The Agents Are Coming: The Agentic Platform

Agentic platform architecture and design patterns aren’t what you think. I helped a hospital system use AI to better manage its ER. The ER experienced patient surges where the wait times would rise, sometimes exceeding an hour or more. When that happens, patient outcomes degrade significantly.

I can reframe this as a systems problem. The hospital’s stock of patients exceeds its stock of staffing or treatment capacity, causing a loss in patient outcomes.

Stocks are resources like patients and staff or your phone’s charge level, water in a glass, and funds in a bank account. Thinking in Systems by Donella H. Meadows is an excellent introduction to systems thinking. I have used systems thinking to reframe problems so they can be solved with technology for almost 20 years.

There’s a physical solution to this problem that doesn’t require technology. Increase staffing stock to a level that would handle all surges. We bump into new stock constraints by evaluating this response. There is a talent shortage, so there may not be enough people in the market to implement this solution. Even if feasible, costs exceed the hospital’s ability to pay.

We’re not just optimizing the system for patient outcomes. Any solution must optimize patient outcomes and staffing levels. The hospital wants to have as close to the exact stock of staff to match its stock of patients.

Why Use AI? The Limits Of Digital Platforms

Without technology, the hospital must wait until the surge happens before responding by increasing staffing levels. People and businesses often intervene in systems to ensure an outcome. We use feedback from the system to understand its current conditions. Patient wait times or the number of patients are feedback points of the ER system.

Waiting for physical feedback causes a lag. Patients overwhelm the ER, and the staff have more work to do than they can manage. Someone finally breaks free to call up to an administrator for extra staffing. The administrator calls around to other units to see who has spare staffing stock and sends down new resources. They call more people to come into work if necessary.

There’s a lag between when the patient surge begins and when the stock of staff rises to the level necessary to manage it. During that lag time, patient outcomes degrade; if we can reduce the lag, we can reduce the loss in patient outcomes. There are two categories of technical solutions that can be deployed: digital and agentic.