Format: 3-Part Instructor-Led Intensive Level: Advanced

I. COURSE DESCRIPTION

In the current AI landscape, the ability to write a good prompt is no longer enough. The next frontier is Agency—building systems that can reason, use software tools, and work together to solve complex business problems. This course is a deep dive into the architecture and assembly of No-Code AI Agents. Participants will move beyond simple linear automations and learn to build Recursive Loops, Multi-Agent Teams, and Self-Correcting Workflows.

Using visual orchestration platforms such as Gumloop, Make, or Relevance AI, you will learn how to connect LLMs to real-world data and APIs, implement sophisticated memory structures, and design human-in-the-loop safety gates. By the end of this course, you will be an AI Architect capable of deploying a digital workforce.

II. KEY COURSE OBJECTIVES AND TAKEAWAYS

• A. ARCHITECTURAL FLUENCY: The ability to look at any manual process and design a no-code agentic system to handle it.
• B. OPERATIONAL SECURITY: Deep knowledge of how to keep agents “in their lane” using guardrails and human oversight.
• C. TOOL AGNOSTICISM: Skills apply to any platform, whether you use Microsoft Copilot Studio, Relevance AI, or Zapier Central.
• D. IMMEDIATE ROI: 30 hands-on labs serve as “starter templates” that can be deployed in a professional environment immediately.
• E. SYSTEMIC REASONING: Transition from simple prompting to building “reasoning engines” that autonomously solve multi-step problems.

MODULE 1: VISUAL LOGIC & THE DECISION ENGINE

• A. DESCRIPTION: Participants will master the transition from linear “if-then” automation to non-linear agentic reasoning. You will learn to architect a “decision engine” that allows an AI to evaluate a complex request, determine the necessary steps, and select the correct path autonomously.
• B. HANDS-ON LABS:
  1. THE ROUTING LOGIC MAP: Build a router that classifies incoming requests into three distinct priority streams.
  2. CONDITIONAL LOOP CONSTRUCTOR: Create an agent that repeats a task until a specific condition is met.
  3. THE INTENT CLASSIFIER: Build a node-based engine that detects “Sales” vs “Support” intent.

MODULE 2: ADVANCED MEMORY & PERSISTENCE

• A. DESCRIPTION: You will learn to overcome the “short-term memory” limitations of standard LLMs by building persistent state systems. This involves structuring external databases so that your agents can reference past interactions and evolving project data across multiple days or weeks.
• B. HANDS-ON LABS: 4. THE USER MEMORY VAULT: Connect an agent to Airtable to store and recall user-specific preferences. 5. MULTI-SESSION CONTEXT TRACKER: Build a system that references a conversation from a week ago during a current interaction. 6. HISTORICAL TREND ANALYZER: Create an agent that compares today’s data entry against the last 30 days of records.

MODULE 3: PROMPT CHAINING & TASK DECOMPOSITION

• A. DESCRIPTION: This module teaches you how to decompose massive, high-stakes business goals into a sequence of atomic, manageable sub-tasks. You will learn to build “chains” where the output of one specialized prompt serves as the filtered input for the next.
• B. HANDS-ON LABS: 7. THE MODULAR GHOSTWRITER: Chain three prompts (Outline -> Draft -> Tone Polish) for consistent long-form content. 8. THE DATA EXTRACTION CHAIN: A sequence that identifies data in text, formats it as JSON, and validates it. 9. THE LOGIC GATEKEEPER: A two-step chain where Agent A generates a solution and Agent B critiques it.

MODULE 4: STATE MANAGEMENT & VARIABLE HANDLING

• A. DESCRIPTION: You will learn the technical “plumbing” of agentic systems: how to move data packets across complex workflows without corruption. You will gain the ability to parse JSON, manage global variables, and maintain “session state” through dozens of processing steps.
• B. HANDS-ON LABS: 10. THE PAYLOAD PASSER: Transfer complex JSON objects between a logic platform and an LLM node. 11. DYNAMIC VARIABLE INJECTOR: Build an agent that pulls real-time variables into its prompt. 12. THE SESSION ID ARCHITECT: Create a unique tracking ID for every agent run to audit data flow.

MODULE 5: TOOL INTEGRATION & THE ACTION LAYER

• A. DESCRIPTION: Participants will learn to give their agents “agency” by connecting them to the live web and internal software suites. You will master “tool-calling,” allowing the agent to decide when to search Google, check a calendar, or post to a communication channel.
• B. HANDS-ON LABS: 13. THE LIVE RESEARCH AGENT: Build an agent that browses the web to find sources for a specific topic. 14. CALENDAR AUTOMATOR: An agent that checks availability and autonomously books a meeting. 15. THE API BRIDGE: Connect an agent to a CRM to update a customer record via natural language.

MODULE 6: THE MODEL CONTEXT PROTOCOL (MCP) & CONNECTORS

• A. DESCRIPTION: You will explore the latest industry standards for secure, standardized data exchange between AI models and local environments. You will learn how to use MCP to allow agents to safely read your local spreadsheets and query private databases.
• B. HANDS-ON LABS: 16. LOCAL FILE NAVIGATOR: Use MCP to allow an agent to read and summarize a folder of local spreadsheets. 17. DATABASE QUERY AGENT: Build a natural language interface for a SQL database. 18. THE CONNECTOR HUB: Sync data between Google Drive, Slack, and Trello via a single agent.

MODULE 7: MULTI-AGENT ORCHESTRATION (THE CREW PATTERN)

• A. DESCRIPTION: This module focuses on the architecture of “digital teams.” You will learn how to design a hierarchical system where a “supervisor” agent delegates specialized tasks to a team of subordinate agents, managing inter-agent communication and conflict resolution.
• B. HANDS-ON LABS: 19. THE EDITORIAL BOARD: A 3-agent team that creates, edits, and fact-checks a blog post. 20. THE DISPUTE RESOLVER: A system where two agents argue different sides of a problem to find an optimal solution. 21. THE PARALLEL TASKER: Orchestrate five agents to perform different research tasks simultaneously.

MODULE 8: AGENTIC RAG & KNOWLEDGE SYNTHESIS

• A. DESCRIPTION: You will move beyond basic document search to “Agentic Retrieval-Augmented Generation.” You will learn to build systems that critique their own search results and perform secondary searches if the first was insufficient to provide a nuanced answer.
• B. HANDS-ON LABS: 22. THE SMART WIKI SEARCHER: A RAG system that “follows up” if the information found is incomplete. 23. THE MULTI-DOC SUMMARIZER: Upload multiple PDFs and have an agent find common themes across them. 24. SEMANTIC CONTENT RE-RANKER: Build a loop that re-ranks results based on the agent’s internal reasoning.

MODULE 9: SELF-CORRECTING WORKFLOWS & ERROR HANDLING

• A. DESCRIPTION: Participants will learn how to build “self-healing” agents that can debug their own failures. You will design loops that catch common errors and prompt the agent to try an alternative strategy, ensuring the task is completed without crashing.
• B. HANDS-ON LABS: 25. THE SELF-HEALING SCRAPER: A web agent that switches strategies if it hits a 404 or a block. 26. THE CODE DEBUGGER: An agent that writes a snippet, tests it, and rewrites it if the test fails. 27. THE RECURSIVE POLISHER: A loop that forces an agent to rewrite output until it hits a specific quality score.

MODULE 10: GOVERNANCE, SAFETY, & HITL

• A. DESCRIPTION: This module addresses the risks of autonomous AI. You will learn how to design and implement “approval gates” and “human-in-the-loop” (HITL) checkpoints to ensure an agent never performs a high-stakes action without explicit authorization.
• B. HANDS-ON LABS: 28. THE REFUND GATEKEEPER: An agent that prepares a refund but sends an “authorize” button to a manager. 29. THE PII SCRUBBING AGENT: A safety node that removes personal data before sending information to a public LLM. 30. THE THOUGHT-LOG AUDITOR: Create a dashboard that exports the “internal monologue” of an agent for review.

MODULE 11: TOKEN MANAGEMENT & COST OPTIMIZATION

• A. DESCRIPTION: You will gain the skills to run an agentic workforce profitably. This involves monitoring token consumption, implementing caching, and strategically switching between large models (e.g., GPT-4o) and smaller models (e.g., GPT-4o-mini) based on task complexity.
• B. REAL-WORLD APPLICATIONS: Analysis of cost-per-run metrics, implementing model-routing logic, and setting up budget-threshold alerts within visual orchestrators.

MODULE 12: DEPLOYMENT & LIFECYCLE MANAGEMENT

• A. DESCRIPTION: In the final module, you will learn how to transition your agents from a sandbox into a production-ready digital workforce. This includes version control, stress testing, and setting up monitoring dashboards to track performance and ROI over time.
• B. REAL-WORLD APPLICATIONS: Hardening builds for live professional use, establishing version control for visual flows, and scaling architectures from single-user to enterprise-wide deployment.