Introduction to Agentic AI

Course Duration

2 Days

Audience

Employees of federal, state and local governments; and businesses working with the government.

Prerequisites

No prerequisites required.

Course Description

This course introduces the principles and practices of Agentic AI, a form of intelligent systems that moves beyond the limits of chat-based AI or productivity assistants. While familiar tools like ChatGPT or Microsoft Office’s Copilot respond to user prompts in the moment, Agentic AI systems are designed to act more autonomously. They can plan steps toward a goal, use tools and data, remember past interactions, and continue working even when not being directly prompted. Across two days, the course introduces the foundations of agent-based systems, showing how they differ from predictive or reactive AI approaches. It explains the core architectures that drive agent behavior, including cycles of perception, planning, action, and reflection, and highlights how prompts and memory shape their decision-making. We will also looks at more advanced setups, such as groups of agents working together, and connects these ideas to real-world applications like research automation, customer support, and business workflows.

Learning Objectives

• Explain what Agentic AI is and how it differs from traditional, predictive, and chat-based AI.
• Describe the core architectures and memory systems that drive agent behavior.
• Recognize how prompts and goal-setting influence agent planning and decision-making.
• Understand how agents make choices, adapt through feedback, and handle uncertainty.
• Identify the features of multi-agent systems and how agents coordinate with each other.
• Relate Agentic AI concepts to real-world application.
• Discuss ethical risks and apply responsible design principles for safe deployment.

Course Outline

1. Foundations of Agentic AI

• Define Agentic AI and its traits: autonomy, persistence, goal-directedness.
• Differentiate Agentic AI from predictive, reactive, and traditional AI.
• Trace the historical path from symbolic AI to agents.
• Identify agent types and the agent-environment loop.
• Distinguish single-agent vs. multi-agent systems.

2. Prompting for Agentic AI

• Write prompts that enable planning and tool use.
• Create templates for dynamic goal-setting and reasoning.
• Use prompts that reference memory or past state.
• Simulate decision chains and tool choices through prompting.

3. Architectures of Agentic Systems

• Recognize the perception → planning → action → reflection cycle.
• Describe short-term, long-term, and episodic memory.
• Explain how agents connect to tools and APIs.
• Compare rule-based vs. LLM-driven planning.
• Understand frameworks like LangChain, ReAct, and AutoGen.

4. Decision-Making and Behavior

• See how agents select actions with scoring, planning, or heuristics.
• Understand feedback, reflection, and reinforcement.
• Recognize uncertainty handling, retries, and fallbacks.
• Explore how environment and feedback shape behavior.
• Identify cases of emergent behavior.

5. Multi-Agent Systems and Coordination

• Compare centralized and decentralized designs.
• Describe communication and message-passing methods.
• Explain negotiation, delegation, and shared tasks.
• Weigh the benefits and risks of multi-agent systems.
• Understand role specialization in agent teams.

6. Real-World Applications

• Categorize uses across domains like support, research, and automation.
• Distinguish between short-term tasks and long-running workflows.
• Examine examples such as AutoGPT, Adept, and AgentOps.
• Match design strategies to specific business needs.
• Note industry adoption patterns and trends.

7. Evaluation and Deployment

• Define success metrics like task completion and reasoning quality.
• Recognize failure modes including hallucinations and false success.
• Describe sandbox testing approaches.
• Compare deployment models: stateless, stateful, or service-based.
• Apply monitoring strategies in production.

8. Ethics and Responsible Design

• Identify risks like autonomy creep and misuse.
• Explain guardrails, monitoring, and human-in-the-loop approaches.
• Consider the balance between autonomy and control.
• Explore strategies for aligning agent behavior with human intent.
• Apply ethical principles of transparency, fairness, accountability, and explainability.
• Anticipate future governance and regulation challenges.