Demo: Medbot, A Practical Tool Built by Clinicians for Clinicians to Leverage AI Agents to Enhance Clinical Practice

Li Lianjie Anthony, Antonio Bandeira, Niroj Bhandari, Rahul Gorijavolu, Aaron Teo Rui Kang, Jasmine Xiaojin Zhang, Alvin Aung Aung Hein, March I-Cheng Chen

Keywords: AI Agent, Sepsis, Bundle, Surviving, Productivity, Automation, AI, Agent, LLM, Demo

TL;DR: This demo paper describes Medbot, an AI Agent platform, designed to help clinicians leverage AI agents to automate manual clinical tasks, reduce adverse outcomes and improve care quality, in complex clinical conditions such as sepsis.

Abstract:

Clinician burnout is a critical issue in modern healthcare, exacerbated by extensive time spent on Electronic Health Records (EHRs), high cognitive load, and numerous manual administrative tasks. While current AI solutions like ambient scribes and Retrieval-Augmented Generation (RAG) systems have shown promise, they do not fully address the complex, labor-intensive workflows inherent in clinical practice. We present Medbot, an agentic AI platform designed by and for clinicians to automate and enhance clinical workflows. Medbot provides a flexible interface for creating, deploying, and evaluating specialized AI agents that can access local datasets, utilize specific tools, and integrate seamlessly with existing EHRs via the FHIR standard. We demonstrate the practical application of Medbot through a real-world use case: an AI agent co-developed with specialists at Tan Tock Seng Hospital, Singapore, to improve the timely initiation and auditing of the sepsis care bundle. This demonstration showcases how Medbot's agentic architecture can directly address clinical needs, reduce administrative burden, and ultimately improve patient outcomes.

For submission under "Topic 1: GenAI Applications and Use Cases in Health" and "Track 2: Demonstration Papers".

Submission Number: 110