AACE 2026: AI moves from hype to reality in diabetes care

"A person with diabetes who's wearing a CGM [continuous glucose monitor] and an Apple Watch is going to generate more health data in one month than their grandparents would have done in their entire lifetime," said David Ahn, MD, during a session at the 2026 AACE Annual Meeting in Las Vegas, NV. 

This proliferation of data creates both opportunity and burden, noted Dr. Ahn, chief of diabetes services at the Mary & Dick Allen Diabetes Center, in Newport Beach, CA, and AACE Endocrine AI board member. 

While CGM systems and automated insulin delivery platforms provide highly structured, high-frequency data, other clinically relevant inputs—such as nutrition, stress, and medication adherence—are less consistently captured and often difficult to quantify. The resulting variability and data volume contribute to clinical complexity but also make diabetes particularly well suited for AI-driven pattern recognition. 

Dr. Ahn outlined several applications of AI that are beginning to demonstrate clinical relevance in diabetes management. 

AI-supported insulin titration. Even relatively simple, rules-based approaches—such as voice-enabled systems that guide basal insulin adjustments—have been shown to help patients reach target doses more efficiently than standard care, likely by improving day-to-day adherence. 

Interpretation of CGM data. Beyond conventional metrics such as time in range, AI-enabled approaches are being developed to identify clinically meaningful patterns, including missed boluses, compression artifacts, and glycemic excursions that clinicians often recognize retrospectively. These approaches may also support identification of glycemic patterns or “phenotypes” that could inform more individualized management strategies.  

Clinical decision support. Dr. Ahn noted that in controlled studies, AI-based insulin dosing recommendations have demonstrated outcomes comparable to those achieved by endocrinologists, with high rates of clinician acceptance. He said that such tools may be particularly useful in settings where access to specialty care is limited.  

Patient engagement and behavioral interventions. Dr. Ahn highlighted a JAMA-published 12-month randomized trial in individuals with prediabetes in which an AI-powered lifestyle intervention achieved outcomes comparable to a traditional, human-led program, with higher initiation and completion rates.  

The difference, he noted, likely reflects the ability of AI systems to deliver timely, personalized nudges based on real-world behaviors.  

Dr. Ahn also discussed emerging agentic AI and neural network–based systems. "Agentic AI takes multiple steps and actions autonomously, meaning it’s doing it on its own without needing human intervention at every step," he said.  

Unlike traditional rule-based algorithms, neural networks learn from large datasets and can be retrained over time, allowing them to recognize patterns and adapt based on prior data. Such approaches may enable more advanced automated insulin delivery and increasingly individualized management strategies, with the potential to incorporate learning across patients.  

Dr. Ahn pointed out that early research suggests neural networks can replicate the performance of existing automated insulin delivery algorithms while requiring less computational power, which could allow for smaller and more efficient devices.  

"This is one of the most exciting aspects of AI—and maybe one of the most advanced aspects," he said.  

Dr. Ahn concluded with several practical insights for endocrinologists: 

• AI is already clinically meaningful: Measurable A1C reductions and improved engagement are being demonstrated today.  

• Assistive AI will dominate near-term adoption: Most tools will remain clinician-in-the-loop for safety and regulatory reasons.  

• Diabetes is uniquely suited for AI: High-frequency data and clear outcome metrics create an ideal testing ground.  

• Patient engagement may be the biggest win: AI’s ability to deliver timely, personalized nudges could address longstanding adherence challenges.  

• Autonomous systems are coming—but not here yet: Fully independent AI management remains a future goal, with ongoing work in neural networks and agentic systems.  

For endocrinologists, the message is clear. AI is becoming an increasingly central component of diabetes management. "Right now, AI is more of a copilot, but the long-term vision is something much closer to autopilot," Dr. Ahn concluded.

He disclosed serving as a speaker and advisory board member for multiple industry organizations.