ITTrainingNow

Agentic AI for Technical Teams: AI Agents, Tools, and Automated Workflows

Build AI agents that plan, call tools, and automate multi-step workflows, with the evaluation and security they need for production.

Course overview

Learn to design and build AI agents and agentic workflows in this two-day introductory instructor-led course for technical teams. Understand the agent loop of reasoning, acting, and observing, give models tools through function calling and the Model Context Protocol (MCP), manage memory and state, and choose between deterministic workflows and autonomous agents. Explore agent frameworks and cloud agent services, orchestrate multi-agent systems, add human approval steps, and evaluate, trace, and secure agents against prompt injection and unsafe actions. Hands-on Python labs build an agent that automates a realistic multi-step IT workflow.

2 daysCourse length
8 modulesHands-on Python labs
IntroductoryNo agent development experience required
Technical streamFor developers and IT professionals

Course coverage

Progress from a single tool-calling agent to orchestrated, evaluated, and secured agentic workflows.

  • 01 Agentic AI foundations: the agent loop, autonomy, and when to use agents
  • 02 Tool calling: function definitions, schemas, results, and error handling
  • 03 Model Context Protocol (MCP): servers, clients, tools, and resources
  • 04 Memory, state, planning, and grounding agents with RAG
  • 05 Agent frameworks and cloud agent services
  • 06 Workflows and multi-agent orchestration with human-in-the-loop
  • 07 Evaluating and observing agents: test tasks, tracing, and metrics
  • 08 Securing agents: prompt injection, least privilege, sandboxing, and cost

Participants will learn to

  • Explain the agent loop and choose between workflows and autonomous agents.
  • Define tools with clear schemas and handle tool results and errors.
  • Build and connect MCP servers to give agents access to systems and data.
  • Manage conversation memory, state, and planning for multi-step tasks.
  • Compare agent frameworks and cloud agent services for real projects.
  • Orchestrate multiple agents and add human approval for risky actions.
  • Evaluate agent success rates and trace tool calls and decisions.
  • Secure agents with least-privilege access, guardrails, and cost limits.
Practical outcome: Build a Python agent that triages IT support tickets by searching a knowledge base, calling tools through an MCP server, drafting responses, and requesting human approval before taking action, then evaluate it against test tasks and trace each step with security controls applied.

Two-day course outline

Day 1 - Agent foundations, tools, and MCP: Learn what makes AI agentic: goals, planning, the reason-act-observe loop (ReAct), tool use, memory, and levels of autonomy, and when a deterministic workflow, a single agent, or multiple agents is the right design. Implement tool calling and function calling from Python with JSON schemas, tool results, retries, and error handling, and build a simple agent loop from scratch. Connect agents to systems and data with the Model Context Protocol (MCP): build an MCP server that exposes tools and resources and use it from an agent and from AI coding assistants such as GitHub Copilot. Manage short-term conversation memory, long-term memory, state, and task planning, and ground agents with retrieval-augmented generation (RAG) as a tool. Labs: build a tool-calling agent from scratch and an MCP server for a ticketing system.

Day 2 - Frameworks, orchestration, evaluation, and security: Compare agent frameworks and services including the OpenAI Agents SDK, LangGraph, Microsoft Agent Framework and Semantic Kernel, CrewAI, Google Agent Development Kit (ADK), Azure AI Foundry Agent Service, and Amazon Bedrock Agents. Build workflows with routing, chaining, parallel steps, handoffs, and multi-agent orchestration, and add human-in-the-loop approvals and checkpoints. Evaluate agents with test task suites, success and tool-accuracy metrics, LLM-as-a-judge, and tracing and observability with OpenTelemetry. Secure agents against prompt injection and tool misuse with least-privilege credentials, sandboxed execution, input and output guardrails, rate and cost limits, loop limits, and audit logging. Final lab: build, evaluate, and secure a multi-step IT support ticket agent with human approval.

Format and audience

For: Developers, software engineers, DevOps and platform engineers, IT automation specialists, data engineers, and technical leads who want to build AI agents and automated workflows. Business staff should consider Agentic AI for Business.

Prerequisites: Programming experience, ideally in Python, and familiarity with JSON and calling web APIs. Large Language Models for Technical Teams is helpful but not required. Participants need a computer with VS Code, Python 3, Git, and an LLM API key (instructor-provided or their own).

How it works: Two days of instructor-led explanations, demonstrations, and hands-on labs. Contact us to discuss dates and virtual or onsite delivery.

Explore the course

Request a quote for the two-day Agentic AI for Technical Teams course and discuss your team's training needs.

Request a Quote

Back to AI & ML training