Choose a trigger: when this happens
One event starts a workflow, and picking the right one is most of the work. Earlier triggers make shorter workflows.
One trigger per workflow. If two different events should do the same thing, that is two workflows — and both will be readable, which is worth more than the small duplication.
The principle that produces good automations: pick the earliest honest trigger. Acting when a chat arrives beats acting when somebody notices it was never answered. Earlier triggers need fewer conditions and fewer steps, because less has gone wrong by then.
Reference
A chat goes a period without a reply
The workhorse trigger for most teams. "A chat goes 30 minutes without a reply" catches the exact failure that costs you deals, and it fires early enough to be recoverable.
A contact is created from a source
"A contact is created from Meta or Google Ads" is how a paid lead becomes an immediate phone call rather than a row in a list. Speed to lead is the metric with the most reliable relationship to revenue in this whole product.
A call outcome is set
"A call outcome is set to Callback requested" turns a promise into a supervised process. "A call outcome is set to Do not call" is how you run a compliance cleanup automatically.
A lead stage changes
"Lead stage changes to Won" is the handover into onboarding, fulfilment or billing — whatever should happen the moment somebody buys.
A form is submitted
Captured submissions land on the activity feed; this trigger lets you act on them rather than hoping somebody notices.
An outside system calls a catch hook
The sixth way in. Your website, your ERP, Zapier or Make posts to a URL and the workflow runs.
Pick the earliest honest one
Compare "a chat is unanswered after 30 minutes" with "a chat is unanswered after four hours". The first gives you three and a half hours to recover the customer.
Put it to use
After a week, look at the run count on each workflow. A trigger firing far more often than you expected is usually broader than you thought — and it is consuming automation actions at that rate too.
The parts people get wrong
- One trigger per workflow. Two events doing the same thing is two workflows, and both stay readable.
- The run count in the workflow list is the fastest way to sanity-check a trigger you have just written.