As of March 2025, over 70% of healthcare institutions across North America and Europe have either begun exploring or experimenting with generative AI applications in clinical and operational settings, according to a recent survey conducted by a major research institute in collaboration with healthcare service providers.
This burgeoning interest is particularly evident in healthcare IT budgets, where up to 25% of the funds are allocated toward AI research and infrastructure, driving rapid developments in AI agents—an evolution from traditional chatbots to systems capable of autonomous actions.
By early 2025, the global healthcare AI market is valued at approximately $3.696 billion, and it is projected to surge to $61.38 billion by 2034. The U.S. market alone is estimated at $8.41 billion, with a compound annual growth rate (CAGR) exceeding 35% over the next decade.
In Europe, the healthcare AI market is expected to reach nearly €50 billion by 2030, with growth rates ranging from 37% to 39%. North America commands nearly 60% of the global market share, while Europe accounts for about 20%. This disparity is closely tied to substantial capital investments and the development of technology ecosystems in these regions.
In the U.S., nearly 80% of hospitals have already implemented some form of AI solution, especially in diagnostic imaging, environmental documentation automation, and predictive scheduling. Insurance companies report a usage rate of 84%, primarily in process approvals and claims risk management, which has improved speed and efficiency while reducing Full-Time Equivalents (FTEs) by about 25%. In Europe, while the adoption rate is slightly lower (around 65%), it is expected to grow to 78% by 2026.
However, what stands out most is the transition from non-agentic AI, which performs tasks such as image recognition or text generation, to "agentic AI." Unlike traditional systems, agentic AI can cross-reference multiple data sources, make decisions, and execute tasks autonomously.
For instance, some systems can now automatically generate reports, schedule follow-up appointments, and even coordinate patient transportation after analyzing diagnostic images—all without human intervention.
Microsoft's recent launch of a "diagnostic orchestration system" illustrates the potential of agentic AI. Early tests showed an accuracy rate exceeding 80% in handling complex cases, significantly outperforming the 20% accuracy rate of traditional doctors, and drastically reducing diagnostic errors and unnecessary tests.
Though not yet clinically deployed, Mustafa Suleyman, one of the system's founders, predicts that within 5 to 10 years, this system will achieve near-zero errors in clinical settings, marking a true "medical super-intelligence."
Another notable example comes from Cedars-Sinai Medical Center in Los Angeles, which, in partnership with K Health, launched the virtual care platform "CS Connect" in 2023. By 2025, the platform had already been used by over 42,000 patients.
This system acts as an AI agent that autonomously collects patient complaints, analyzes medical histories, suggests preliminary diagnoses, and generates treatment plans for doctors to review or revise. According to a study published in the Journal of Internal Medicine in 2025, 77% of AI recommendations were deemed to be the best treatment options, compared to just 67% from human doctors.
This was especially evident in cases involving recurrent urinary tract infections. The system not only alleviates administrative burdens but also allows doctors to dedicate more time to making nuanced, human-centered decisions and interacting with patients.
Ellipsis Health, headquartered in San Francisco, also represents a breakthrough in the field. Its AI platform, Sage, secured over $45 million in Series A funding from investors including CVS Health Ventures, Salesforce, and Khosla Ventures. The Sage platform uses an emotional recognition engine (dubbed the "Empathy Engine") to actively engage with patients during post-discharge periods.
It monitors medication adherence, coordinates transportation, interprets discharge instructions, and even reports emotional fluctuations or potential risks in real time. This has notably enhanced chronic disease management and reduced readmission rates.
In the field of radiology, Israeli company Aidoc has deployed AI systems in top hospitals across Europe and the U.S. to detect conditions like stroke, pulmonary embolism, cervical fractures, and abdominal emergencies.
A multicenter study in Europe revealed that its pulmonary embolism detection algorithm had a sensitivity rate of 84.8% and a specificity rate of 99.1%. Similar high performance was observed in brain hemorrhage detection, with sensitivity and specificity both exceeding 90%.
These systems operate "around the clock" in practice, sending alerts to radiologists after analyzing images while also summarizing cases and optimizing workflows. For example, in Mercy Health System, the deployment of Aidoc reduced ICU stay durations and controlled overall treatment costs, demonstrating the significant practical value of these systems.
A recent Pew survey reveals that American adults' trust in medical AI varies by age: 44% of those aged 18–29 reported feeling "comfortable" with AI, while only 35% of individuals over 65 felt the same.
Overall, 85% of users reported positive experiences with AI health tools, although 41% hesitated to provide sensitive information due to privacy concerns. Among chronic disease patients, 72% believed AI could aid in monitoring and improving medication adherence.
Among healthcare professionals, the usage rate of AI has surged from 38% in 2023 to 66% in 2024, with over two-thirds of doctors reporting clear benefits from AI. The potential of AI to reduce medical errors and improve efficiency is widely recognized.
For instance, a study conducted by OpenAI in collaboration with Kenya's Penda Health clinics showed that AI reduced diagnostic errors by 16% and treatment errors by 13% during clinical consultations, and was deemed to have significant educational value for junior physicians.
Despite its promising future, agentic AI in healthcare presents new and unprecedented risks. Data security remains a paramount concern, as AI agents must access sensitive personal medical records and treatment systems.
If these systems are misused or compromised, the consequences could be severe. In regions like Europe and the U.S., strict regulations such as GDPR and HIPAA govern the use of data, placing high demands on AI systems to protect privacy and transparency.
Another critical issue is accountability. When agentic systems autonomously execute scheduling, diagnosis, or treatment recommendations, who bears responsibility when things go wrong?
Is it the developer, the healthcare provider, or the regulatory body? This question remains unanswered, and resolving it will be essential for the safe integration of agentic AI in healthcare.
AI itself is not infallible. Like humans, it can make mistakes due to incomplete training data or algorithmic biases. Determining when AI can take full control of processes and when human oversight is required becomes crucial.
European regulators are pushing for standards in AI explainability, transparent decision-making, and auditability to ensure a balance between agentic AI capabilities and regulatory requirements.
Finally, ethical concerns cannot be overlooked. Whether it's algorithmic fairness, emotional sensitivity, or ensuring that patients' needs for empathy are met, developers and healthcare providers must ensure that technology does not strip away the inherently human aspects of care.
Overall, North America and Europe are transitioning from the exploratory phase of agentic AI in healthcare to accelerating its adoption. The market is expanding rapidly—nearly 80% of the combined North American and European markets—driven by both capital investments and technological advancements.
Clinical applications are broadening beyond traditional image recognition and document automation to include proactive treatment recommendations, process automation, and smart patient interactions.
Platforms like Cedars-Sinai's CS Connect, Ellipsis Health's Sage, and Aidoc's deployments in top healthcare centers are not just theoretical examples—they are tangible applications that showcase the transformative potential of agentic AI.
Meanwhile, the involvement of tech giants like OpenAI and Microsoft, along with Europe's InvestAI initiative, signals that agentic AI is becoming a strategic industry trend.
Looking ahead, we can expect agentic AI to dramatically enhance healthcare efficiency, increase access to services, drive personalized treatment, reduce errors, and optimize resource allocation. It will shine particularly in remote care, chronic disease management, and clinical research.
However, this progress hinges on establishing robust data security frameworks, clarifying legal and ethical boundaries, and ensuring that human care remains central to the patient experience.
As agentic AI and healthcare professionals collaborate more closely, we can anticipate a future where the "AI + Doctor" hybrid model becomes the norm, offering a sustainable, fair, and precise approach to healthcare that benefits all.