The AI Moonshot: What We Need and What We Do Not

• artificial intelligence
• family medicine
• EHR

Earlier this year, Annals of Family Medicine sponsored a workshop at the International Conference of Artificial Intelligence in Medicine (AIME24) in beautiful Salt Lake City, Utah. Dr Caroline Richardson, MD, Editor of Annals of Family Medicine, started the workshop with a thought-provoking presentation on artificial intelligence (AI) and its possible use in primary care.

To set the stage, she told how family medicine has become nearly impossible as a job, with excessive time spent on electronic health records (EHR). Family physicians use that time documenting visits and managing messages, but she presented a third reason for this: Noise. Dr Richardson gave the example of seeking the results of an electrocardiogram (EKG) in the medical record, finding 5 entries for that result, but after clicking on all 5, there was no result. When patients request medications, the EHR notifies a physician. If the physician doesn’t immediately respond, the EHR may notify another physician, causing redundancy, confusion, and necessary calls to the pharmacy to ensure the patient gets the correct medication. These time-consuming tasks are an artifact of asynchronous communication and multiple eyes on the same record. They create work but do not reduce the time needed for other tasks.

Family physicians do so much more than what is in the medical record. We create and sustain meaningful relationships with patients, decide with them how our interventions (medications and others) can best serve the patient, and help them navigate a complex and poorly designed health system based on reimbursement instead of health. We work with patients to provide solutions that align with their values and help them navigate the challenging and false dichotomy between physical and mental health.

Family physicians (and our patients) would be better served if AI could write our notes for us, perhaps listening to our patients when they speak in a language other than English yet recording English words in the EHR. Artificial intelligence could even organize it into a Subjective, Objective, Assessment, and Plan (SOAP) note, allowing physicians to talk to the patient instead of typing. Artificial intelligence scribes can make explicit the often implicit and complex and multifactorial (socio-ethical-familial) care support system choices of investigation plans and treatment plans. AI could possibly sort out the confusing and, at times, contradictory recommendations from specialists. Deploying AI to organize the record into something useful for our patients would create time for relationship building and providing solutions. If we are bold, AI could facilitate communication, reduce redundancies, and develop coherent collaboration within teams. There are many other things that AI can do for us, as presented in Table 1.

Artificial intelligence has enormous potential to make our lives easier in very many ways. And yet, AI is currently creating things that we do not need. Family physicians do not need another risk calculator, and we do not need more diagnostic assistance. This is our moonshot moment: if we articulate well what we need and what we do not, this tool has the potential to restore joy to practice. Let’s articulate our vision of what we want from AI and make it into something that serves us. With the right AI, more of our brain space will be freed up to do what family physicians do best: take care of people.

• Received for publication December 5, 2024.
• Accepted for publication December 5, 2024.

• © 2025 Annals of Family Medicine, Inc.