Implementing AI in the Physician’s Office

WHEN ENTERING “Top Benefits of AI in Physician’s Offices” into ChatGPT, the response highlights the increasingly rapid growth of this burgeoning technology. From expanded access to patient care, faster and more efficient office workflows, streamlining automated tasks and the creation of personalized care plans, not to mention its potential for diagnostic support, artificial intelligence (AI) in physician offices is here to stay.

The Benefits of AI

Experts have written extensively about AI’s potential in diagnostics, but many find AI’s ability to improve administrative efficiency offers a fantastic initial application for this powerful technology. Enhanced capabilities take the place of time-consuming administrative tasks and enable a greater focus on one-to-one patient care.

In fact, a 2024 American Medical Association (AMA) survey showed nearly two-thirds of physicians reported using AI in their practices, a 78 percent increase from the year prior. Documentation of billing codes, creating discharge instructions, translation services and transcription and charting are among the most common uses. More than half of respondents said the reduction in administrative burdens was the biggest opportunity for AI in their practices.¹

Concerns Remain

Concerns for this technology as barriers for entry remain, however, including data privacy, poor integration with electronic health records, dispensing incorrect information, liability and employee training requirements.

The design of AI tools such as ChatGPT lack the robust privacy requirements of platforms designed for healthcare information. Furthermore, because AI uses historical datasets for training, these tools may incorporate erroneous information or biases into their algorithms. They may also exclude current clinical recommendations or new findings. In one well-documented example, developers shut down a chatbot designed to support the management of eating disorders when it dispensed diet and exercise tips to users. The weight loss information dispensed by the chatbot was reportedly not part of the data on which the tool was meant to be trained.2 This cautionary tale highlights the importance of human oversight with the use of any technology.

The learning curve for AI’s use may be steep. After all, medical schools only recently introduced AI into their curricula. In some cases, staff may also be uninterested or uncomfortable learning how to incorporate AI, concerned it will take over their jobs. The time required to effectively use AI can also be cumbersome, taking away from important daily duties. That being said, many with higher levels of data literacy and health informatics are jumping at the opportunity to incorporate AI methods so they can take advantage of the ability to streamline tasks.

AI’s Entry Point: Chatbots

Incorporating chatbots into office functionality can be an easy entry into the world of AI. Chatbots are computer programs or software that simulate human-like conversations. Used by anyone with an Internet-enabled computer or smartphone, 10 percent of AMA survey respondents reported adding patient-facing chatbots to their medical practices.

From a patient standpoint, chatbots can act like virtual assistants and enable scheduling appointments, link those to the patient’s own digital calendar, and send prescription and refill reminders. Chatbots help patients triage their symptoms through text chats to determine whether a condition warrants medical attention and, if so, where the nearest treatment facility is located. Many patients find chatbots easier to navigate than their own patient portals or telehealth appointments. That makes them an attractive proposition from a provider standpoint. The ability for a bot to respond to patient administrative needs, with the patient’s consent, saves time and lowers costs, allowing staff to focus on responsibilities that require problem-solving, emotional intelligence, interpersonal skills and judgment.

Chatbots have a relatively low cost of entry for medical practices, with fees ranging from approximately $150 to $400 monthly. Custom-designed chatbots can cost tens of thousands to upwards of $100,000. The market for chatbots continues to grow, with estimates reaching $1.2 billion by the year 2032.²

Despite chatbot’s growing success, many caution that more research is required before they are standardized in medical practices, particularly given the concerns for data privacy. Although the level of required privacy between chatbot hosts and their users is dependent upon the types of conversations held, providers must be vigilant in choosing chatbot software and hosts whose security practices align with their own accepted data integrity practices. Similarly, providers should inform patients about their data privacy, explaining data collection and storage, and whether responses can be tracked and linked to other online data.²

Realizing AI’s Full Potential

Chatbots are just one opportunity to incorporate AI efficiencies into medical practices. Billing, coding and generation of patient discharge sheets are the biggest uses of AI per the AMA survey.¹

Automated billing and coding software can improve invoicing and collections, thanks to the elimination of manual transcription errors. Natural language processing (an AI language model using human-like speech) can listen to patient/provider conversations, enter information into the patient’s electronic health record, and then search for and identify relevant information to assign billing codes for the patient’s visit. Additionally, AI-created customized patient discharge sheets can translate lengthy provider instructions into actionable summaries for greater patient understanding and improved compliance.

Back to the power of AI in diagnostics, AI can also facilitate research into difficult-to-diagnose conditions and treat them, including rare diseases. Providers can scan large data sets quickly to find similar symptoms and speed the process of diagnosis, including identifying examples of treatments that have been effective in similar cases.¹

The Future Is Near

Clinical trials looking at the efficacy of AI and chatbots are underway, including their potential to support medical adherence, improve sleep and improve diabetic foot care, as well as many others. And no wonder, with the enthusiasm for AI’s power and its seemingly limitless potential. Regulatory requirements also continue to be debated, an important consideration for nearly half of the AMA respondents who stated improved regulatory oversight would increase their trust in adopting AI.1 As more and more medical practices incorporate this new technology, adherence to safety and data privacy is imperative, as is focusing attention on that much-needed human approach that makes the practice of medicine such a meaningful discipline.