Build AI Agents and Workflows (No-Code)
Design, build and test automations and simple AI agents with visual tools, and learn how to prove they work before you trust them with real work.
One payment of $39. Instant online access to the full written course. No subscription. Refund policy.
Design and build automations and simple AI agents with no-code tools, and test them before you trust them.
Who it is for. Beginners who do repeated computer tasks for themselves, an employer, or clients (inbox handling, form follow-up, data entry, lead sorting, simple customer questions) and want to automate part of that work with visual, no-code tools. No programming experience needed. If you can use email, a spreadsheet and a web form, you can take this course.
You finish with. The Tested Helper: One Real Workflow or Agent, Built, Tested and Documented. Choose one real, repeated task and deliver a working no-code workflow (with or without an AI step) or a simple agent, together with the evidence that you tested it: a written spec, a test log, a safety review, a cost estimate and a runbook. You may build in your own accounts for your own task, or for a willing friend, colleague or employer who has given you permission. Use only data you are allowed to use.
Certificate. Finish every lesson, resolve every quiz question with at least 50% right on the first try, and tick the capstone checklist — Apex Flow Academy issues a verifiable Certificate of Completion with a unique ID and a public verification page. It is a certificate of completion, not a degree, licence, accreditation or exam result.
01 · Automations, Workflows and Agents: The Map4 lessons
Learn the real difference between an automation, a workflow with an AI step, and an agent. See how a workflow is put together, pick a first job that is worth automating, and set up a safe workshop.
- 1.1Automations, Workflows and Agents: Three Different ThingsFREE
- 1.2The Anatomy of a Workflow: Triggers, Actions and Data
- 1.3Pick a Job Worth Automating
- 1.4Set Up Your Workshop
02 · Building Reliable Workflows4 lessons
Build your first multi-step workflow, then add conditions, paths, data handling and app connections. This is the base that every AI step and agent later stands on.
- 2.1Your First Real Workflow: Form, Sheet, Alert
- 2.2Filters, Paths and Conditions
- 2.3Handling Data: Clean It, Look It Up, Never Duplicate It
- 2.4Connecting Apps: Permissions, Webhooks and APIs Without Code
03 · Adding AI Steps to a Workflow4 lessons
Decide where an AI step belongs, write prompts that work like a specification, turn AI output into clean data your workflow can trust, and ground the AI in your own information.
- 3.1Where an AI Step Belongs (and Where It Does Not)
- 3.2Writing Prompts That Work Like Specifications
- 3.3Turning AI Output into Data Your Workflow Can Use
- 3.4Giving the AI Your Own Information
04 · Designing and Building Simple Agents4 lessons
Understand what changes when an AI picks its own steps. Write an agent spec, build a simple agent with tools and knowledge, and add human approval and simple handoffs.
- 4.1From Workflow to Agent: What Actually Changes
- 4.2Write the Agent Spec Before You Build
- 4.3Build a Simple Agent in a No-Code Builder
- 4.4Human Approval, Handoffs and Simple Multi-Step Patterns
05 · Test Before You Trust4 lessons
Build a real test set, run and debug your build, put safety guardrails in place for permissions, private data and prompt injection, and judge AI quality with a rubric before anything goes live.
- 5.1Build a Test Set Before You Trust Anything
- 5.2Run, Read and Debug
- 5.3Safety: Permissions, Private Data and Prompt Injection
- 5.4Judging AI Quality and Knowing When to Stop It
06 · Run It, Maintain It, Ship It4 lessons
Estimate costs and limits, monitor and maintain what you built, document it so someone else can run it, and complete your final project with a tested, documented workflow or agent.
- 6.1Costs, Limits and Reliability
- 6.2Monitoring and Maintenance
- 6.3Handing It Over: Documentation and Honest Scoping
- 6.4Final Project: Assemble and Review Your Tested Helper
Automations, Workflows and Agents: Three Different Things
You will be able to
- Tell an automation, a workflow, an AI step and an agent apart.
- Place any project on a four-level ladder.
- Pick the smallest level that does the job.
Why this matters
People use these words as if they meant the same thing. Then they build something too big, too costly or too risky. Once you know the four levels, you can choose the smallest tool that works. That saves you time, money and worry.
Learn it
The four words
An automation is one rule: when this happens, do that. For example, when someone sends a form, add a row to a sheet. The event that starts it is the trigger. The thing it does is the action.
A workflow is a chain of steps. Each step passes its result to the next step. You draw every path yourself.
An AI step is one step in a workflow that uses a language model. That is a program trained to read and write text. It can do jobs a fixed rule cannot, like reading a messy message and picking a label. The rest of the workflow stays fixed.
An agent is different in one big way. The AI picks its own next step. You give it a goal, some instructions and a set of tools. A tool is an action it is allowed to use, like looking up a price. The agent uses a tool, reads the result, and decides what to do next. You do not draw its path. You draw its fence.
The four-level ladder
| Level | What it is | Who picks the path | Main risk |
|---|---|---|---|
| 1. Rule automation | Trigger, then action | You | Low. It fails loudly. |
| 2. Workflow with branches | Several steps and conditions | You | Low to medium |
| 3. Workflow with AI steps | Fixed path, AI handles messy parts | You, plus the AI at one step | Medium. The AI can be wrong. |
| 4. Agent | AI chooses steps and tools | The AI, inside your limits | Higher. Paths change each run. |
The main rule
Climb only as high as the job needs. If a plain rule can do the job, use a rule. A rule is cheaper and faster. It also gives the same answer every time. Add AI only where rules run out. Add an agent only when the path truly changes from case to case.
How each level fails
Rules fail loudly. A missing field stops the run and you see an error. AI steps fail quietly. The run shows success, but the label is wrong. Agents can fail in more ways. They can call the wrong tool, loop, or answer with confidence from nothing. That is why this course spends so much time on testing.
How to use this course
Each lesson ends with a short task. Do the task. The tasks add up to one final project. Keep a build log, which is a plain document where you note what you changed and what happened. When the course names a tool, it is only an example. Check its current features and pricing first, because tools change often.
See it in action
Example (illustrative numbers). Fieldstone Bike Repair is a made-up shop with two staff. Customers fill in a web form with a name, an email, a request type and a message. On a busy day about 20 arrive. The owner reads each one, copies it into a sheet and writes back.
Here are three designs for the same problem.
- Design A, level 1. When a form arrives, add a row to the sheet and post a note in the team chat. No AI. This removes all the copying.
- Design B, level 3. Do everything in Design A. Then add an AI step that reads the message and labels it: quote, question, complaint or other. A complaint alerts the owner at once. The AI does one small job.
- Design C, level 4. An agent reads the message, looks up prices, checks the booking sheet and drafts a reply. It picks its own tools. It is the most flexible design. It is also the hardest to test.
Design A solves a real problem on its own. Many jobs never need more. You will build all three levels for this shop, so you can feel the trade-offs yourself.
Common mistakes
- Reaching for an agent first. Agents sound exciting. Fix: start at level 1. Move up only when you can name what the lower level cannot do.
- Calling any AI feature an agent. The words blur together. Fix: ask who picks the next step. If you do, it is a workflow. If the AI does, it is an agent.
- Trusting a green success mark. An AI step can succeed and still be wrong. Fix: check the content of the result, not only the status.
You are done when
You can explain the four words to a friend in your own sentences. You have also placed three tasks from your own week on the ladder, and for each one you can say why a lower level would not be enough.
You finished the free lesson
That is one lesson from the course. The full course gives you every remaining lesson, a quick check and a hands-on task in each one, and the workbook of templates and checklists.
Full course$39
- A computer with a reliable internet connection
- A free account on ONE no-code automation platform. Examples: Zapier, Make, or n8n (n8n can also be self-hosted, which is more technical). Check each tool's current free plan and limits before you pick; free plans change and usually cap runs, steps or active workflows
- A free spreadsheet tool such as Google Sheets, a free form tool (Google Forms or the form feature of any tool you already use), and a free email account you can use for testing
- A free AI chat assistant, used to test prompts by hand. This is the free path for every AI lesson
- Optional and NOT required: a paid plan, or an AI provider key that bills per use. Some AI actions inside automation platforms are limited or unavailable on free plans, and API usage can cost real money. Everything core in this course can be completed on free tools plus the manual testing path taught in module 3
- A notes document or notebook to keep your build log, test log and specs (the workbook gives you the templates)
- Comfortable using email, a web browser, and a spreadsheet (rows and columns; formulas are not needed)
- Ability to create free online accounts and follow on-screen setup steps
- One repeated task from your own work or life that you understand well enough to describe step by step
The Tested Helper: One Real Workflow or Agent, Built, Tested and Documented
Choose one real, repeated task and deliver a working no-code workflow (with or without an AI step) or a simple agent, together with the evidence that you tested it: a written spec, a test log, a safety review, a cost estimate and a runbook. You may build in your own accounts for your own task, or for a willing friend, colleague or employer who has given you permission. Use only data you are allowed to use.
- One-page process map of the task as a human does it today, with the steps you chose to automate highlighted
- Workflow or agent spec sheet (purpose, trigger, tools, data, boundaries, escalation, success criteria)
- The working build, shown with screenshots or a short screen recording of a full successful run and a caught failure
- Test log with at least 15 cases covering normal, edge, messy, out-of-scope and hostile inputs, with expected results and actual results
- Safety and permissions review (what it can read, what it can change, what needs a human, what the kill switch is)
- Cost and limits estimate using your real plan's units, with your assumptions written down
- A one-page runbook: what it does, how to pause it, how to check it weekly, who owns it
A website built from your brief
Take the website brief you wrote in this course to Apex Flow Digital and see the website options for your business.
See website options- Automation
- One rule that does a job for you when something happens, like adding a row to a sheet when a form arrives.
- Workflow
- A chain of steps where each step passes its result to the next. You draw every path yourself.
- Trigger
- The event that starts a run, such as a new form response, a new email or a schedule.
- Action
- A thing a workflow does after it starts, like creating a row or sending a message.
- Run history
- A log of each run that shows what went into and came out of every step. It is the first place to look when something breaks.
- Field
- One named piece of information, like email or message, that a step produces and later steps can use.
- Mapping
- Choosing which field from an earlier step goes into which box of a later step.
- JSON
- A plain text way to write data as names and values, with curly braces around one record and square brackets around a list.
- Webhook
- A web address that starts your workflow when another app sends it data. Treat it like a secret.
- API
- A documented way for programs to swap data with a service. An HTTP request step uses one.
- Language model
- A program trained to read and write text. It powers AI steps and agents.
- AI step
- One step in a fixed workflow that uses a language model, for example to label a message.
Browse the full Academy encyclopedia
15 checks were run and recorded while writing this course (code, formulas, commands and facts), and it lists 4 official sources it was checked against. Prices, features and policies of outside tools can change, so check each tool's own website.
Created by Apex Flow Academy with AI assistance. For education only; not legal, tax, financial or medical advice. Results depend on your effort and circumstances.
Python From Zero: Write Programs That Work
You write and run real Python programs from the first lesson, and you finish with three small working tools that you built yourself, tested with checks you wrote, and can change without starting over.
Python Automation: Make the Computer Do Your Boring Work
You automate files, spreadsheets, email drafts and web lookups that eat your week. You finish with your own script that plans before it acts, keeps a log, is tested against sample data, and runs on a schedule.
SQL: Ask Any Database Any Question
You will be able to write SQL that answers business questions from a database: choose and sort columns, filter rows, summarise and group, join tables, and check that your numbers are right. You finish with a report you built and checked yourself.