The automation stopped working six weeks ago. Nobody noticed, because a thing that stops happening does not announce itself. There is no error on anyone’s screen, no complaint, no red mark. Just an absence, and absences are invisible.
This is the characteristic failure of small-business automation, and it is worse than never having automated at all, because the manual process it replaced is gone and nobody is checking.
Why nobody notices
Automation succeeds by removing something from human attention. That is the entire point, and it is also the vulnerability.
When a person handled it, they knew when they had not done it. When software handles it, the only evidence of success is that nothing went wrong, which looks precisely the same as nobody looking.
Reminders that stopped sending. Records that stopped copying. A weekly report nobody reads, so its absence goes unremarked for a quarter.
What actually breaks them
- An expired connection. The most common cause by a distance. These run under someone’s account, credentials expire or a password changes, and the connection silently stops authorising.
- Somebody left. If it ran under a departed employee’s account, disabling that account disables the automation. Nothing warns you.
- Something upstream changed. A column renamed, a folder moved, a form field added. The automation looks for something no longer there.
- A quiet limit. Volume caps and throttling. It works for a year, then a busy month exceeds a threshold and it stops mid-run.
The second is worth dwelling on. Anything important should run under an account that belongs to the business rather than to a person, and this is decided at build time or not at all.
The question to ask about anything automated
How would we find out if this stopped?
Ask it about every automation currently running. If the honest answer is “a customer would eventually complain”, that is not monitoring, that is your customers doing quality assurance for you.
The answer should be a specific mechanism with a specific person attached.
What good monitoring looks like
It does not need to be sophisticated. It needs to exist.
Alert on failure, to a person. Most platforms can email someone when a run fails. This is usually a checkbox nobody ticked. Send it to a shared address rather than an individual, so it survives a departure.
Watch for silence, not just errors. The harder and more important half. A failed run raises an error; a run that never started raises nothing. If something should happen every day, something should notice when a day passes without it.
A periodic check by a person. Once a month, someone opens the list and confirms things ran. Ten minutes, and it catches what automated alerting misses.
Write down what exists. One page: what runs, what it does, whose account it runs under, who to tell. Most businesses cannot produce this, which means nobody can check something they do not know about.
The audit worth doing today
List everything automated. Include the small things somebody set up years ago, because those are the ones running under a departed employee’s account.
For each, answer three questions. When did it last run successfully? Whose account does it use? Who finds out if it fails?
Businesses doing this for the first time routinely find at least one thing that has not run in months. That is the normal result, not an embarrassing one, and finding it is the point. It is also a useful moment to check that what is running is still worth automating at all, because processes change and automations rarely get retired.
The honest summary
Automation is not finished when it works. It is finished when someone would know within a day that it had stopped.
Anything that matters needs a business-owned account, a failure alert going somewhere a person actually reads, and a note that it exists. That is a small amount of work at build time and a genuinely expensive omission afterwards. It is the same discipline that makes approval processes survive past month two.
If there are things running in your business and nobody is quite sure what they are or whose account they use, describe what you know about them. That is a common position and a solvable one.
Common questions
How would we know an automation has stopped?
That is exactly the question to ask about each one. If the honest answer is “a customer would eventually complain”, that is not monitoring, it is your customers doing quality assurance for you.
What actually breaks them?
Expired connections most often, then a departed employee whose account it ran under, then something upstream being renamed or moved, then a volume limit hit during a busy month.
What is the minimum worth setting up?
Failure alerts going to a shared address rather than an individual, something that notices silence and not just errors, and one page listing what runs and whose account it uses.