Building Agents with Generative AI

Build Practical AI Agents for Real-World Workflows

Generative AI is making it easier for professionals without traditional programming backgrounds to design and build useful AI agents that can follow instructions, use tools, maintain context, and interact with workplace systems. This hands-on introductory course teaches you how to build practical AI agents using guided templates, AI-assisted development, and commonly available AI tools rather than conventional programming. You’ll learn how to define agent goals and instructions, design workflows, use tools, manage memory and state, incorporate human approval, and create simple multi-agent workflows.

Through instructor-guided activities and hands-on projects, you’ll work with tools such as Visual Studio Code, Claude Code, GitHub, Claude, Gemini, and Slack. You’ll also learn how to test agent behavior, apply guardrails, and refine your workflows for practical professional use. The course culminates in a working AI agent project designed to automate or augment a real professional task.

What You Will Learn

By the end of this course, you will be able to:
 

• Explain what an AI agent is and how it differs from a traditional prompt or chatbot.
• Identify opportunities to use AI agents in professional workflows.
• Define an agent’s goals, instructions, inputs, outputs, tools, and limitations using structured templates.
• Use AI-assisted development tools to create, modify, document, and manage agent projects without building programs from scratch.
• Build AI agents that can follow multi-step instructions, use tools, and maintain context or state.
• Connect agents to workplace tools such as Slack for notifications, requests, and human approval.
• Design single-agent and basic multi-agent workflows using planning, reflection, and handoffs.
• Apply human-in-the-loop approaches and guardrails to improve agent reliability and control.
• Test and refine agent behavior using expected and failure-case scenarios.
• Build and document an AI agent that addresses a practical professional use case.

Topics Covered

• Prompt Templates for AI Agents: Define agent goals, roles, instructions, inputs, outputs, and limitations using structured templates.
• AI-Assisted Agent Development: Use Visual Studio Code, Claude Code, Markdown, and GitHub to create and manage agent projects with AI assistance.
• Agent Workflow Patterns: Explore practical workflow patterns such as chaining, routing, planning, and reflection.
• Memory and State: Learn how agents maintain information and context across multiple steps and interactions.
• Tool Use and Planning: Enable agents to select and use approved tools and capabilities to accomplish tasks.
• Claude and Gemini APIs: Explore basic API usage, account setup, access keys, and cost considerations when working with AI models.
• Workplace Integration with Slack: Connect agents to Slack for notifications, requests, and human approval workflows.
• Human-in-the-Loop Workflows: Design workflows that incorporate human review, approval, and intervention when needed.
• Multi-Agent Collaboration: Explore basic multi-agent workflows, including agent roles, handoffs, and coordination.
• Guardrails, Testing, and Agent Evaluation: Test agent behavior, handle failure cases, apply guardrails, and refine agent workflows.
• Final Agent Project: Build and document a working AI agent that automates or augments a real professional task.

Course Details and Next Steps

• Course Typically Offered: Online during the Winter and Summer quarters. 
• Next steps: Upon completion of this class, consider enrolling in other courses in Artificial Intelligence. 
• More information: Contact infotech@ucsd.edu to learn more about this course.