Autonomous Trained Agents
Overview
Sully AI is actively developing autonomous trained agents designed to streamline healthcare workflows by automating complex, multi-step tasks traditionally handled by clinical staff. While full autonomy is not yet deployed in production, our current architecture includes foundational components such as real-time medical scribing, task triage, and decision support—each of which contributes to agent specialization over time. These agents are built to adapt to role-specific workflows (e.g., receptionist, medical assistant, care coordinator) and are architected to incorporate behavioral patterns and institutional routines into their task execution logic.
Importance
Autonomous trained agents combine structured learning, contextual memory, and task-specific reasoning to simulate human decision-making in healthcare workflows. These agents ingest unstructured input, such as patient-provider conversations or incoming EHR data, and convert it into structured actions by referencing known medical ontologies, prior agent memory, and observed provider preferences. Over time, agents refine their behavior by learning from task feedback, repeat interactions, and localized workflow patterns, allowing them to make more relevant and efficient decisions without constant human prompting.
Current Capabilities
Transcribe and summarize clinical encounters with >98% accuracy.
Extract structured elements (labs, meds, diagnoses) in real-time to support downstream task automation.
Classifies incoming information by intent and clinical urgency, routing tasks to the appropriate internal tool or staff member.
Supports modular agent chaining based on observed workflows and provider role preferences.
Provide differential suggestions, order prompts, and referral insights based on context and known guidelines.
Operate within scoped environments with configurable guardrails and auditing for transparency.
Agents continuously observe provider decisions across workflows and use structured feedback to adjust responses.
Over time, agents can apply known provider routines (e.g., referral habits, phrasing preferences, common differentials) to new patient contexts to reduce friction and increase relevance.
Agents produce standardized outputs such as referral summaries, follow-up scheduling recommendations, and draft orders.
These outputs are tied to intent signals derived from conversation data and historical patterns.
Implementation Workflow: Autonomous Trained Agents
Role-Based Activation
Agents are enabled based on clinical roles (e.g., scribe, coordinator, receptionist).
Context Setup
Agents are configured with access to transcripts, visit history, and provider preferences.
Optional EHR and scheduling integrations enhance context.
Task Detection & Triggering
Agents monitor real-time input and trigger actions based on keywords, clinical patterns, or workflows.
Reasoning & Output
Agents generate structured outputs—notes, orders, referrals—tailored to context and role.
Feedback Integration
Provider feedback is logged and used to refine agent behavior over time.
Memory & Personalization
Agents adapt to individual provider routines while maintaining strict data isolation.
Audit & Oversight
All actions are logged and reviewable for transparency and compliance.
Future Integration
Autonomous trained agents are a natural evolution of Sully AI’s current infrastructure. With role-based orchestration, custom in-house LLM scaffolding, and secure provider-specific memory, we are positioned to support agents that learn from local behavior while remaining compliant and auditable. These agents will unlock scalable, safe task delegation across the care journey—beginning with high-trust domains and eventually extending into more dynamic, multi-agent collaborations.
Last updated