Automation for · n8n
n8n, built to survive a Tuesday.
n8n is a good choice and it is not the interesting decision. Anyone can draw a workflow that moves a record from one place to another; the question is what that workflow does at three in the morning when the far end returns a 502 and nobody is watching.
Most of what I am asked to fix is not a missing feature. It is a workflow that works on the happy path and has no answer for anything else.
What I actually build
Five things that are not on the canvas.
- 01 Retries and backoff where a retry helps, a fallback queue where it does not BUILT
- 02 One shared error handler, so an alert reaches one place BUILT
- 03 Idempotency, so a repeated event does not produce two of everything BUILT
- 04 Service accounts with scoped access, never pasted into a node BUILT
- 05 Rate limiting under the quotas of whatever sits underneath BUILT
A sync that dies at the API limit loses records silently, which is why the last one is a build item rather than a tuning exercise.
Whose system is it afterwards
“The tool matters less than who can debug it at two in the morning.”
That is the question I ask before recommending anything: when a step fails, who is alerted, can the run be replayed, can you see what ran and why, and if the pricing changes at your volume, can you move.
Everything runs in accounts you own. There is no part of the system that depends on me being reachable.
The largest of these is a suite of 30 workflows across twelve services, twenty of them live and 607 nodes in all, running in production for a professional services firm. It cleared 25,007 executions in the ten days to 18 August 2026, one of which failed.
Three in the morning
What does your workflow do when nobody is watching?
Tell me what breaks, or what you are afraid will break quietly. I will tell you what I would look at first.
Tell me what feels too manual (opens in a new tab)or email dan@burdetsky.xyz