Agentic AI Explained: What Hospitals Need to Know

Agentic AI is one of the most talked-about, and least understood, terms in healthcare technology right now. In this Revecore conversation, Angela Troccoli, Head of Marketing at Revecore, sits down with Firoze Lafeer, who leads Revecore's data science team, to break down what an AI agent actually is and what hospital leaders need to think about when building an AI strategy. 

The discussion covers the practical difference between a chat-based AI tool and an agentic system that monitors, detects patterns, and acts on its own; why that distinction matters operationally; and what foundational steps organizations should take before investing in AI internally. 

Watch the Full Conversation

What Is Agentic AI?

Angela Troccoli: We've briefly talked about the different types of AI, including large language models and machine learning. One area receiving a lot of attention is agentic AI. Can you explain what it is, why it matters, and how it differs from the other forms of AI we use every day? 

Firoze Lafeer: It's a confusing topic right now because I'm fairly certain everyone who uses the word "agent" has something different in mind, which is completely expected. 

Our definition is that agentic AI is an AI system that has agency, meaning it does something. That's the difference between a chat tool that answers questions and a system that moves a workflow forward or performs ongoing risk analysis autonomously or semi-autonomously. 

How Is Agentic AI Different from Chat-Based AI?

Angela: I was going to ask whether it's autonomous. Can it be both? 

Firoze: It depends on the use case, but semi-autonomous is a good way to think about it. 

For example, imagine you're working on a Toyota assembly line. With a non-agentic system, you might notice something wrong with a vehicle, go to a chat interface, ask questions, retrieve the manual for that particular car, and determine whether everything is correct. 

An agentic system would continuously monitor all the cars coming off the line. You could build a system of agents to monitor what's happening across the entire line without someone explicitly asking a question. The system could identify cars that appear to have the same issue. If a worker finds a problem with one car, the system might locate the other 300 Toyota Camrys on the line that could have been affected by the same earlier issue. 

Why Agentic AI Matters for Hospital Leaders

Angela: It's doing that without being prompted. I think that's what makes it so different from chat-based AI. It recognizes a pattern and takes the next step. 

Firoze: That's right. ChatGPT made the idea of asking questions and receiving answers extremely popular. Agentic AI shows that you can do much more than that. It brings other technologies into the fold and extends beyond a chat interface. That's a useful general definition of what an agent is. 

When building a strategy around this, it's important to understand that within a year, or perhaps even sooner, saying you use agents may be like saying you use computers. That's wild to consider because the technology is so new and moving so quickly. 

A company wouldn't create one department that uses computer programs while everyone else avoids them. You also wouldn't say that because one vendor uses computer programs, you don't need other vendors that use them. 

An agent is simply something that has some form of agency. You'll have agents within your company, your vendors will have them, and many of the other tools you use will incorporate them as well. That reality is either already here or coming very soon. 

You shouldn't think of agentic AI as one particular technology for a single use case. I think we can accept that this is simply how the world will operate. 

The question now is how agentic AI fits into your broader staffing and vendor strategies. You want to ensure your vendors are using these technologies because they are highly effective. They can help reduce costs and improve the effectiveness of what you're trying to accomplish, including increasing the number of claims that receive additional payment. Success depends on working with people who understand how to apply these capabilities. 

What Hospitals Need Before Implementing AI

Angela: On that same note, many organizations that are not traditionally technology companies are getting excited about using AI and investing in their own internal AI strategies. I think that's wonderful, but what foundational elements should they have in place to set themselves up for success? What does it take to successfully launch an AI strategy? 

Firoze: First and foremost, organizations need to understand what AI is and what it isn't. They also need to ensure they're receiving guidance from people who understand those distinctions. 

Second, you need to introduce your organization to the tools. Practical experience is extremely helpful because it allows the people doing the work to imagine how they could perform their jobs differently using those tools. 

But that can't be the only component. You also need to bring in people who have experience with the technology. That's how technology projects succeed. 

Organizations also need to consider how they collect information and protect it. In our work, data safety is especially important. You need to collect information securely, using the right methods and organizational structure, so your machine learning and AI techniques are useful and effective. 

Data governance, data architecture, and engineering disciplines are all critical. That's the difference between experimenting with AI in your garage or using ChatGPT at home to plan a vacation and applying AI to high-stakes business processes. 

Deciding whether to send letters to payers, determining whether a claim was paid correctly, or pursuing a complicated case requires a different standard. 

One potential blind spot is focusing on the science and AI while overlooking the engineering required to operate at scale. As with computers and the internet, scaling technology requires strong engineering support. That must be part of the strategy. 

Angela: I love it. Firoze, thank you so much. I appreciate the conversation. We'll have more questions and answers about AI coming soon. Thank you again for your time. 

Firoze: Thanks, Angela.  

See How AI-Powered Recovery Performs at Scale 

Download Faster Revenue Recovery. Greater Precision. to see how Revecore's ReClaim™ platform pairs AI with expert review to drive results like a 79% reduction in average days to bill and a 6–7% year-over-year uplift in underpayment collections. 

Download the Whitepaper