What happens when a workflow fails

Every automation we build assumes it will eventually fail — a third-party API goes down, a field changes name upstream, a rate limit gets hit — and is built with a specific response to that, not just hope.

Retries

A workflow's first response to a failure is to retry automatically, with a short delay between attempts, since most real-world failures are temporary — a service being briefly unreachable, a momentary rate limit. Most failures never surface to you at all because the retry succeeds.

Alerts

If retries are exhausted, the workflow raises an alert rather than failing silently — you get notified that a specific step did not complete, with enough detail to see what was affected. See the workflow that failed silently for six weeks for why this specific design choice matters as much as it does.

The manual fallback

Every workflow that touches something time-sensitive — order processing, customer notifications — has a documented manual fallback: the exact steps to do by hand if the automation is down for an extended period, so a failure never means the underlying task simply does not happen.