The three building blocks of every automation

Every automation, from a two-step Zapier “Zap” to a complex workflow, is built from three pieces. Learn to see tasks in these terms and you can design automations in any tool.

• Trigger: the event that starts the automation. A new email arrives, a form is submitted, a row is added to a spreadsheet, a time of day is reached.
• Actions: the steps that run in response. Create a task, send a message, update a record, call an AI model.
• Conditions: rules that decide whether, or which, actions run. “Only if the email is from a customer”, “if the deal is over $5,000, notify the manager; otherwise add to the queue”.

Where AI fits

Traditional automation is excellent at moving data that is already structured, like a form field into a spreadsheet column. It struggles with unstructured input: free-text emails, documents, notes. An AI step fills that gap. It can classify (what kind of email is this?), extract (what is the company name and budget?), summarize and draft. The output then feeds ordinary conditions and actions.

Trigger:   new email in support inbox
AI step:   classify as billing / technical / sales / other; extract urgency
Condition: if urgent = true -> post to team chat
Action:    add a row to the triage sheet with category and summary

What AI should not do unattended

AI output can be wrong, and confidently so. Keep a person in the loop for anything that sends messages to customers, moves money, deletes data or makes commitments. In this course, every build has a review point: the automation drafts or labels, and a person approves before anything consequential happens.

Try this

Write down three repetitive tasks from your week. For each, name the trigger, the actions and any conditions. Mark which step involves unstructured text, because that is where an AI step might help.