Building Simple AI Agents for Workflow
Covers the building blocks of simple AI agents — perceiving, deciding, acting — and walks through a Python example that connects to email via IMAP and uses an LLM API to summarise unread messages.
{"contentFormat":"slides.v1","completion":{"requireAllSlides":true,"requireQuiz":true},"slides":[{"kind":"title","eyebrow":"Module 5: AI Automation and Workflows","title":"Building Simple AI Agents for Workflow","body":"Learn how to create AI agents that automate repetitive tasks—without writing complex code. We'll explore no-code platforms like Zapier and Make, and show you how to integrate AI steps to build powerful workflows.","outcomes":["Define AI agents and their role in workflow automation","Identify real-world use cases for AI agents across industries","Build a no-code email summarization agent using Zapier and OpenAI","Understand key concepts like APIs, prompts, and rules-based systems","Evaluate ethical and security considerations for AI agents"],"narration":"Welcome to this lesson on building simple AI agents for workflow automation. By the end, you'll be able to create your own AI agents using no-code tools, no programming required."},{"kind":"content","heading":"What is an AI Agent?","body":"An AI agent is a software entity that perceives its environment, makes decisions, and takes actions to achieve a specific goal. Think of it as a digital assistant that can automate tasks like sorting emails, monitoring social media, or generating reports.\n\nKey characteristics:\n- Autonomous: Operates without constant human input\n- Goal-oriented: Designed to accomplish a defined task\n- Interactive: Can use APIs to connect with other services\n\n> Example: An AI agent that scans incoming emails, identifies urgent messages, and sends you a Slack notification.","callout":{"variant":"note","title":"No-Code Focus","text":"In this lesson, we'll build agents using no-code platforms like Zapier and Make. You don't need to write any code—just configure triggers and actions."},"narration":"An AI agent is like a digital assistant that works autonomously. It can connect to different apps via APIs and perform tasks based on rules or AI models. We'll focus on no-code tools so anyone can build one."},{"kind":"content","heading":"Real-World Use Cases","body":"AI agents can automate a wide variety of tasks across industries. Here are some practical examples from around the world:","cards":[{"title":"Email Sorting & Prioritization","text":"An agent scans incoming emails, flags urgent ones (e.g., containing 'urgent' or 'deadline'), and forwards them to your team."},{"title":"Social Media Monitoring","text":"Track brand mentions on Twitter or Facebook. The agent alerts you to negative sentiment or customer inquiries in real time."},{"title":"Data Entry Automation","text":"Extract information from PDF invoices or forms and populate a Google Sheet or database automatically."},{"title":"Meeting Scheduling","text":"Analyze calendars, suggest optimal meeting times, send invites, and handle rescheduling requests."},{"title":"Content Generation","text":"Draft blog posts, social media updates, or product descriptions using an AI model like GPT-4o."},{"title":"Customer Support Triage","text":"Classify incoming support tickets by urgency and route them to the appropriate team."}],"narration":"AI agents are used everywhere—from automating email sorting in Kenya to scheduling meetings in Brazil. The possibilities are endless."},{"kind":"content","heading":"Key Concepts for Building AI Agents","body":"Before we build, let's review essential concepts:","callout":{"variant":"insight","title":"Why No-Code?","text":"No-code platforms like Zapier and Make allow you to build complex workflows without writing a single line of code. They provide pre-built connectors and AI modules that handle the heavy lifting."},"table":{"headers":["Concept","Description","Example"],"rows":[["Workflow","A sequence of steps to complete a task","New employee onboarding: send welcome email → create accounts → assign training"],["Automation","Using technology to perform tasks with minimal human intervention","Auto-reply to common customer queries"],["Rules-Based System","Agent follows predefined if-then rules","If email contains 'urgent', then flag it"],["NLP","Natural Language Processing enables understanding of human language","Summarize an email body into one sentence"],["API","Digital connector that lets apps talk to each other","Zapier uses APIs to connect Gmail to Slack"],["Prompt Engineering","Crafting effective instructions for AI models","Write a prompt like 'Summarize this email in one sentence'"]]},"narration":"Understanding workflows, automation, APIs, and prompt engineering is key to building effective AI agents. No-code platforms make it accessible to everyone."},{"kind":"content","heading":"Step-by-Step: Build a No-Code Email Summarization Agent","body":"Let's build an AI agent that summarizes incoming emails using Zapier and OpenAI. No code required!\n\nPrerequisites:\n- A Zapier account (free tier works)\n- An OpenAI API key (or use Anthropic's Claude)\n- A Gmail account (or any email supported by Zapier)\n\nSteps:\n\n1. Create a new Zap in Zapier.\n2. Trigger: Choose Gmail → New Email (or New Attachment). Set the trigger to watch for new emails in your inbox.\n3. Action 1: Add an OpenAI step → Send Prompt. Configure the prompt:\n - Prompt: Summarize the following email in one concise sentence: {{body}}\n - Model: gpt-4o (or gpt-4o-mini for cost savings)\n - Max tokens: 50\n4. Action 2 (optional): Add a Slack step → Send Channel Message to post the summary to a Slack channel.\n5. Test and turn on your Zap.\n\nExplanation:\n- The trigger fires whenever a new email arrives.\n- The OpenAI step uses the email body to generate a summary.\n- The Slack step sends the summary to your team.\n\n> Security Note: Zapier uses OAuth for Gmail—no passwords stored. Your API key is encrypted. As of 2026, IMAP password authentication is deprecated; always use OAuth.","callout":{"variant":"exercise","title":"Try It Yourself","text":"Set up the Zap above using your own email and a test email. Modify the prompt to ask for a bullet-point list instead of a sentence. See how easy it is to customize!"},"narration":"With Zapier and OpenAI, you can build an email summarization agent in minutes. The trigger is a new email, the AI step summarizes it, and you can send the result to Slack or any other app."},{"kind":"content","heading":"Alternative: Using Make (Integromat) with AI Modules","body":"If you prefer Make (formerly Integromat), the process is similar:\n\n1. Create a new scenario in Make.\n2. Module 1: Gmail → Watch Emails (choose folder, filter by unread).\n3. Module 2: OpenAI → Create Completion. Map the email body to the prompt field.\n4. Module 3: Slack → Send a Message (or use any other app).\n5. Run the scenario once to test, then schedule it.\n\nAdvantages of Make:\n- More granular control over data transformation\n- Built-in error handling and retries\n- Supports many AI providers (OpenAI, Anthropic, Google Gemini)\n\nExample prompt for Claude:\n
\nHuman: Summarize the following email in one sentence:\n{{email.body}}\n\nAssistant:\n