8th Global Gastroenterology and Artificial Intelligence Summit Separating AI Hype from Clinical Reality | September 2026-4 - Kowalcyzk _ Embedding AI into the EHR and Endoscopy Workflow

4 - Kowalcyzk _ Embedding AI into the EHR and Endoscopy Workflow

Pdf Summary

The presentation argues that AI in gastroenterology only succeeds when it is designed around clinical workflow, not just model accuracy. The key message is: “Start with the workflow, not the algorithm,” because adoption depends on how and where the tool is delivered.<br /><br />It emphasizes three constraints of the endoscopy setting: the technology must fit the physical room, operate with clinically acceptable latency, and preserve endoscopist control. The system should capture intentionally rather than passively, provide guidance within the same second, and allow the clinician to remain in control without interrupting the procedure.<br /><br />The talk also distinguishes between different procedure types. EGD and colonoscopy require fast, real-time detection and structured reporting, while advanced procedures such as ERCP, EUS, and EMR need more documentation-focused support, including device details, sizing, staging, and post-procedure synthesis. One workflow will not work for all cases.<br /><br />A central point is that AI should include a human-in-the-loop oversight layer. Experts should monitor outputs, correct errors, and update models so the system improves safely and continuously.<br /><br />Finally, the speaker stresses that AI should be embedded into the EHR and connected through open integration, including FHIR-enabled data exchange. The EHR should serve as a central documentation hub, with the flexibility to bring in different models or solutions. Overall, AI earns its place by disappearing into the workflow while remaining expert-monitored, EHR-connected, and procedure-aware.

Keywords

artificial intelligence

gastroenterology

clinical workflow

endoscopy

real-time detection

human-in-the-loop

electronic health record

FHIR integration

procedure-aware AI

endoscopy reporting