By Rick Valachovic, DMD, MPH, Clinical Professor and Director of the NYU Dentistry Center for Oral Health Policy and Management

 

FDA announces sweeping changes to oversight of wearables, AI-enabled device

The changes could allow unregulated generative artificial intelligence tools into clinical workflows

 

Utah and Doctronic Announce Groundbreaking Partnership for AI Prescription Medication Renewals

Utah becomes the first state to safely evaluate autonomous AI for prescription renewals for chronic conditions

 

Hospitals Are a Proving Ground for What AI Can Do, and What It Can’t

Healthcare is going all-in on artificial intelligence, from reading patient scans to fighting insurance denials

 

Headlines from STAT’s AI Prognosis newsletter and the Wall Street Journal (WSJ)

 

These are just three of the headlines that caught my eye at the start of this month, and dozens more have appeared since — in my news feeds alone! 

Less than two years have passed since I first wrote about ChatGPT, the large language model (LLM) that took the world by storm following its release in November 2022. At the time, I noted that “the pace of change is accelerating at an unprecedented rate.” Little did I know. Keeping up with everything written about LLMs and other forms of artificial intelligence (AI)  has become a Herculean task. Grasping all of the ways these technologies are transforming our lives…impossible. 

Nevertheless, we can draw some insights from the ocean of information flooding the popular press, trade publications, and professional journals. Here are a few takeaways from my reading that may prove helpful to other dental educators and health professionals as we navigate the AI-infused reality that lies ahead.

 

1. Pilot projects are demonstrating AI’s ability to improve diagnosis and clinical decision making, but delivery of these technologies at scale appears a long way off.

Imagine…

• An AI tool that analyzes coronary plaque in CT scans to determine a patient’s risk of heart attack or stroke without undergoing an invasive angiogram procedure.
• An AI-enabled wearable device that gently wakes patients with atopic dermatitis when they begin scratching in their sleep. 
• A new class of algorithms that could soon turn the screening mammogram into a tool for predicting an individual’s risk for cardiovascular disease as well as breast cancer.

These tools already exist, as do predictive algorithms that analyze the vital signs and charts of hospitalized patients each hour to determine their risk of deterioration. Kaiser Permanente employed this AI tool in twenty-one hospitals and reports it saved five hundred lives over the course of a single year. I also heard about an equally impressive use of AI at my old stomping grounds, Boston Children’s Hospital. Genetics researchers at the hospital’s Manton Center for Orphan Disease Research worked with OpenAI to create a bespoke ChatGPT model that had diagnosed rare diseases in 15 children as of last October. 

These cases demonstrate that clinical applications of AI have come a long way and have tremendous promise. That said, it’s notable that each example constitutes a discrete application of the technology on a single condition or treatment. None offers a solution with the (admittedly failed) ambition of IBM’s Watson supercomputer or the tricorder from Star Trek — tools that aimed to revolutionize the practice of medicine, and by extension dentistry. Such giant leaps may be far off, but intermediate steps are likely just a matter of time. The federal Advanced Research Projects Agency for Health (ARPA-H) launched an effort this month to create AI agents that would serve as “clinician extenders” capable of autonomously providing 24/7 cardiovascular care to millions of Americans.

A December 2025 report from KLAS Research affirms that most health care organizations have embraced AI but primarily for administrative, not clinical, tasks and not at scale. The reasons are manifold — patient-safety and liability concerns, a lack of governing rules, and uncertainty about their return on investment. If you have time for a deeper dive into this topic, I recommend a recent European Commission report on the deployment of AI in health care. It discusses these and other implementation barriers and proposes steps that could enable “the sustainable integration” of AI into health systems. 

 

2. Dental use cases are fewer, but more are surely coming. 

Dental applications of AI are fewer and less diverse, but they do exist — at least according to Google’s LLM, AI Mode, which emerged last month as the top tool for providing up-to-date information in a chatbot battle staged by The Washington Post. That query pulled up a 2025 scoping review in Bioengineering, which noted the following use cases:

• Radiographic diagnostics, including reading cone-beam computed tomography (CBCT) scans, intraoral photographs, and radiographs. And identifying periapical lesions on 2D CBCT slices.
• Optimizing restoration design and implant positioning
• Predicting the risk of caries, periodontal disease, and cancer in order to personalize care planning 

The authors’ conclusion? “AI is revolutionizing dentistry with enhanced diagnostic accuracy, predictive planning, and efficient administration automation.”

Regarding this last item, industry sources suggest that dental practices are adopting AI tools for administration and revenue cycle management, with dental service organizations leading the way. If a 2025 report on digital dentistry from the United Kingdom’s National Health Service is any guide, the integration of these tools could significantly increase efficiency. “Practices utilizing AI-powered management systems report average efficiency improvements of 27%, allowing reallocation of staff time from administrative tasks to patient care,” the report states.

 

3. Reducing the burden of documentation is AI’s biggest impact on health care to date. 

As you may have noticed during a recent appointment of your own, ambient scribes — AI tools that capture conversations between clinicians and patients and turn those recorded encounters into clinical notes — are becoming a fixture of many practices. Health care organizations have eagerly adopted these tools in the hope of reducing clinician burnout linked to the demands of digital documentation. 

Do the tools work as advertised? A study of clinicians in seven health systems reported a significant reduction in burnout and cognitive load after a month working with an AI scribe, suggesting their use can reduce administrative burden and improve clinicians’ wellbeing. Yet writing in STAT, Mario Aguilar and Brittany Trang report on studies that show the time actually saved by these tools varies considerably and does not necessarily align with clinicians’ perceptions. Additionally, one scoping review found that errors remain an issue even as speech recognition improves, with errors of omission being especially problematic. A separate study comparing the accuracy of five unnamed AI scribes found that all five tools regularly produced AI-generated notes with errors deemed capable of producing moderate-to-severe harm, raising significant concerns about patient safety.  

Despite these limitations, I suspect health systems will continue to invest in these tools. STAT reports that companies marketing AI scribes have begun raising capital and wooing customers on the promise that AI-assisted clinical documentation will unearth more billable conditions and increase revenue for adopters. The WSJ reports about 1,000 hospitals are already using Epic’s generative-AI tool to mine patient records and draft insurance appeals. An executive at Northwestern Medicine told the paper that staff now spend about 23% less time processing each denied claim. A similar effort is paying off at New York’s Mount Sinai, which gained $12 million in revenue by using AI to successfully overturn insurance denials. Who will pick up the extra cost to the health sector writ large remains unclear, but is it any wonder that health care organizations are purchasing commercial AI licenses at more than twice the rate of other U.S. businesses?

Next month I’ll Iook at how AI might influence health professions education and scientific research and grapple with its destructive potential. Stay tuned!