Hey Reader,

Introduction

Now that we have transitioned from 2024's "year of AI" into what experts are calling 2025's "year of AI Agents,". We are seeing AI as the interface with more and more systems. But, is it time to look beyond AI Agents to the environments where these agents operate. While AI Agents might capture headlines, the real transformation lies in how we design intelligent systems that blend human expertise with targeted automation. This is done with "Agentic Design". Healthcare organizations stand at a juncture where understanding agentic design principles will determine success in operational efficiency and patient care.

TLDR

Agentic design provides a framework to model healthcare workflows as perceive-think-act cycles, enabling strategic automation that enhances rather than replaces human capabilities.

Key Concepts Defined

Large Language Models (LLMs): AI systems trained on vast text datasets that can understand and generate human-like text through natural language interactions.

Hallucinations: Incorrect or fabricated outputs from AI systems that appear plausible but lack factual basis.

​

Prompt Engineering: The discipline of crafting precise instructions (prompts) for LLMs to minimize errors and improve output quality.

​

AI Agents: Constrained AI systems designed to achieve specific goals and perform defined tasks with some degree of autonomy.

​

Agentic Design: A framework for modeling problem-solving processes as agent-based cycles (perceive-think-act), applicable with or without AI technology.

Digital Twin: A virtual representation of physical systems, processes, and environments that serves as the operational space for both human and automated agents.

Detailed Explanation

Beyond the AI Hype Cycle

While 2024 introduced healthcare to conversational AI through tools like ChatGPT, Claude, and Gemini, 2025 will be about practical implementation through AI agents. The distinction is crucial: unlike general-purpose LLMs that can answer any question (accurately or not), agents are purpose-built for specific tasks within defined parameters.

The Security Challenge

Cloud-based AI platforms present security concerns when handling sensitive healthcare information. Uploading confidential operational strategies or patient data to external AI services risks breaches, regardless of the efficiency gains. This reality has slowed adoption in healthcare settings where data protection is paramount. Agentic design can help by ensuring that Agents operate on a "need to know" basis. AI is best applied to areas where there is uncertainty, but if we remove the uncertainty, AI is not necessary. Removing all uncertainty is impossible so AI always has some utility.

The Agent Paradigm

What makes the agentic approach transformative is its simplistic, cyclic process: sense and perceive (gather information), think and decide (based on events and states), and act toward an outcome (generating new states). This mirrors human decision-making but can be distributed across both technological and human components.

The brilliance of agentic design lies in separating the agent concept from AI technology. An agent can be entirely human, entirely automated, or—most effectively—a hybrid system leveraging the strengths of each. This holistic view allows healthcare organizations to model existing workflows, identify automation opportunities, and maintain appropriate human oversight.

The Digital Twin Foundation

For healthcare operations, the digital twin serves as the environment where agents function. It provides the situational awareness necessary for effective decision-making and creates a sandbox for testing automation before implementation. When viewed through this lens, technologies like Real-Time Location Systems (RTLS) become the sensory organs of your operational environment rather than isolated point solutions.

Practical Application: Loss Prevention

Consider a hospital's asset tracking workflow: An IV pump leaves the building, triggering an RTLS alert. A security officer reviews camera footage, recognizes the individual as a vendor, and confirms with Central Equipment that the pump is scheduled for maintenance. The officer cancels the alert, resolving what was simply a documentation oversight.

In this scenario:

• Sensing and Perception occurs through RTLS sensors and security cameras
• Thinking and Deciding happens primarily through the security officer's situational awareness and contextual knowledge
• Action is the cancellation of the alert and outcome is the incident is neutralized

By modeling this process through agentic design, we can identify opportunities for improvement without immediately jumping to AI solutions. This structure makes it easier to see where automation can be most effectively applied. For instance, perhaps an additional sensor at Central Equipment could trigger a reminder to document outbound equipment. Or maybe computer vision could recognize vendor uniforms to prioritize alerts differently.

Multi-Agent Systems: The Future State

As more workflows are modeled and optimized, organizations can develop multi-agent systems (MAS) where specialized agents handle specific aspects of operations while communicating with each other. A monitoring agent might manage alerts, while a utility agent interfaces with the maintenance management system. These interacting agents, orchestrated toward common goals, create an intelligent operational ecosystem—where AI becomes essential to managing complexity.

The Human-in-the-Loop

Critical to agentic design is acknowledging human cognitive load. By quantifying the mental demands of workflows, we can identify not just where automation adds value, but where it might introduce new challenges like alert fatigue. The goal isn't to remove humans from the equation but to enhance their capabilities through strategic automation.

Key Takeaways

• Agentic design provides a framework to analyze healthcare workflows independent of AI implementation, creating a pathway to strategic automation.
• Digital twins serve as the environment for agent operations, making technologies like RTLS more valuable as components of an integrated system.
• Effective healthcare solutions will combine AI and human elements, leveraging the strengths of each while mitigating limitations.
• Modeling existing workflows as agents reveals immediate improvement opportunities while establishing the foundation for future AI integration.
• Multi-agent systems represent the future of healthcare operations, with specialized agents handling different aspects of complex workflows.

Conclusion

As we embrace 2025 as the "year of agents," healthcare organizations have a unique opportunity to move beyond AI hype and into strategic implementation. By applying agentic design principles to existing workflows, hospitals can take control of their operational AI strategy while building the backbone of their digital twin. This approach reveals both immediate improvement opportunities and long-term strategic direction without requiring deep AI expertise.

Ready to explore how agentic design could transform your healthcare operations? Connect with us to learn more about implementing these principles in your organization.

Until next week,

Paul E Zieske ​Location Based Services Consulting ​​​