All courses · Platform Mastery · Multiple

Build No-Code AI Agents: Scheduled Tasks, Connectors and Safe Automation

Build No-Code AI Agents: Scheduled Tasks, Connectors and Safe Automation is a self-paced AIvanced Academy course (20 lessons, 5h 17m).

Design, test and safely run a small AI agent on a schedule, with only the access it needs, a person who approves before anything is sent, a log and an off switch.

Length: 5h 17m
20 lessons in 6 modules
Level: Advanced
Certificate of completion
Price: $29.99 one-time, lifetime access

Also included with Academy Membership ($29.99/month or $249.99/year) and with a Teams plan. See pricing

About this course

An AI agent is a helper that works through several steps on its own, often on a schedule, and can use your other apps. That makes it useful. It also means it can make mistakes while nobody is watching. This course is the next step after AI Workspace Building Blocks: Projects, Artifacts, Skills and Agents. You will design, test and (if your plan allows) build one small AI agent that runs on a schedule, uses only the access it needs, waits for a person before anything is sent, and has a log you can check and an off switch you can find.

You will write agent instructions that hold up when nobody is there, label every action Auto, Approve or Never, pick and narrow triggers, connect apps the safe way, test with made-up data and planted traps, and learn how hidden instructions in emails and web pages can try to trick an agent. Then you will practice running one: reading the run history, stopping an agent that went wrong, keeping it healthy and sharing it with your team. You finish by designing, testing and documenting a new agent from scratch.

Every exercise uses one made-up company, Halden Moving & Storage, with made-up data and a model answer.

What you need: to build for real, an AI plan with scheduled tasks, such as a paid ChatGPT or Claude plan, or a Microsoft or Google work plan your company has turned on. Free plans may have a limited version or none, so check your plan. Connecting work email, files or calendars needs your company's OK, and often an IT contact or admin. No plan with scheduled tasks? Every exercise also has a path you can do in any normal AI chat with the made-up data, and a path on paper with no AI tool at all. Real work data only goes into a tool and setup your company has approved. Never connect work accounts to a personal AI plan.

Before you start: finish AI Workspace Building Blocks: Projects, Artifacts, Skills and Agents, or make sure you can explain a project, a skill, a connector and a scheduled task in one sentence each. ChatGPT Plugins and Skills: Package Your Best Work, Write Your Company's AI Policy and AI Privacy, Safety, and Responsible Use pair well with this course.

Topics: ai agents, scheduled tasks, connectors, automation, chatgpt, claude, copilot, gemini, zapier, prompt injection, human approval, small business

What you'll learn

Explain what an AI agent is and isn't, and pick a safe first job for one with a six-question test.
Write agent instructions that run alone, with clear steps, a stop-and-flag rule and a never-do list.
Label every action an agent might take as Auto, Approve or Never, and match your tool's settings to the labels.
Write a one-page Agent Spec with an owner, a narrow trigger, a test plan, a log check and an off switch.
Dry run and schedule a small agent with made-up data, and ask IT for the least access it needs.
Test an agent with planted traps and a harmless prompt injection note, then decide "launch" or "fix first."
Read run history, stop an agent that went wrong in the right order, and keep it healthy or retire it.
Know when a plain no-code automation is a better choice than an agent.

A Try It exercise called "Your first prompt": a short scenario about moving a team meeting from 2 p.m. to 3 p.m., then numbered steps to send a prompt in ChatGPT, Copilot, Gemini or Claude, paste the reply, and note one thing to keep or change. — An example from AI Foundations, Lesson 1. Lessons in Build No-Code AI Agents work the same way: you do the task in your own AI tool, then a quick check explains the answer. Try this course's first lesson free

A quick check question about an AI tool that confidently states a rent late fee. The answer "It may be a guess, because the tool can't see Fernhill's leases" is marked correct, with an explanation of why asking again proves nothing and the fact should be checked in the lease. — An example from AI Foundations, Lesson 1. Lessons in Build No-Code AI Agents work the same way: you do the task in your own AI tool, then a quick check explains the answer. Try this course's first lesson free

Course outline

Module 1: Agents Without the Hype

What an Agent Is (and Isn't), Plus Your Ready Check (free preview)
Where Agents Live Now: ChatGPT, Claude, Copilot, Gemini and Automation Tools
Guided Check: Is This a Good First Agent Job?

Module 2: Write It Down Before You Build

Writing Agent Instructions That Hold Up When You're Not There
Actions and Permissions: Auto, Approve or Never
Guided Check: The One-Page Agent Spec

Module 3: Build, Schedule and Connect

Your First Scheduled Agent, Step by Step
Triggers: Schedules, New Emails and Other Events
Connecting Apps Safely: Test Folders, Shared Accounts and Least Access
Structured Simulation: The Quote Sorter's Pilot Week

Module 4: Test Before You Trust

Test Plans for Agents: Fake Data, Traps and Dry Runs
Prompt Injection: When the Data Talks Back
Guided Check: The Pre-Launch Check
When to Use a No-Code Automation Tool Instead

Module 5: Run, Watch and Fix

Logs and Run History: How to Check an Agent's Work
Structured Simulation: The Agent Went Wrong
Keeping an Agent Healthy: Owners, Reviews and Retirement
Sharing an Agent With Your Team

Module 6: Final Project

Final Practice Exercise: Design, Test and Document the Move-Day Reminder Agent
Final Knowledge Check

Start with the free preview lessons, no account required. Questions about buying