Description:

Artificial intelligence is rapidly evolving beyond chatbots and single prompts. A new generation of systems - known as AI Agents - can pursue goals, use tools, maintain memory, and execute multi-step workflows with minimal human intervention.

This 90-minute session introduces the core architecture and design principles behind agentic AI systems. Participants will explore how modern agents plan tasks, interact with external tools, and adapt their behavior based on feedback. The session provides a technology-agnostic blueprint that applies across both developer frameworks (such as LangGraph and multi-agent frameworks) and no-code orchestration tools like n8n. Rather than focusing on a single platform, the workshop emphasizes conceptual understanding and system design - equipping participants with the knowledge needed to evaluate, design, and deploy AI agents that can meaningfully assist with real-world work.

By the end of the session, participants will understand how to move from prompt-based interactions to goal-driven AI systems capable of completing complex tasks.

Duration:

90mins

Course Code: BDT 543

Learning Objectives:

After this course, you will be able to:

1. Differentiate between Chatbots and Agents: Understand the leap from "Reactive" to "Proactive" AI
2. Identify Agentic Components: Deconstruct an agent into its Brain (LLM), Tools (APIs), and Memory
3. Master Goal-Based Delegation: Learn to write "Objectives" rather than "Instructions."

Business professionals, entrepreneurs, product managers, developers, and technology enthusiasts who want to understand how AI systems are evolving from simple chat interfaces into autonomous digital collaborators.

Basic familiarity with AI concepts such as prompting is recommended. Prior exposure to workflow thinking or visual automation tools is helpful but not required.

Course Outline:

1. Beyond the Chatbot: What is an AI Agent?
   1. The Autonomy Scale: From scripted bots to "Agentic" systems that plan their own steps
   2. The Reasoning Loop: How agents "Think-Act-Observe" (The ReAct pattern) to solve multi-step problems
   3. Real-World Impact: From answering a refund FAQ to processing the refund across three different software systems

2. The Anatomy of an Agent (The Universal Blueprint)
   1. The Brain (LLM): Choosing the right model for reasoning vs. speed
   2. The Toolbox (Capabilities): Giving the agent "hands" (e.g., Web search, database access, email, or custom internal tools)
   3. Memory Systems - Short-term: The current conversation context, Long-term: Remembering user preferences and past project outcomes

3. Agentic Orchestration: Single vs. Multi-Agent Systems
   1. The Specialized Worker: Building a single agent for a specific role (e.g., The Researcher)
   2. The Digital Dream-Team: How "Multi-Agent" frameworks (like CrewAI or AutoGen) allow different agents to talk to each other, debate, and peer-review work
   3. Decision Nodes: When to use a fixed flow vs. when to let the agent decide the next move

4. Safety, Ethics, and the "Human-in-the-Loop"
   1. The "Kill Switch": Designing guardrails and resource budgets to prevent "infinite loops" and high API bills
   2. Verification Layers: Why agents need a human "Supervisor" for high-stakes actions (e.g., moving money or deleting data)
   3. The Future of Work: Transitioning from "Doing the work" to "Managing the fleet."

Training material provided: Yes (Digital format)