Triggers & actions

Time to get concrete. In every platform you'll assemble workflows from a catalog of triggers ('when this happens...') and actions ('...do this') that each connected app exposes. Gmail offers 'new email matching search' as a trigger and 'send email' as an action; Sheets offers 'new row' and 'add row'; Slack offers 'new message in channel' and 'post message'. Workflow building is picking one trigger and chaining actions — the craft is in the picking.

Choosing the trigger (more decisions than it looks)

• Event triggers ('new email at support@') give instant response — the 2am ticket logs at 2am. Prefer them when latency matters. Watch the scope: 'new email' fires on newsletters too; 'new email matching to:support@ -label:automated' fires on what you meant. Precise triggers prevent 80% of embarrassing runs.
• Schedule triggers ('Mondays 7am') suit digest work: reports, summaries, reminders. The design question is what the run reads when it wakes — usually 'everything new since last run', which some platforms track for you and others make you implement yourself with a timestamp or status column. Problem set 2's classic double-processing bug is exactly this done wrong.
• Manual triggers (a button) are underrated: automation you invoke deliberately. Riley's invoice-reminder drafts run on a button — automated labor, human timing. Also the perfect training wheels while you're building trust in a new workflow.

One latency reality check before you promise anything: on many free tiers, event triggers aren't instant — the platform polls for new events every ~15 minutes, and true instant triggers (webhooks) are typically a paid feature. If your Module 1 brief's success sentence promised 'appears within 1 minute', re-word it against your plan's actual polling interval before a lab holds you to it.

Setup that saves future-you

1. 1Pick your platform: whatever your workplace already licenses; otherwise start on a free tier — but check it honestly first. Free tiers vary: some gate multi-step workflows, branching, or premium connectors behind paid plans. Confirm yours allows a multi-step workflow with a filter before starting the labs; run volume is the other ceiling you'll eventually hit.
2. 2Connect apps with the right account — a shared/service account (ops@, not riley@) wherever the workflow acts for the team. Automations bound to personal accounts break on vacation and die at offboarding; this one habit prevents both. No admin rights to create one? Do the course on personal accounts and hand the service-account step to IT when you productionize.
3. 3Check the permission screen: actually read what the consent screen requests. Many providers offer no scope choices at all — if the access requested is broader than the workflow needs (manage all mail, to read one mailbox), note it and flag it for your security folks. Knowing exactly what you granted is the habit; narrowing it is often not on offer.
4. 4Name things properly from workflow #1: [dept] what it does - v1 ('OPS email->ticket v1'). Six months from now, 17 workflows named 'My Zap 3' is an archaeology project.