AI in Clinical Medicine, ISSN 2819-7437 online, Open Access
Article copyright, the authors; Journal compilation copyright, AI Clin Med and Elmer Press Inc
Journal website https://aicm.elmerpub.com

Review

Volume 2, August 2026, e35

Optimizing the Use of AI Agents in Electronic Health Record Workflows in Healthcare: Clinical Integration and Human Factors Review

Figures

↓  Figure 1. Multimodal data fusion architecture with AI agents (Microsoft Copilot, ChatGPT, & Clinical Chatbots).
Figure 1.
↓  Figure 2. PRISMA 2020 flow diagram illustrating identification, screening, eligibility, and inclusion of studies evaluating AI agents (ambient documentation systems, Microsoft Copilot, ChatGPT, and clinical chatbots) in clinical workflows.
Figure 2.
↓  Figure 3. A four-layer concentric framework for healthcare AI agents, with Data & Context at the core, surrounded by Agent Reasoning & Action, Human-AI Interaction & Verification (featuring the three HFE patterns: Visual Provenance, Progressive Disclosure, and Attestation Interlocks), and Organizational Learning & Governance at the outermost layer.
Figure 3.

Tables

↓  Table 1. Technical and Operational Comparison of Clinical AI Modalities
 
Feature/dimensionAmbient Scribes (Abridge/DAX)Microsoft Copilot for HealthcareChatGPT (OpenAI Enterprise)Specialized clinician chatbots
Primary clinical focusPassive ambient consultation recording & SOAP note generation.EHR chart summarization, inbox draft synthesis, & Microsoft 365 workflow automation.General conversational reasoning, patient letter generation, & complex differential query support.Interactive chart QA, clinical protocol retrieval, & real-time decision support.
Interaction mechanismBackground microphone audio capture (hands-free).Sidebar EHR pane & embedded native buttons.Conversational chat window/API integration.Interactive chat interface embedded in EHR.
HIPAA/data privacyBAA compliant, audio processed in HIPAA enclave.Enterprise HIPAA compliance via Azure Health Data Services.Enterprise BAA required; zero data retention for training.Dedicated HIPAA-compliant clinical infrastructure.
Primary strengthsEliminates manual typing during patient encounters.Seamless deep integration with Epic, Cerner, and Office 365.Superior complex reasoning, multi-language translation, & drafting flexibility.High precision on domain-specific medical guidelines & literature.
Key HFE & Usability risksAcoustic distortion in noisy rooms; verification fatigue.Context switching if integrated across multiple sidebar tabs.Risk of ungrounded hallucinations if prompt isn’t tied to RAG.Alert fatigue and chat friction if answers are verbose.

 

↓  Table 2. Operational Workflow Transformation Matrix Across EHR Dimensions
 
Workflow dimensionLegacy EHR workflowAI agent & Copilot workflow
DocumentationManual typing/dictation during or post-encounter.Real-time ambient capture with DAX/MS Copilot notes drafting.
Chart review5–15 min spent navigating fragmented tabs.Conversational clinical chatbot queries and single-page LLM summaries.
In-basket triageFirst-in, first-out manual review of all messages.Automated urgency routing + MS Copilot generative draft response.
Safety alertsHigh-volume, static pop-ups leading to alert fatigue.Targeted, risk-calculated predictive CDS interventions.
Prior auth/codingLabor-intensive manual chart data extraction.Automated MS Copilot evidence extraction matched to payer rules.

 

↓  Table 3. Quantitative Workflow and Efficiency Impact Metrics
 
Performance metricPre-AI baselinePost-AI agent implementationOperational shift (%)
Documentation time (per patient)10–12 min2–4 min75% reduction
Daily administrative time (per FTE)2.5 h1.0 hour60% reduction
After-hours (weekly)8–10 h2–3 h70% reduction