Operations
Five questions to ask before you automate anything
FLOW AI · 5 August 2026 · 6 min read

Most automation projects don’t fail because the tool was wrong.
They fail because nobody wrote the process down first. The work lived in someone’s head. It had exceptions nobody mentioned, and a step that only happens when a customer calls twice.
One person knew about that step. Then it got automated. And the step quietly disappeared.
You don’t find out on launch day. You find out four months later, when somebody finally checks.
The build is rarely the hard part. The thinking before it is where these go wrong, and it takes about twenty minutes. Here’s what we ask first.
1. What exactly happens today, step by step?
Write it out. Every step, in order. Including the ones too obvious to mention.
Most people can’t do it on the first try. That’s the useful part. A process that can’t survive being written down won’t survive being automated either.
Watch for the words and then I just. That’s where the undocumented judgment hides. And then I just check whether they’re a repeat customer. That isn’t a small step. It’s a rule, and a machine needs it spelled out.

2. How often does it actually run?
Not how often it feels like it runs. Count it.
Ten minutes, twice a day, is about eighty hours a year. Load that at $45 an hour and there’s $3,600 sitting inside one small task. Now take the two hour job that runs once a quarter. Eight hours a year. You’ll spend longer building the automation than you ever get back.
Frequency beats duration nearly every time. The small annoying thing that happens constantly is the better target.

3. What happens when it fails?
Every automation breaks eventually. An API changes. Someone renames a field. A customer submits something nobody planned for.
So the question isn’t whether it breaks. It’s who finds out, and how fast.
Failing loudly is fine. Failing silently is worse than never automating at all, because you’ve stopped watching. You think it’s handled. Three weeks later you learn nothing has run since the fourth.
Decide the failure path before you build the happy path.

4. Who owns it after it goes live?
Automation isn’t something you finish. It’s something you keep.
Someone has to notice when it breaks. Someone has to update it when the business changes. If that person isn’t named before you build, the answer is nobody. And nobody is how automations rot.
This is the question that gets skipped most. It also costs the most.

5. What does done look like?
Pick a number you can check.
Hours back per week. Response time in minutes. The share of tickets closed without a person touching them. Something you could put on a screen and know inside a month whether this worked.
It’ll save time isn’t a target. It’s a hope. And it’s why so many automations run for a year with nobody able to say whether they were worth building.

None of this needs a developer. It needs twenty minutes and a willingness to write down what you actually do.
Most people skip it because building is more fun than thinking. Then they build the wrong thing beautifully.