID while juggling clinic duties, and you need efficient literature synthesis without endless PubMed searching.

Staying current on the medical literature amid information overload is a perpetual challenge, but AI tools such as Elicit act as a research assistant. This resource searches millions of academic papers and clinical trials, with new data sources being added continuously. Based on your query (e.g., “What are the signs, symptoms, and evidence-based treatment options for long COVID?”), these tools can query studies, summarize findings, and extract relevant data. In the above use case, Elicit could help you summarize key papers on post-viral syndromes in hours, not days. It's free for the basic version, with fees for upgrades. Here again, LLMs such as ChatGPT, Gemini, or Grok can also be helpful for research and analysis, but these general tools gather information from diverse sources and have a higher risk of generating misinformation.

Tip: Even if you aren't doing research for presentation purposes, these AI tools can support personal learning and make it more evidence based. Start with the free version and begin experimenting; you can always upgrade later to access more features if needed.

ALL THAT IS GOLD DOES NOT GLITTER

J.R.R. Tolkien wrote in The Fellowship of the Ring, “All that is gold does not glitter.” In other words, while AI may be glittering brightly right now, the old-fashioned relationships we have with our patients and their families are still golden. Any AI tools we adopt should support us in caring for the people who need us.

In family medicine, where time is our scarcest resource, AI tools can help us reclaim our time without compromising care. In my dual role as clinician and educator, I've seen AI enhance decisions, cut burnout, and enrich teaching. Patients get evidence-based care faster, residents learn the “why,” and physicians reclaim their evenings. The key is to start small: Identify your pain point, pick one tool, pilot it, and assess its impact on your workflows and patients.

AI is no panacea, but thoughtfully integrated, it can amplify our art.