What are the most pressing operational challenges health systems are facing today?

Increasing costs and decreasing reimbursements — often because rising labor costs and reimbursement shortfalls from Medicare and Medicaid — are the top customer concerns across the board.

The complexity of resource management is another major challenge. As demand grows, hospitals struggle with creating scalable capacity. Managing patient flow, bed availability, and care coordination at scale are increasingly difficult with workforce shortages and burnout. As burnout drives provider and staff resignations, health systems become more reliant on contract workers.

AI works by transforming how these teams manage resources, predict, and coordinate care. Hospitals can use AI's predictive analytics tools to study historical and real-time data to forecast ED arrivals, patient volumes, admission volumes, and discharge patterns.

That means hospitals can anticipate bed demands, optimize capacity, and coordinate patient discharges. And because patient volumes can vary by season, AI tools can help hospitals detect patterns across the year and plan for surges in patient visits.

Using AI to track patient flow in real-time can help health systems anticipate these patterns to identify and prevent operational bottlenecks. We can then send alerts and recommended actions to the right person and escalate the issue if it isn't resolved.

Finally, AI agents can be used to automate operational tasks — identifying patients who need bed assignments, transport, orders, or discharges. Those agents can also automate and notify staff about administrative tasks that need to be executed, allowing staff to focus on patient care and complex issues.

What's the difference between an AI that is more informational or makes recommendations and an AI that is more autonomous?

AI agents are not only providing recommendations but are also acting. It's the difference between the AI making staff members aware of what the issue is and how to resolve it, and the AI actually executing the administrative task — freeing the staff to do other professional work.

An AI agent's actions are based on pre-defined standard operating procedures, so they can take action without human intervention.

A good example is bed placement. A patient comes into the ED and needs an inpatient bed, but the operational process is complex: Not only do staff need to assign them a bed, they also need to assign a transporter to move them to that bed. Staff also needs to ensure that the room is cleaned. Autonomous agents can work on staff's behalf to find the right bed with the right level of care and complete the other administrative tasks — so staff doesn't need to do them manually.

At the same time, human interaction with AI agents is very important here. AI agents are really focused on following guidance and directions, but when they can't resolve an issue or need more guidance, AI agents escalate the issue to a person. For example, if there are no beds, the AI agent needs to ask a human to figure out how to find a better solution.

This is where it starts to get interesting. Now that I — as a provider or staff member — have solved a problem, how do I go back and train the AI agent to learn how the human has solved the issue, so it can act based on what it's learned? That's the end goal for AI agents.

What risks are involved with implementing AI solutions, and how can health systems ensure they innovate safely?

The big risks are really about access to protected health information (PHI). The AI shouldn't have full direct access to EHR data. To ensure patient privacy, health systems should regularly audit AI tool's data access.

Even though the AI that accesses data is not a person, health systems still need to maintain HIPAA compliance: Systems should still track which information an AI tool can see and restrict it from sensitive patient data. For example, through Care Logistics' system, the AI doesn't have direct access to the EHR, so it can't compromise PHI.

And while certain decisions can't be made by AI agents alone, they need human guidance and approval. Built-in safeguards should be present so AI agents are trained to escalate issues to staff members, and that requires planning.

To do that in our system, we work with each health system to build standard operating procedures (SOPs) that can also be used as guardrails for AI agents. That is, we use custom SOPs to train AI agents in how to do their tasks, when to make recommendations, and when to ask a human for guidance.

I think mature AI-supported solutions will speed up and sustain the return-on-investment (ROI) so hospitals and health systems can continue to provide their communities with access to quality care delivery.

What do you think most people get wrong about AI and healthcare?

Many people expect AI to target a specific problem without solving the underlying issue. For example, we design our operational model first, before integrating the AI. The AI is really there to help execute the model well and optimize the entire patient journey, not solve an isolated issue.

The other piece is that AI tools don’t replace people, but do make people more effective. Humans still need to be involved in complex decision-making, because they understand patient care — and other humans — much better. Because people will always have more data points available to them than AI, they will make better decisions. Simple and routine tasks can be handled by AI.

What does the future hold for solving operational problems in this space?

I think there are several things to look forward to. First, AI will continue to take over more operational tasks.

More and more, staff will be able to focus on patient care and less on operational issues.

Second, AI tools may help shift patients more to outpatient and home-based care, because it can be used to track, identify, and manage patients outside of the acute care space. Part of that will be identifying the patients who don't need an inpatient bed in the hospital, based on what we know about the patient. That may help providers and staff move those patients to outpatient or home-based care. That will allow hospitals to have more capacity — which, in turn, will help reduce burnout and optimize the workforce. 

On the next level of development, AI could help staff manage a lot of patients recovering in home-based care. AI could potentially help a smaller number of medical staff manage a larger number of patients — often on a chronic illness care or similar plan — on a remote monitoring system. That can also help with capacity, because if you manage patients well outside of the hospital, they stay healthy and don’t need to be readmitted to the hospital.

Care Logistics^® helps health system leaders who are frustrated with their organization’s constant operational challenges. Our proven operational model simplifies healthcare operations and provides a foundation for improved performance. The supporting CareEdge™ technology delivers a complete view of your operations, enabling leaders to identify patterns and trends that would otherwise be invisible and make more informed decisions about patient care and resource management. Real-time and predictive insights with recommended actions empower you to anticipate and address emerging threats and opportunities quickly and effectively. The result? A solid return on investment and efficient operations that help you deliver the best patient care possible.

Learn more about Care Logistics

This article is sponsored by Care Logistics^®, an Advisory Board member organization. Representatives of Care Logistics helped select the topics and issues addressed. Advisory Board experts wrote the content, maintained final editorial approval, and conducted the underlying research independently and objectively. Advisory Board does not endorse any company, organization, product or brand mentioned herein.

To learn more, view our editorial guidelines.