State Representation for Agents Navigating Stateful Clinical Applications
Agents need structured state, not screenshots, to navigate clinical software safely.
Tomás Ibáñez-Reyes
Contributing Editor, Deployment & Operations
Tomás Ibáñez-Reyes
Tomás has covered health IT infrastructure and enterprise deployment strategy for over a decade, with earlier stints as a systems administrator for a multi-site hospital network informing his ground-level perspective. His reporting focuses on the operational realities of rolling out agent-based tooling in clinical environments where downtime carries direct patient risk.
6 stories
Tool Calling vs Pure Vision in Computer-Use Agents
Hybrid agents need both approaches—each covers failures the other can't fix alone.
Tomás Ibáñez-Reyes
Modal Dialog and Interrupt Handling in EHR Workflows
Computer-use agents must detect and classify modal interrupts to survive real EHR workflows.
Tomás Ibáñez-Reyes
Session Timeout Detection and Recovery in Clinical Software
AI agents need built-in timeout detection to prevent silent session failures in clinical software.
Tomás Ibáñez-Reyes
Deterministic vs. Probabilistic Planning in Agents
Deterministic systems prove what happened; probabilistic ones report what they think happened.
Tomás Ibáñez-Reyes
Vision Model Selection for Clinical UI Parsing
Most vision models fail below 9% accuracy on real clinical software systems.
Tomás Ibáñez-Reyes