Part 1: Stop waiting for a healthcare workforce miracle—embrace Agentic AI now

Health systems and health plans must decide now: do you automate care delivery or accept system collapse?

The US currently faces a shortage of 55,000 doctors, while the prevalence of chronic conditions is reaching historic highs. We are also only a few short years from having more seniors than young-uns for the first time in the nation’s history (see Exhibit 1). This intersection of healthcare demand and supply is headed toward an unprecedented public health catastrophe that will make the pandemic look like a zit on a teenager. However, agentic AI (AI systems to deliver specific outcomes with limited human direction) will bridge the supply gaps to address the aging and sick, potentially more consistently, empathetically, and less expensively. So, enterprise health leaders can no longer hire their way out of systemic clinical shortages—agentic AI is the only viable scale lever.

Exhibit 1: The US is growing older and sicker faster than any other place on this planet and needs a different healthcare paradigm

Source: Centers for Disease Control and Prevention (CDC), US Census Bureau, HFS Research, 2025

The glide path to the acceptance of Dr. Agentic AI remains fluid

Health system and plan leaders (CEO, CFO, strategy) must incorporate agentic AI in their overall business reimagination construct. They must recognize that agentic will be the answer to overcoming the constraints under which they have played so far. Consider that agentic AI can be categorized into four types:

• Task-based agentic: The ability to complete specific tasks beyond just completing a process, e.g., prior authorization, patient intake
• Multi-agentic: The ability to execute multiple tasks at an enhanced level of cognition, e.g., identifying a molecule followed by additional simulations, testing different variations, and refining predictions
• Clinical agentic: The ability to predict clinical outcomes or deliver proactive care based on lifestyle, genetics, medical data, etc., e.g., clinical intervention before symptoms manifest and ongoing chronic care management
• Autonomous agentic: The orchestration of independent actions with limited to no human intervention to deliver specific outcomes, e.g., continuous follow-up and interventions in a post-acute care setting

The quadruple aim of care (reduction of costs, enhancing the experience of care, improving health outcomes, and addressing health equities) is a logical and reasonable way to measure the performance of the healthcare ecosystem (funding, cure, and care delivery). Enterprise leaders must reframe the quadruple aim as a business case for agentic investment—not just a clinical framework (see Exhibit 2). This thesis is based on evidence that shows consistent accuracy (radiology detections of cancer), higher empathy (younger generations prefer chatbots over human clinicians), and reduced dependence on expensive human clinicians. While the data is preliminary and is continuously refreshed, it directionally validates the enormous potential for agentic in healthcare to address the quadruple aim of care.

Exhibit 2: Agentic AI will give us the option to either augment or replace human clinicians

Source: HFS Research, 2025

Technology (diagnostics, EMR/EHR, wearables) and machines (imaging, monitors, ventilators) are ubiquitous in healthcare. Over the last fifty years, their adoption has set a precedent for adopting agentic AI. However, there is an adoption curve that will be driven by generation (younger clinicians and patients are likely to embrace faster), data (inaccurate or inequitable data will reduce efficacy and trust), and regulations (burdensome regulations could slow adoption and make it expensive). The economics will drive the angle of that adoption curve. If agentic AI can improve the financial health of health systems, hospitals, and consumers, there is no reason not to adopt or delay—the safety factors will sort themselves out as the AI models will improve faster than humans upgrade their clinical knowledge.

The lack of a 2026 agentic deployment roadmap today will mean you will be uncompetitive by 2027

Staffing challenges (clinical and non-clinical), the complexity of regulations and processes, and the increased cost of care are pushing health plans and health systems toward adopting more technologies, particularly AI. That is reflected in the current state of healthcare, which is marked by AI enablement to automate some processes and low-risk communications with augmented content.

Health plans use generative AI to summarize member service calls and for approvals of low-cost prior authorizations. Health systems have been piloting ambient listening in exam rooms and incorporating intuitive workflows in their EHRs. However, there is no evidence of at-scale adoption beyond proof of concepts and some power usage in the context of pilots.

That said, as AI models refine and incorporate expanded tranches of data into their training, we will see higher levels of task-based agentic AI adoption in 2026. Exhibit 3 illustrates some tasks and activities that agentic could be deployed to address across health plans and health systems. The task-based agentic will likely be biased toward the cost attribute of the quad-aim. It will not be a catalyst for reimagination; instead, the everyday blocking and tackling can be done faster and possibly cheaper (the cost of agentic is yet to be determined).

Exhibit 3: Task-based agentic is a low-risk effort to validate its potential to improve operations

Source: HFS Research, 2025

Agentic is real and is being applied by health plans and systems around the world

If you’re confusing automation for agentic, you’re likely underinvesting in cognitive capabilities. Time and sustained investments will bring clarity as agentic will drive execution vs. process compliance. There is already evidence of that behavior (see Exhibit 4) as health systems and plans continue experimenting. The task-based agentic has saved time and money, which is encouraging. However, more work is required to understand the net savings, given that the cost of agentic development and usage (compute) is unclear. Still, progress must be seen in the context of validating the potential.

Exhibit 4: Agentic is more than just promise—tangible, quantifiable outcomes must strengthen enterprise leadership convictions to operationalize broadly

Source: HFS Research, 2025

Multi-agentic is also promising as it tackles complex process challenges with fraud, waste, and abuse that can often contribute up to 25% of the cost of care. Hospital operations are also seeing the benefits with discharge and surgery planning. These early examples validate the potential.

More exciting is the progress being seen with clinical agentic, as the early outcomes suggest a reduction in mortality, missed diagnoses, and hospital lengths of stay. It also shows promise in diagnosing mental health and potentially will intrinsically address equity by reaching across economic and social biases. 

The Bottom Line: Agentic AI is the only viable strategy to solve the quadruple aim of care, and enterprise leaders must stop piloting and start rapidly operationalizing—before spiraling costs and workforce shortages force their hand.

Agentic has shown lots of promise in non-scaled production with outcomes headed in the right direction. Enterprise leaders must move beyond pilots and fund agentic AI to operational scale—before their workforce gaps become existential. At the same time, there are still many questions to answer, including the cost of computing, the accuracy of the models, the biases, and consumer and clinician adoption. These questions will find answers over time and should not hinder continued adoption. Our subsequent perspectives will address  the other three evolutionary steps along the agentic glide path.