Automation

When not to automate.

RTOPilot, built by Nexora, hands its team back more than 15 hours of admin every week, live at rtopilot.com.au. That is what a good candidate looks like. Most tasks are not that, and the arithmetic that separates them takes about a minute — which is why this site publishes a calculator that will tell you not to bother.

How do you know what is worth automating?

Frequency times minutes, against the hours it would take to build. A task that takes an hour once a quarter costs four hours a year, and no build worth doing repays that. One that takes four minutes forty times a week costs nearly 140 hours a year, and that is a different conversation entirely.

Why do the small tasks win?

Because nobody notices them. The big annual job everyone complains about is visible precisely because it is rare. The tasks worth removing are the ones so routine that the people doing them have stopped counting — re-entering the same record, rebuilding the same schedule. RTOPilot's 15 hours a week came from that kind of work.

What does the arithmetic leave out?

Maintenance, which varies more than any average would admit, and the errors the manual version makes, which are real and push the case the other way. Treat a break-even figure as the pessimistic end rather than the answer.

What if the payback is longer than a year?

Leave it alone. Past twelve months the task is usually rarer than it felt, or the build is larger than it needs to be, and both are better found before anyone quotes than after.

Questions we get asked

Is a task that only one person does worth automating?
Sometimes, but the arithmetic is harsher: the saving scales with the number of people doing it, so a one-person task needs to be either frequent or slow to repay a build.
What if we cannot estimate the build hours?
Start from the bands this studio works to — about two weeks for a landing page, three to four for an AI assistant, six to ten for a full product — and put smaller automation below all three.