What Are Automated Insulin Delivery Systems?

Automated insulin delivery (AID) systems represent a paradigm shift in diabetes care. Often referred to as artificial pancreas systems, these technologies integrate three core components: a continuous glucose monitor (CGM) that measures interstitial glucose levels every one to five minutes, an insulin pump that delivers rapid-acting insulin subcutaneously, and a control algorithm that processes CGM data and commands the pump to adjust insulin delivery in real time. The goal is to maintain blood glucose within a target range—typically 70–180 mg/dL—with minimal manual intervention from the user.

Traditional diabetes management requires individuals to perform fingerstick blood glucose checks, calculate insulin doses based on carbohydrate intake, current glucose level, and anticipated activity, then manually inject insulin or adjust pump settings. This burden is not only time-consuming but also prone to human error. AID systems automate much of this decision-making by creating a closed loop: when the CGM detects rising glucose, the algorithm increases basal insulin delivery; when glucose falls, it reduces or suspends delivery to prevent hypoglycemia. The system can also deliver automatic correction boluses when glucose exceeds a threshold.

Commercial AID systems available as of 2025 include Medtronic's MiniMed 780G with SmartGuard technology, Tandem Diabetes Care's t:slim X2 running Control-IQ, and Insulet's Omnipod 5 integrated with the Dexcom G6 CGM. Each system employs a proprietary algorithm, but all rely on IoT principles: wireless communication between devices, cloud-based data storage, and remote access for users and clinicians. The FDA has cleared multiple AID systems, reflecting growing confidence in the technology's safety and efficacy.

The Role of IoT in Enhancing These Systems

The Internet of Things (IoT) is the backbone that makes closed-loop insulin delivery practical outside of clinical research environments. IoT refers to the network of interconnected devices—CGMs, pumps, smartphones, cloud servers—that continuously exchange data. In AID systems, IoT enables real-time sensing, algorithmic computation, and actuation to occur with sub-minute latency, replicating the homeostatic function of a healthy pancreas.

Real-Time Data Sharing and Remote Monitoring

One of the most transformative IoT capabilities is continuous data transmission to cloud platforms. Modern AID systems upload CGM traces, insulin delivery logs, and system status to secure servers, where they can be accessed by patients via smartphone apps and by healthcare providers through clinical dashboards. This remote monitoring allows diabetologists to review glycemic patterns, adjust therapy settings, and intervene proactively when a patient experiences recurrent hypoglycemia or hyperglycemia. For parents of children with type 1 diabetes, the ability to check glucose levels remotely while the child is at school provides significant peace of mind and enables timely corrective actions.

IoT also powers automated alerting. Systems can generate push notifications when glucose is trending dangerously low, when infusion sets become occluded, or when sensor life is expiring. These alerts reduce the cognitive load on users and help prevent acute complications such as diabetic ketoacidosis or severe hypoglycemia. Studies have shown that remote monitoring in AID systems reduces caregiver burden and improves time-in-range.

Personalized Treatment Algorithms

The continuous data stream enabled by IoT allows machine learning models to identify individual-specific patterns in insulin sensitivity, circadian rhythms, activity levels, and meal responses. For instance, the system can learn that a particular user experiences a pronounced dawn phenomenon and preemptively increase basal rates in the early morning. Other users may have exercise-induced insulin sensitivity that requires temporary reductions in delivery. Over time, these algorithms become increasingly tailored, leading to tighter glycemic control and fewer manual overrides. Some systems already incorporate predictive models that anticipate glucose excursions 30 to 60 minutes ahead, enabling preemptive adjustments before a deviation occurs.

Interoperability and Ecosystem Integration

IoT extends beyond the AID system itself to integrate with a broader ecosystem of connected health devices. Fitness trackers, smartwatches, smart scales, and food logging apps can feed contextual data into the insulin algorithm. For example, if a wearable detects that the user has started a vigorous workout, the algorithm can automatically reduce insulin delivery to prevent exercise-induced hypoglycemia. Similarly, data from a smart scale can be used to adjust mealtime boluses based on actual carbohydrate content. Achieving this level of integration requires standardized communication protocols—such as the Personal Health Devices (PHD) profile and Continuous Glucose Monitoring (CGM) standards—and secure application programming interfaces (APIs). Platform solutions like Directus serve as a content management and data orchestration layer, enabling developers to build interoperable IoT applications with structured data models and secure access controls.

Current State of the Technology

As of early 2025, the AID market has matured significantly. The Medtronic MiniMed 780G, launched in 2022, offers a hybrid closed-loop system that automatically adjusts basal insulin every five minutes and can deliver automated correction boluses up to once per hour. It integrates with the Guardian 4 sensor, which requires no fingerstick calibration. The Tandem t:slim X2 with Control-IQ uses a predictive algorithm that incorporates both current and projected glucose levels; it features an exercise mode that raises the target range to reduce hypoglycemia risk during physical activity. The Omnipod 5 is unique in being a tubeless, waterproof patch pump controlled entirely via a smartphone app, paired with the Dexcom G6 CGM. All three systems have demonstrated significant improvements in time-in-range and reductions in HbA1c in clinical trials.

Beyond commercial offerings, an active open-source community has developed do-it-yourself (DIY) closed-loop systems such as OpenAPS (Open Artificial Pancreas System) and Loop. These systems allow technically proficient users to combine compatible CGMs and pumps with community-developed algorithms. A landmark study published in Diabetes Care found that Loop users achieved a mean time-in-range of approximately 75%, comparable to or exceeding commercial systems. The open-source movement has pressured manufacturers to adopt more open interfaces and has accelerated innovation in the field.

Despite these advances, all current commercial systems are "hybrid" closed loops: they still require user input for meals (announcing carbohydrate intake) and sometimes for exercise. Fully autonomous systems that eliminate the need for meal announcements remain a research goal. The transition from hybrid to fully closed-loop is one of the most anticipated milestones in diabetes technology.

Future Developments: Smarter, More Autonomous Systems

AI and Machine Learning Integration

The next generation of AID algorithms will move beyond simple proportional-integral-derivative (PID) control and model-predictive control (MPC) to incorporate deep learning and reinforcement learning. These AI-driven approaches can learn complex, nonlinear patterns from large datasets—including historical glucose traces, insulin delivery, meal logs, activity data, sleep quality, stress levels, and even menstrual cycle phases. By combining these inputs, future algorithms will be able to predict glucose excursions with high accuracy and preemptively adjust insulin delivery before any deviation occurs. For instance, a model might learn that a user's postprandial glucose spike after a pizza meal is delayed by three hours and requires a dual-wave bolus—and it will execute this automatically without any meal announcement from the user.

Fully Closed-Loop Systems

The ultimate goal is a fully automated closed-loop that requires zero user intervention for meals, exercise, or correction doses. Achieving this will likely require a multi-hormone approach. Bi-hormonal systems that deliver both insulin and glucagon can prevent hypoglycemia by releasing glucagon when blood glucose drops, mimicking the natural counter-regulatory response. Several research groups, including the team at Boston University and the University of Virginia, have conducted clinical trials with bi-hormonal pumps, showing improved time-in-range and reduced hypoglycemia compared to insulin-only systems. Ongoing clinical trials are evaluating these systems in outpatient settings. IoT connectivity will be essential for coordinating the delivery of two hormones from a single pump and for enabling algorithms to adapt to the user's physiology in real time.

Integration with Smartphones, Wearables, and Smart Home Devices

Future AID systems will become deeply embedded in users' digital lives. Smartwatch apps will display glucose readings, allow quick bolus adjustments, and provide haptic alerts. Smart home assistants such as Amazon Alexa or Google Home could offer voice-activated status updates and emergency notifications. Data from smart scales (for precise carbohydrate tracking), continuous heart rate monitors (to detect stress or exercise), and smart beds (to monitor sleep quality) will feed into the algorithm to provide context-aware insulin adjustments. This level of integration demands a robust IoT infrastructure with low-latency data pipelines, reliable device pairing, and secure over-the-air firmware updates. Fleet management platforms that handle device provisioning, monitoring, and remote updates will become critical as the number of connected diabetes devices grows exponentially.

Challenges to Overcome

Data Security and Privacy

As AID systems become more connected, they become more vulnerable to cybersecurity threats. An attacker who gains control of an insulin pump could alter delivery rates with potentially fatal consequences. Manufacturers must implement end-to-end encryption, secure boot processes, hardware-backed key storage, and multi-factor authentication. Over-the-air (OTA) update capabilities must be designed with cryptographic signing to prevent malicious firmware installation. The FDA has issued comprehensive cybersecurity guidance for medical devices, and compliance is mandatory for premarket approval. Furthermore, patients' glucose and insulin data constitute sensitive health information protected under HIPAA in the U.S. and GDPR in Europe. Data storage and transmission must adhere to these regulations.

Device Interoperability and Standardization

The diabetes device ecosystem remains fragmented. CGMs, pumps, and algorithms from different manufacturers often cannot communicate directly because of proprietary data formats and closed APIs. This limits patient choice—if a person prefers a particular CGM, they may be forced into a specific pump ecosystem. Industry-wide adoption of interoperability standards, such as the IEEE 11073 Personal Health Devices standard and the Diabetes Device Interoperability (DDI) specification developed by the JDRF, is essential. Regulatory agencies are increasingly requiring interoperability as part of device approval, but progress has been slow. Open platforms like Directus can help bridge this gap by providing a standardized data layer that abstracts away device-specific protocols.

Regulatory Hurdles and Clinical Validation

Bringing a fully autonomous, AI-driven AID system to market requires rigorous clinical evidence. Adaptive algorithms that change over time based on user data present a challenge for traditional regulatory frameworks designed for static software. The FDA's pre-certification program for software as a medical device (SaMD) aims to streamline approval, but manufacturers must still conduct large, randomized controlled trials to demonstrate safety and efficacy. Post-market surveillance is equally important to detect rare adverse events and algorithm drift. Balancing innovation with patient safety requires close collaboration between developers, regulators, clinicians, and patient advocates.

Cost and Accessibility

Current AID systems are expensive. The initial hardware costs for a pump and CGM can exceed $5,000, and ongoing consumables—sensors, reservoirs, infusion sets—cost several thousand dollars per year. Insurance coverage varies widely, and many patients in lower-income brackets or with inadequate insurance cannot afford these systems. Expanding access requires competitive pressure from multiple manufacturers, value-based reimbursement models, and policy changes that mandate coverage for all diabetes devices. IoT infrastructure can help reduce overall healthcare costs by enabling remote monitoring and reducing hospitalizations for acute complications, but the upfront financial barrier remains daunting.

The Impact on Quality of Life

Beyond glycemic metrics, AID systems deliver profound improvements in quality of life. Users consistently report reduced diabetes distress, less anxiety about hypoglycemia, better sleep quality, and greater freedom to engage in spontaneous activities such as exercise or dining out. The constant mental arithmetic of carbohydrate counting, insulin dosing, and glucose trend prediction is offloaded to the algorithm, freeing cognitive bandwidth for other pursuits.

IoT-enabled remote monitoring also reduces the need for frequent clinic visits. Telehealth consultations, supported by data from the AID system, allow clinicians to manage patients more efficiently. This is especially valuable for those living in rural areas or with limited access to endocrinologists. Caregivers of elderly patients or children can participate in management without being physically present, improving safety and reducing stress for family members.

Clinical evidence continues to accumulate. A meta-analysis of hybrid closed-loop systems published in Diabetes Technology & Therapeutics found that users achieved an average of 12 percentage points higher time-in-range compared to sensor-augmented pump therapy, with significant reductions in nocturnal hypoglycemia. Long-term improvements in HbA1c are associated with reduced risk of microvascular complications, ultimately lowering the burden of comorbidities such as retinopathy, nephropathy, and cardiovascular disease.

The Role of IoT Infrastructure in Scaling AID Systems

To deliver on the promise of automated insulin delivery, the underlying IoT infrastructure must be reliable, secure, and scalable. This includes device management platforms that can handle millions of connected pumps and CGMs, data ingestion pipelines capable of processing terabytes of time-series glucose data daily, and cloud analytics engines that extract population-level insights to improve algorithms. Fleet management systems enable manufacturers to push OTA firmware updates, monitor device health, and proactively replace failing components before they impact patient safety.

A hybrid architecture combining edge computing and cloud processing is essential. Time-critical safety decisions—such as suspending insulin delivery when glucose is dropping rapidly—must execute locally on the pump or a dedicated controller to avoid network latency. Meanwhile, complex machine learning models that require training on large datasets can run in the cloud, and updated model parameters can be pushed to devices during non-critical times. This split architecture ensures both responsiveness and continuous improvement.

Security must be baked into every layer. End-to-end encryption between devices and the cloud, role-based access control for clinicians and patients, and comprehensive audit trails for all data access events are non-negotiable. Regular penetration testing and compliance with standards like ISO 27001 and HIPAA build trust among users and regulators. Platforms like Directus provide a flexible content management and data orchestration layer that can enforce these security policies while enabling rapid development of interoperable IoT applications.

Conclusion

The convergence of IoT technology and automated insulin delivery is reshaping diabetes care. Real-time connectivity, personalized algorithms, and integration with wearables and smart home devices are driving a shift from reactive management to proactive, automated regulation of blood glucose. While current hybrid closed-loop systems already improve outcomes and quality of life, the path to fully autonomous, multi-hormone artificial pancreas systems requires continued investment in AI, interoperability, cybersecurity, and accessibility.

Collaboration among device manufacturers, software developers, regulators, and patient communities will be critical to overcoming the remaining hurdles. As IoT infrastructure matures and open standards gain adoption, the vision of a true artificial pancreas—invisible, adaptive, and reliable—moves closer to clinical reality. For the millions of people living with diabetes, the promise of less burden and better health has never been more attainable.