Beyond the Exam Room: AI Across the ENT Patient Journey

Srinivas R. Kaza, MD, Chair-Elect, AAO-HNS Practice Management Education Committee

Consider a 50-year-old man who has been snoring for years and is increasingly congested. His primary care physician, part of a large regional health system, refers him to the otolaryngology department within that same system. The next available appointment is six weeks out.

On the same evening, he receives the referral and opens his phone. An AI-driven search algorithm identifies a nearby independent otolaryngology practice in his top results fed by his location, search history, and targeted marketing that had already identified him as a likely candidate based on demographic and behavioral signals. He clicks through to the practice website, where an AI chatbot walks him through symptom screening, insurance verification, and new patient intake. He schedules himself online that night. An automated confirmation lands in his inbox within minutes, followed by a series of AI-generated text and email reminders in the days before his visit.

He never called during business hours. He never left a voicemail. The practice never had to staff that interaction. By the time he arrived, his health history was already in the system. That is where the patient journey begins now—not in the waiting room, but in an algorithm.

Inside the Visit

At his appointment, ambient AI scribe technology listened unobtrusively as the otolaryngologist took his history and examined him. The physician faced the patient rather than a keyboard. In-office nasal endoscopy findings were documented in real time. Possible contributors to his snoring were reviewed, and AI-based clinical decision tools cross-referenced potential therapies—decongestants, nasal steroids, surgical options—against his medication list, surfacing interaction flags and relevant literature as the conversation unfolded. At checkout, he received a plain-language visit summary and AI-generated patient education materials he could understand and share with his spouse.

Behind the scenes, ICD-10 and CPT codes had already been generated. Nursing and scheduling messages were routed automatically. Prior authorization requests for any planned studies were submitted through AI-facilitated adjudication with his insurer. The superbill was cleaned, scrubbed, and submitted before the end of the day. This is not a hypothetical situation. It is where AI-enabled otolaryngology practice stands today—uneven in adoption, but unmistakably present.

What's Working Now

The most mature and immediately deployable AI tools in otolaryngology are ambient documentation systems. Approximately 90 commercial platforms now exist to capture clinical encounters and automatically generate notes, billing codes, and after-visit summaries with minimal clinician input. These systems integrate with major electronic health record (EHR) platforms and represent the only AI application with 100% adoption activity among surveyed health systems becoming a meaningful signal of practical value.¹ For many of us, documentation burden is one of the most corrosive forces in daily practice. Ambient scribing directly addresses that.

On the administrative side, AI is reshaping revenue cycle management, appointment scheduling, and, gradually, prior authorization (PA). A 2024 AMA survey found that physicians and staff spend an average of 12 hours per week managing PA requests, with each physician completing roughly 43 authorizations.² A 2025 study in JAMA Network Open demonstrated that AI-integrated PA workflows reduced median authorization time by 33.9%, from 4.2 to 2.8 business days—a reduction that translates directly into faster patient access to care.³ Across the health system broadly, AI-driven operational efficiencies project annual savings of 5 to 10%.¹

AI-generated after-visit summaries written in plain language deserve particular mention in the context of older or more medically complex patients. Patients juggling multiple chronic conditions often leave appointments uncertain about what was discussed and what comes next. A clear, readable summary, customized to health literacy level, can meaningfully improve adherence and reduce the callback burden on nursing staff. This is one of the more immediately valuable and under-appreciated applications of generative AI in an otolaryngology office.

The Clinical Frontier

Although clinical AI tools in otolaryngology show genuine promise, most remain in earlier stages of validation. Machine learning algorithms have been trained to predict nodal metastases in oral cavity and laryngeal cancers, assess margins in oropharyngeal malignancies, and analyze narrow-band imaging patterns to identify mucosal lesions.⁴ In some studies, radiomics-based AI analyses have matched or exceeded expert readers in identifying extranodal extension on diagnostic imaging.⁴ AI models are also being explored for snoring and sleep-disordered breathing analysis—a natural fit given our patient population—as well as digital voice biomarkers and sinus disease grading on imaging.

These applications hold true value, but they face real barriers. EHR interoperability remains inconsistent. Standardized data protocols are lacking. Algorithmic bias and data privacy concerns add complexity, particularly when training datasets underrepresent certain populations. Most clinical AI tools in our specialty remain in proof-of-concept phases, and the gap between publication and validated clinical integration is still wide.

The AAO-HNS Artificial Intelligence Task Force has been clear on this: AI should not replace mastery of clinical and surgical knowledge, nor eliminate the need for continuous learning. When clinicians use AI tools, the clinician remains responsible for exercising sound clinical judgment according to standards of care.⁵ That accountability framework matters. It is easy in the current moment to treat AI outputs as settled rather than as inputs requiring interpretation. They are not infallible, and the physician's judgment, earned through training and refined through experience, is what gives any AI recommendation its appropriate weight. 

Barriers, Privacy, and the Black Box Problem

Adoption barriers are real. A recent survey found that 77% of health systems cite immature tools as a primary concern, 47% point to financial constraints, and 40% identify regulatory uncertainty as an impediment.¹ Data privacy is a recurring theme. Traditional data aggregation approaches (so-called “data lakes”) create risk when patient data cross institutional lines.

Federated learning has emerged as a promising alternative: AI algorithms are trained across multiple institutions without patient data ever leaving each site.⁶ Model parameters are generated locally and aggregated centrally. This architecture may prove especially important for specialty-specific AI development in otolaryngology, where annotated datasets are inherently smaller than those available in primary care or radiology.

Transparency in how decisions are made by AI (what AI ethicists call explicability) is crucial to ensuring fairness and equity in care provided. The “black box” problem describes AI systems whose internal logic is not visible to the clinicians relying on them.⁷ As these tools grow more capable, the pressure to explain their reasoning to patients and payers alike will increase. For any otolaryngology practice pursuing broader AI integration, governance frameworks addressing accountability, transparency, and liability are foundational, not optional. 

A Word on Sustainability

One dimension of the AI conversation that rarely surfaces in clinical settings is environmental impact. Training a single large natural language processing model GPU emits an estimated 626,000 pounds of CO2 equivalents—more than the lifetime emissions of five average American cars.⁵ In 2023, AI already accounted for roughly 1% of global carbon emissions, a figure projected to rise as adoption accelerates.⁵ This does not argue against AI use, but it is worth acknowledging as part of an honest accounting of costs and benefits as our professional societies develop sustainability commitments.

Augmented, Not Artificial

The patient in our opening vignette found his way to an independent otolaryngology practice through an AI-driven search, completed intake through an AI chatbot, and experienced AI-facilitated care from encounter to billing, all while the physician remained focused on him. That is the balance worth pursuing.

For practices evaluating where to begin, the evidence is reasonably clear: ambient documentation systems offer an immediate, tangible workflow benefit with manageable implementation risk. Revenue cycle and prior authorization automation follow closely, particularly for practices frustrated by administrative overhead. Clinical AI tools deserve monitoring and measured adoption, with validation preceding routine use.

Throughout, there is a growing and important reframing worth embracing: perhaps AI should stand not for artificial intelligence, but for augmented intelligence. The distinction is more than semantic. Artificial suggests a replacement; a system that acts in place of the physician. Augmented suggests something closer to what these tools do at their best: extend the clinician's reach, surface information faster, reduce friction, and free the physician to do what only a physician can do. The judgment, relationships, and accountability all remain ours.

The practices that will integrate these tools most effectively are those that approach them neither with uncritical enthusiasm nor reflexive skepticism, but with the same disciplined, evidence-driven habit of mind that defines good clinical care. AI has arrived across the otolaryngology patient journey. Our job is to make sure it serves the patient, and the physician, rather than the other way around.

Disclosure: The author has no financial relationships relevant to this article to disclose.

AI disclosure: AI-based tools were used in the preparation of background materials informing this article. The article itself was written by the author.

References

1. Ayoub NF, Rameau A, Brenner MJ, et al. American Academy of Otolaryngology–Head and Neck Surgery (AAO-HNS) Report on Artificial Intelligence. Otolaryngol Head Neck Surg. 2025;172:734-743.
2. American Medical Association. 2024 AMA Prior Authorization Survey. Chicago, IL: AMA; 2024.
3. Chen WC, Carpenter C, Sidiqi B, et al. Integrating Prior Authorization Into Clinical Workflows for Care Access and Practitioner Experience. JAMA Netw Open. 2025;8(12):e2549093.
4. Panwar A, Brenner MJ. Artificial Intelligence and Otolaryngology–Head and Neck Surgery: Implications for Clinical Practice, Innovation, and Education. AAO-HNS Bulletin. January 2024.
5. Ayoub NF, Rameau A, Brenner MJ, et al. AAO-HNS Report on Artificial Intelligence. Otolaryngol Head Neck Surg. 2025;172:734-743.
6. Angus DC, Khera R, Lieu T, et al. AI, Health, and Health Care Today and Tomorrow: The JAMA Summit Report on Artificial Intelligence. JAMA. 2025;334(18):1650-1664.
7. Morley J, et al. The Ethics of AI in Health Care: A Mapping Review. Soc Sci Med. 2020;260:113172.