Course A · Module A3 No-Code Automation · Lesson 9 of 17

Multi-Step & Safe Failure

One-step automations are fine. Two-step automations with a safety net are how you actually save an hour a day.

~14 min · Have Make or Zapier open · Prereq: A8 first automation built

Your first automation did one thing. Real business workflows do five — and sometimes they hit a snag halfway through. This lesson's single win: by the end you'll know how to chain steps together, add a branch that says "if X, do this; if not, do that," and build in a human-in-the-loop checkpoint so the automation can never do something embarrassing to a real customer without your eyes on it first.

Chaining steps: one trigger, many actions

A multi-step automation is just a recipe with more than one instruction. The trigger fires once; every action after it runs in order.

Step 1Trigger — a new form submission arrives.
Step 2Action — look up the contact in your CRM.
Step 3Action — ask the AI to draft a personalised reply.
Step 4Action — add a row to your tracking sheet with the draft attached.

Build one step at a time. Run the scenario after each addition. If step 2 breaks, you know exactly where to look — you didn't add four things at once.

Branching: if this, do that

Most platforms call this a Router (Make) or a Filter / Path (Zapier). It lets you split the flow based on a condition.

Think of branching like a receptionist's decision tree: "If the enquiry says refund, route it to the billing folder. If it says new client, route it to onboarding. Anything else, put it in the general inbox." One trigger, three destinations — no manual sorting.

Common branch conditions that are easy to set without code:

Testing before you go live

Never activate an automation on real, live data until you have tested it with a fake example first. Every platform has a test / run-once mode. Use it. The checklist:

Before activating, confirm:
  1. Does each step show a green check? A red icon means something is misconfigured — fix it before going live.
  2. Is the output what you expected? Check the actual data the automation produced, not just that it ran.
  3. Did it touch anything real? If step 3 sends an email, point it at yourself during tests, not a client.
  4. What happens if a field is blank? Feed the automation an intentionally incomplete record and see if it breaks gracefully.

Error handling and the human-in-the-loop checkpoint

Two habits keep automations from running amok.

Error handling

Set your platform's error-notification option so it emails or Slacks you the moment a step fails — before you wonder why nothing happened for three days. In Make: Scenario settings → Send a notification. In Zapier: Zap history alerts.

Human-in-the-loop

For anything customer-facing, the automation drafts or suggests — it never sends. Insert an approval step: the draft lands in a Slack message, a Google Doc, or a draft email folder. A person approves, then it ships. One extra click. Infinite peace of mind.

The rule of thumb: if a mistake would make a customer confused, unhappy, or charged incorrectly — add a human checkpoint. If a mistake just means a wrong tag in an internal spreadsheet, it's safe to let it run fully automated.

Check yourself

Answer before opening. Recalling it now is what makes it stick.

You want different emails to go to existing clients vs. new leads. What do you add?

A branch (Router / Filter / Path). You split the flow on a condition — for example, "contact exists in CRM = yes" goes one path; "no" goes another. One trigger, two tailored actions.

Your automation ran but the AI draft landed in the wrong folder. What's your first debugging move?

Open the run history and check the data at each step. Most platforms show exactly what each step received and produced. Find the step where the folder field got the wrong value — that's where to fix the mapping.

Why should a customer-facing automation always draft rather than send directly?

Because a human-in-the-loop checkpoint is your safety net. AI can misread context or get a name wrong. One approval click before sending means a mistake embarrasses no one. Fully automated sending is only safe once you've reviewed many error-free drafts.

Read this next (primary source)

Make Help Center — the official documentation covers multi-step scenarios, routers, filters, and error handling with step-by-step screenshots. Start with "Scenarios" in the left nav. See RESOURCES.md for more.

I'm your teacher — ask me anything. Paste in a screenshot of your scenario or describe the branch logic you're trying to build. I'll sketch out the exact steps in plain English, including which condition field to use and where to drop the human-checkpoint action.

Key terms from this lesson — trigger, action, human-in-the-loop — are defined in the course glossary. Keep it open.