Introduction
- Understanding Agentic AI concepts and healthcare automation challenges
- Patient Utilization Review use case overview
- AI agents vs traditional integrations: adaptive reasoning vs fixed workflows
- Workshop objectives and technology stack
Agentic AI Design Patterns
- Understanding how integrations become tools for AI agents
- Exploring thinking patterns: ReAct and Plan and Execute
- Model Context Protocol (MCP) for tool discoverability
- Agent orchestration patterns and best practices
Import an OIC Project
- Importing the pre-built Patient Care Utilization Review project
- Exploring five healthcare integrations and decision tables
- Configuring connections: REST, FTP, and OpenAI LLM adapters
- Activating integrations and validating project setup
Register Integration as Tool
- Understanding agentic AI tool requirements
- Registering five healthcare integrations as tools:
- Fetch Patient Record
- Match Clinical Guideline
- Check Guideline Validity
- Escalation Decision
- Recommend Care Plan
- Configuring tool identifiers, descriptions, and parameters
- Setting guidelines for LLM interpretation
Discover Integrations as Tools from MCP Client
- Enabling MCP server capabilities in the project
- Understanding MCP protocol for external tool discovery
- Testing tool discovery from MCP clients (Postman, Langflow)
- Verifying tool metadata and availability
Build OIC AI Agent
- Creating the Clinical Quality Assessment Agent
- Configuring AI thinking patterns and LLM connections
- Adding and orchestrating multiple agentic AI tools
- Defining agent role, guidelines, and behavior parameters
- Setting up agent guardrails and execution controls
Run and Test the AI Agent
- Executing the agent with sample utilization review cases
- Monitoring agent reasoning and tool invocation sequences
- Analyzing agent activity streams and decision-making
- Validating audit reports and quality assessments
- Understanding agent observability and troubleshooting