Introduction to Agentic AI

• Explain the fundamental concepts and defining characteristics of Agentic AI
• Differentiate AI agents from traditional AI, generative AI applications, copilots and automated workflows
• Identify the core components of an AI agent — models, goals, instructions, tools, memory and state
• Explain how agents reason about goals, make decisions, use tools and perform multi-step tasks
• Design prompts and instructions that guide agent planning, decision-making and behavior
• Compare common architectures and patterns for building agentic systems
• Distinguish deterministic workflows from autonomous agent-based approaches
• Describe how multiple specialized agents collaborate, delegate tasks and coordinate activities
• Identify enterprise use cases suitable for agentic AI and the appropriate level of autonomy
• Evaluate agent performance using task outcomes, execution trajectories, tool interactions, reliability and cost
• Apply testing, tracing, observability and AgentOps practices to agentic systems

Identify security risks — prompt injection, excessive permissions and tool misuse — and apply guardrails, least-privilege access and human-in-the-loop controls

Evaluate agentic AI solutions from the perspectives of security, governance, accountability and responsible AI

Designed for software developers, architects, technical leads, IT professionals, AI practitioners, business analysts and technology decision-makers who want to understand the concepts, architectures and practical applications of agentic AI. A general understanding of AI and generative AI is helpful but not required; no prior experience developing AI agents or with a specific framework is necessary. The course is technology-neutral, focusing on concepts and patterns that apply across modern agentic AI platforms and frameworks.

• Module 1: Foundations of Agentic AI
  • Places modern AI agents within the broader evolution of AI. Defines agentic AI and its key characteristics (autonomy, persistence, goal-directed behavior); differentiates it from predictive, reactive and traditional AI; traces the evolution from symbolic AI to LLM-powered agents; covers agent types (reactive, deliberative, hybrid), the agent-environment interaction loop, single- vs multi-agent systems, and how assistants, copilots, workflows and autonomous agents differ.
• Module 2: Introducing AI Agents
  • How agents extend generative AI beyond request-and-response by combining models with goals, instructions, tools, memory and decision-making. Covers the core components of an agent; how agents interpret goals and progress toward them; the agent execution loop and the relationship between reasoning and action; tool use for interacting with external systems and data; the role of memory and state; and common patterns for combining models, tools and application logic.
• Module 3: Prompting for Agentic AI
  • Prompting techniques designed for agentic workflows. Writing prompts that establish clear goals and responsibilities; designing instructions that guide planning, reasoning and decision-making; reusable prompt templates for dynamic goals and runtime information; prompts that guide tool selection; incorporating memory and prior state; and using constraints and behavioral boundaries to reduce unintended behavior across multi-step tasks.
• Module 4: Architectures and Patterns for Agentic Systems
  • The building blocks of agentic architectures and how models, agents, tools, memory and workflows work together. Comparing deterministic workflows with autonomous agents; common patterns (chaining, routing, parallelization, orchestration, agent loops); distinguishing conversational context, agent memory and application state; and selecting an appropriate architecture based on requirements and risk.
• Module 5: Tools, Decision-Making and Agent Behavior
  • How agents decide what actions to take, select tools, evaluate results and adapt. Goal decomposition; tool selection and invocation; distinguishing model reasoning from external tool execution; the roles of planning, execution, reflection and evaluation; handling tool failures; recognizing agent loops, ineffective decisions and false task completion; and balancing autonomy with deterministic controls.
• Module 6: Multi-Agent Systems and Coordination
  • Multiple specialized agents collaborating rather than one general-purpose agent. When multi-agent designs help; defining agent roles and responsibilities; communication and coordination patterns; task delegation; combining and evaluating results; sharing context, memory and state; identifying coordination failures; and weighing the benefits against operational complexity.
• Module 7: Agentic AI in Real-World Applications
  • Practical enterprise applications and a framework for deciding when an agent-based approach fits. Common use cases; distinguishing problems needing simple LLM integration, deterministic workflows or autonomous agents; how agents interact with APIs, databases, documents and enterprise systems; agentic patterns for research, document processing, customer support, operations and decision support; determining the right level of autonomy; and where human-in-the-loop belongs.
• Module 8: Evaluating, Deploying and Operating AI Agents
  • Testing, evaluation, deployment, observability and AgentOps for production. Success metrics (task completion, accuracy, reliability, latency, cost); evaluating agent trajectories (decisions, tool calls, observations, outcomes); recognizing failure modes (hallucinations, agent loops, tool failures, false success signals); sandbox and simulation testing; stateless/stateful/service-based deployment; tracing and observability; auditability; and continuous evaluation.
• Module 9: Security, Governance and Responsible Agentic AI
  • Managing the risks agents introduce beyond traditional generative AI. Security and operational risks of autonomous systems; prompt injection; excessive permissions and unrestricted tool access; data security and privacy; least-privilege access; guardrails and validation controls; human-in-the-loop and human-approval controls; balancing autonomy with accountability; transparency, fairness and explainability; and governance, compliance and regulatory considerations.