Future of AI in Healthcare: Agentic AI and the Shape of Things to Come

In the AYDIJ article, “AI in Action: The Pragmatic Path to Success,” Holly Hallman shared lessons from her AI journey—how ideas became pilots, pilots matured into practice, and practice transformed care for patients and providers alike. Her pragmatic perspective illustrated how organizations can generate sustained, tangible value with GenAI across clinical care, operations, patient experience, and research.

That shift from pilots to practice has accelerated even faster than expected. A new wave is now emerging, focused on what many see as the next frontier in AI: agentic AI. To explore this evolution, I am excited to introduce Justin Coran as the latest contributor to the AYDIJ series.

Justin is a distinguished thought leader at the intersection of AI and healthcare delivery. I first met him at an analytics and AI conference roundtable, where his ability to move seamlessly from strategic vision to practical application captured the attention of everyone in the room. Few leaders match his clarity on what it really takes to transform care with AI.

After serving as chief analytics officer at Renown Health, Justin is now chief data and AI officer at Medical Concepts, where he brings his visionary outlook to a broader platform, anchored by the same pragmatic, real-world perspective that defines his work to improve lives.

In this article, “Future of AI in Healthcare: Agentic AI and the Shape of Things to Come,” Justin examines how healthcare organizations are using AI today and where agentic AI is poised to take us next, from smart hospitals to diagnostic agents and administrative automation to patient engagement. His perspective offers a compelling glimpse into a future where AI systems don’t just respond but act—and reshaping healthcare delivery at its core.

– Ryan Sousa

The healthcare industry is going through a profound period of innovation and evolution. Health systems are reimagining their facilities with smart hospital technology to boost patient care and improve provider efficiency. Physicians are using ambient listening technology and automatic clinical documentation tools to capture more information than ever while engaging with patients.

Insurers have deployed automation, prediction, customer service bots, and other advanced technology to improve how they do business. Pharmaceutical companies are leveraging AI to speed up drug discovery and investigate how to lower drug prices for the patient. In fact, the term artificial intelligence—or “AI”—has become mainstream, where the average person is intrigued to know when or how they will interact with AI in their daily life.

Soon, patients will interact with AI and robots in their healthcare facilities. This article will discuss the current era of AI and potential areas where AI may transform healthcare in the next few years.

What is the difference between AI and an AI Agent?

AI has evolved through several eras, beginning in the 1930s when Alan Turing laid the groundwork with concepts like the Turing Machine (1936), and continuing to today, when people can speak into their smartphones and ask a voice assistant nearly any question and receive an informed answer. Many people have interacted with products that were produced using statistical techniques such as machine learning and natural language processing. More recently, the world has been taken by storm by generative AI, a type of artificial intelligence that can create original content, such as text, images, video, audio, or software code, in response to a user’s prompt or request. Many of these tools can be commanded through voice recognition.

Generative AI (GenAI) is commonly associated with the applications ChatGPT (OpenAI), Grok (xAI), Gemini (Google), and Siri (Apple). GenAI is composed of large language models, and during development, it attempts to consume as much information as possible to produce accurate answers to prompts. Usually, large language models use all the information available on the internet, plus private data provided by the development team.

AI in healthcare refers to the use of technologies, such as machine learning, natural language processing, and computer vision, to improve medical processes, diagnostics, treatment, and patient care. The term “AI” can include a broad range of clinical and administrative applications. For example, in diagnostics, AI can analyze medical images or patient data to detect diseases like cancer, Alzheimer's, or physical anomalies. Within personalized medicine, AI can tailor treatment plans based on patient preferences, genetic profiles, and previous medical history. AI accelerates drug development by predicting molecular interactions or identifying new chemical compounds.

Regarding administrative operations, AI can automate tasks like scheduling, billing, or medical record management to reduce costs and errors, or act as a virtual health assistant and provide patient support, answer health queries, or monitor symptoms remotely. Marketing staff can use AI to help create content for the company, including translating current content into multiple languages. Video editing is a breeze with AI, as is the creation of new programming code for IT staff.

The next era of AI involves AI agents. Agentic AI describes AI systems that are designed to autonomously make decisions and act, with the ability to pursue complex goals with limited supervision. It brings together the flexible characteristics of large language models with the accuracy of logical programming.

The key difference between agentic AI and GenAI is its proactive AI-powered approach, whereas GenAI is reactive to the user’s input. Agentic AI can adapt to different or changing situations and can make decisions based on new contexts. AI agents are not specific to healthcare, but some examples that have been deployed in healthcare include diagnostic agents, virtual health assistants, clinical decision support, remote monitoring, chatbots, and personalized care assistants.

To summarize, AI in healthcare is a broad categorization that covers any AI use in the medical field. An AI agent is a specific system designed to act autonomously, which may or may not be used in healthcare.

Current State of AI in Healthcare

Today, there is an arms race among companies to be the leader in deploying AI for the healthcare industry. Arguably, the current leaders are likely Epic and Microsoft due to the fact they had existing business relationships with the entire healthcare industry, but the space is crowded with hundreds of companies that claim to have a product that can aid healthcare. In the U.S., Epic provides the market-leading electronic health record system (EHR), and they currently have over three dozen AI applications that enhance features in their EHR. These tools include Art for Clinicians (automated response technology); Emmie, a patient-facing chatbot; and Penny, a revenue management assistant.

At Epic’s 2025 User Group Meeting, Epic announced the development of a new generative medical event model called CoMET to help doctors use real-world evidence to improve patient treatment and care decisions. Epic’s innovation work has large-scale implications for leveraging AI to improve patient outcomes, reduce costs, and enhance the overall process of delivering care. However, the benefit may be limited due to Epic’s business model of only servicing large health systems or hospitals with a certain threshold of beds. This means that many physicians who are not employed by these institutions will not be able to access Epic EHR software or some of the best AI integration work occurring in healthcare technology today.

Microsoft is deploying AI across its entire product catalogue. The healthcare industry uses multiple products from Microsoft for operations. For backend IT infrastructure, there is Microsoft Cloud for Healthcare. This innovation connects care experiences, enhances team collaboration, empowers healthcare workers, and unlocks clinical and operational insights.

Microsoft has released new AI models in Azure AI Studio, demonstrated how Fabric can provide different capabilities for healthcare data solutions, and created a new healthcare agent service in Copilot Studio. A Microsoft company, Nuance, can integrate with Epic to provide ambient listening technology and transcribe patient encounters into clinical documentation, saving physicians hours in their day.

In July, Microsoft released its latest Healthcare AI paper, “Sequential Diagnosis with Language Models,” which Microsoft claims is “the path to medical superintelligence.” In the paper, researchers reported the creation of a new benchmark, SDBench, based on clinical cases. Unlike most scenarios, performance was based on diagnostic accuracy and the total cost of reaching the diagnosis, and the new model appears to achieve 80% accuracy. This is the closest a research team has come to creating an AI agent that can make multiple diagnoses and perform competitively with physicians.

Although the previous paragraphs highlighted some of the key advances from Microsoft and Epic, other notable companies are making an impact in healthcare with their AI technology. Stryker, the largest medical device manufacturer, is creating devices that interconnect together and will form the backbone of the smart hospital of the future. Care.AI provides remote patient monitoring and virtual nursing opportunities while AI Doc integrates with medical images to flag anomalies within the images for review by clinicians. Medtronic, Johnson & Johnson, and NVIDIA are additional companies that have popular products with integrated AI features for healthcare. Below are other key companies and their AI innovations that may experience rapid growth in the near future:

Automating Medical Administration

• Augmedix: Their latest announcement was the launch of a fully automated, generative AI medical documentation tool for emergency departments. This builds on their Augmedix Go solution, a clinician-controlled mobile app that uses generative AI to instantaneously create a fully automated draft medical note after each patient visit.

• DeepScribe: Their AI-powered platform uses ambient voice technology and natural language processing to automate the creation of medical notes. DeepScribe says they have built their tool using the world’s largest database of natural patient conversations, and state that it is significantly (32%) more accurate than GPT4.

Disrupting Medical Imaging

• Butterfly Network: Created a portable handheld device that uses an ultrasound-on-chip technology to replace the traditional transducer system with a single silicon chip, emulating any type of transducer (linear, curved or phased), allowing for whole-body imaging from a single probe. By combining semiconductors, artificial intelligence, and cloud technology in a pocketable form, the Butterfly iQ is making remote medical imaging a reality.

• Enlitic: Uses the power of deep learning technologies, specifically its prowess at certain forms of image recognition, to harvest the data stemming from radiology images and apply it in unique medical cases. Their system can interpret a medical image in milliseconds and integrates seamlessly into any existing health system.

• Tempus Radiology: Formerly known as Arterys. Originally focused on cloud-based platforms for faster radiology image examination and reduced missed detections. Arterys is now a key component of Tempus Labs’ precision medicine platform. Their Pixel platform aids radiologists and oncologists by automating image analysis, quantifying lesions, tracking disease progression, and generating detailed reports.

Health Management

• Ada Health: Ada’s app has almost 14 million users. It takes reported symptoms, matches them with symptoms of patients of similar age and gender, and reports the statistical likelihood that the patient has a certain condition. It is currently available in English, German, Spanish, Portuguese, Swahili, Romanian, and French. The app became available in Epic App Orchard in May 2022 and received a Class IIa medical device certification in Europe in the fall of 2022.

• MySense AI: MySense is a well-being application. It collects data related to the activities of daily living through passive IoT sensors. Its AI algorithm learns an individual’s behavior patterns to establish what ‘being well’ looks like for that person. The platform allows patients to monitor their health at home and identify declines in health in real time.

• Skinvision: Developed an app to remotely evaluate suspicious skin lesions. Users take a picture of the lesion in question, upload it to the app, fill out a short questionnaire and will receive a preliminary evaluation prepared by the AI algorithm in a few minutes. This preliminary diagnosis will be followed by a final evaluation conducted by a human dermatologist in 1-2 days.

Care Coordination and Disease Detection

• Viz.ai: A leading company in the application of artificial intelligence to disease detection and care coordination. Viz.ai One is an AI-powered platform that rapidly analyzes medical images to identify suspected diseases, such as strokes, aneurysms, and intracranial hemorrhages. This technology significantly accelerates the diagnostic process, enabling faster treatment decisions and potentially improving patient outcomes.

• Hippocratic AI: Hippocratic AI focuses on developing generative AI “agents” that assist healthcare professionals by handling non-diagnostic tasks through empathetic, conversational interactions. In addition, Hippocratic AI has an AI nursing program that costs approximately $9 per hour.

AI transforms how clinicians and patients experience healthcare. AI also transforms how administrative staff work in healthcare. AI tools aid in day-to-day tasks. Tools such as Microsoft CoPilot help create PowerPoint slides, create summary documents about a meeting, and aid in email communications.

Data engineers and other types of IT professionals use AI tools to help them write and qualify code, speeding up the development process by hours and in some cases, days or weeks. This includes software development. AI will aid existing IT staff in creating new software on behalf of the company. Previously, health systems rarely had expertise in software development and would typically engage a third party to develop on their behalf. Soon, AI agents will be the preferred development partners for IT staff working for health systems and providers.

Agentic AI and the Smart Hospital of the Future

Provider organizations are in the process of building new healthcare facilities (i.e., hospitals, ambulatory surgical centers, clinics), designing them with the ability to evolve into a “smart hospital.” Overall, a smart hospital incorporates telemedicine, connected medical devices and remote monitoring as part of its technology integration plan.

Smart hospitals are transforming healthcare by integrating the Internet of Things (IoT) to enhance patient care. By connecting medical devices and systems, IoT enables real-time monitoring, data analysis, and improved communication among healthcare providers. This integration leads to more efficient operations and better patient outcomes.

Within a smart hospital, wearable devices track vital signs and alert providers to changes; IoT tags help locate equipment and manage inventory; sensors adjust lighting and temperature for patient comfort; and AI agents help with predictive maintenance of equipment and can schedule timely repairs.

Companies are choosing to build smart hospitals due to their advantages over less technologically integrated facilities. These advantages include:

• Continuous data collection allows for early detection of issues.

• Automation reduces manual tasks, freeing staff to focus on patient care.

• Real-time analytics support that informs clinical decisions.

• Tailored therapies based on real-time data.

• Improved patient engagement and more active participation in care due to increased access to personal health data and faster communication with staff.

From a business point of view, if a smart hospital facility is designed and constructed with intention, then it should create a competitive edge in the market versus other facilities. The cost to retrofit an older building to a modern specification, matching today’s smart hospitals, could be cost-prohibitive, and if that cannot occur, then the older facility will be unable to match the same patient experience or technological expectations provided at the smart hospital.

Smart hospitals treat data as an asset and will invest in the underlying infrastructure to ensure all technological systems that run the business are interconnected. That reality starts with the facility’s power and networking infrastructure, and care should be taken to think about the facility from a patient’s perspective.

For example, when purchasing furniture for patients for waiting areas, ensure that each furniture piece includes power, USB, and USB-C connections, just like the seats and lounges at the airport. Why should a healthcare facility’s waiting room be inferior to an airport lounge? To include modern furniture, the facility would need to plan for network and power cables at each furniture location. The same consideration is needed for each connected device in a smart hospital, such as wall panels, TVs, computers, mobile stations, and medical devices.

Building a smart hospital is an opportunity to be innovative. Future-forward designs have started incorporating piezoelectric floor tiles in busy areas. The floor tiles convert the kinetic energy from footsteps into electricity, powering local devices such as lights and displays. This technology is being pioneered in Japan and has seen success in Shibuya and Tokyo Stations. Within the next three to four years, Optimus Robots or a similar technology will be advanced enough to work within medical facilities. In the beginning, robots will start conducting tasks associated with patient transport, environmental cleaning, and supply transportation. For robots in healthcare facilities to become reality, they will need space to charge batteries overnight and enough network bandwidth to operate their AI brains.

Essential technologies for a smart hospital include predictive analytics, ambient AI, smart hospital TVs, smart hospital beds, and digital door signs. Smart TVs in hospitals and medical facilities have been underutilized for years. Part of that reason is due to the lack of connectiveness across technologies that ultimately would be displayed on those TVs. In addition to using smart TVs for entertainment, patient education, and videoconferences with doctors and family members, some health systems will also include virtual rehab content and exercise recommendations following a procedure.

LG and Samsung have partnered with health systems to build an operating system for smart hospital TVs with integrated AI. At the nurse’s verbal request, they can send a signal to the TV to display the patient’s lab results or procedure time. The system has built-in voice and facial recognition security, and once the nurse is “logged in,” the system will recognize the nurse via camera or microphone and confirm access permission to view the patient’s medical information.

In addition, the phones of medical staff can connect to an indoor positioning system to display information (such as their name and role) on a sidebar on the TV, when they walk into a patient’s room. Smart TVs can act as digital signage for nearly any data source connected to the hospital.

For a smart hospital to achieve its full potential, each business system should be interconnected where data flows freely between systems and into a centralized data lake. Systems such as the EHR, radiology system, lab system, HR, or billing are not designed to hold and centralize data from disparate systems. Therefore, an enterprise data warehouse or cloud-based lakehouse environment is critical for providing the infrastructure to centralize the company’s data and use the features of a Smart TV to display that information meaningfully to the patient or staff member.

Getting Started in the Healthcare AI Era

Each provider organization will get started on a different pathway but eventually end up at the same destination in their journey to implement AI to better their operations and outcomes for the future. The number of staff, degree of financial resources, and access to industry connections will differ whether your organization is part of (1) hospitals and health systems; (2) ambulatory surgical centers; (3) physician clinics; (4) insurance industry; or (5) pharmaceutical industry. Whether your company is large, medium-sized, or small will have an influence on how fast the company can onboard new technology and AI.

Before starting the journey to implement new technology, such as AI, set clear goals within the company such as a financial budget for capital and operational costs, define what operations or outcomes are expected to be influenced by the technology, agree on which staff members will have access to the technology, and determine an evaluation method to assess the realized value of the new technology.

The maturity of a company’s IT, data, and analytics strategy will be key to successful AI implementations. Poor data quality across data repositories will elongate and complicate an AI implementation timeline. Data fields must be standardized and have clear definitions when used in conjunction with AI technology. Mature companies will establish or have an existing data and analytics governance program, which is a perfect forum for AI policy creation. AI tools may require end-user support, which in turn may require specific education and skills among staff.

Two considerations need to be discussed prior to getting started on your journey: build vs. buy and hybrid strategy.

Build vs. Buy

For hospitals, health systems, and physician clinics at the beginning of their data, analytics, and AI journey, I would recommend buying an AI product or service that enhances clinical practice. Examples include ambient listening and clinical documentation (e.g., Nuance), AI features built into an EHR and available via an EHR company (e.g., Epic, Cerner), or AI products related to Radiology (e.g., Tempus Radiology). Numerous products can aid administrative tasks; however, the costs of implementing AI into administrative functions may be cost-prohibitive for smaller or lower-revenue companies. AI deployed into clinical operations will be easier to track return on investment and value of implemented product.

Hybrid Strategy

Insurance companies, pharmaceutical companies, or any large company with sufficient revenue and budget for AI should consider a hybrid strategy where the best technology available is purchased, while in-house research and development teams build next-generation AI alongside clinical colleagues. Different types of AI technology such as large language models and AI agents can be built by development teams within either cloud-based or on-site server infrastructure.

Cloud-based infrastructure from Microsoft, Amazon, or Google is ideal due to the delivery of an infrastructure as a service business model, ease of scaling infrastructure in the cloud, access to best-in-class development tools, familiarity and skillset to operate cloud-based resources in the cloud, and the potential to partner with Microsoft, Amazon, or Google when building or operating one of their clouds for the first time. Agentic AI deployments and development will benefit from cloud-based infrastructure or renting time on a supercomputer from a datacenter or specialized technology provider (e.g., Switch, IBM, NVIDIA).

Companies that choose a build strategy should consider who will conduct the development. Some companies outsource AI projects while others recruit a development team of data scientists, data engineers, software developers, and AI specialists. AI development and associated IT teams will likely request tools and applications available in cloud environments to conduct their work. A select group may prefer on-site servers and enterprise software to accomplish similar AI development tasks.

Any company that decides to undertake an in-house AI development project should think carefully about its ability and resources to achieve success. AI programs composed of large language models require large data centers to support the computing power required to ask the AI questions and receive accurate answers. AI data centers are typically only available to Fortune 50 companies such as Microsoft, Google, OpenAI, xAI, and NVIDIA.

Alternatively, small language model development may be a good starting point for health systems and providers wishing to undertake their first in-house development project. Developing AI agents can be accomplished in both small- and large-scale infrastructure, making those projects accessible to health systems and providers. The development of AI agents can be affordable but requires extensive expertise among the development team and collaboration with the clinical workforce.

Final Thoughts

We have entered into the AI era of agentic AI, and this new technology will transform how we think about healthcare and how we deliver services in the healthcare facility of the future. The race to acquire AI technology has just begun, and no company has been left behind in 2025. Companies wishing to get started have multiple entrance points. One recommendation is to deploy an AI tool for front-line clinical staff and follow that success with administrative AI tools to make operations more efficient and value-driven.

More analytically mature companies have the opportunity to buy the best technology on the market while experimenting with in-house AI development. AI does not have to be cost-prohibitive, as the primary cost drivers are the amount of data and the amount of usage. It can be deployed to benefit both staff and patients. It is only a matter of time before the general public will expect healthcare to embrace AI in the same fashion as other industries.

Exciting times ahead for providers and patients. Enjoy the journey!

References

King, D., and Nori, H. (2025, June 30). The Path to Medical Superintelligence. https://microsoft.ai/news/the-...;

Mucci, T. (2024, October 21). The history of AI. IBM. https://www.ibm.com/think/topi...;

Teaganne Finn, T. and Downie, A. (2025). What are the key differences between agentic AI and generative AI? https://www.ibm.com/think/topi...;

Turing, A. M. (1936). On Computable Numbers, with an Application to the Entscheidungsproblem. Proceedings of the London Mathematical Society, Series 2, Vol. 42, pp. 230–265.

Mike. (2025, August 26). The Complete Guide to Healthcare TVs: LG & Samsung. https://elevate-b2b.com/health...;