What to automate first, and what to leave alone

Automation projects do not fail because the technology is not good enough. They fail because the wrong task was chosen first.

The typical bad choice is something impressive that runs once a month and that nobody checks. Six months later nobody even knows whether it is still running.

The three question filter

Before automating a task, ask:

How often does it run? Once a month means the saving is small while the failure surface is the same. Daily or weekly is worth it.

What does it cost if it breaks quietly? This is the important one. A broken report assembly is annoying. A broken lead handoff means nobody called anyone back for two weeks. The second kind needs an alert on failure, not only a confirmation on success.

Does it need judgement? If part of the task is somebody deciding what is correct, you do not need automation, you need preparation. The machine gathers the data, a person makes the call.

What is usually worth doing first

Data assembly for reporting. Not the pretty dashboard, just not having to copy from three systems by hand. It runs daily, needs no judgement, and if it breaks you see it immediately.

Lead routing and notification. Get the form submission to where the work actually happens, and tell somebody. Simple, and it makes money directly.

Watchdog runs. A daily job that tells you when the conversion count dropped to zero, when the feed did not refresh, or when a campaign spent its daily budget by nine in the morning. This kind of automation does nothing except notice. It is often the most profitable kind.

What not to do first

Anything that writes or publishes in your brand's name without human approval. Not because it does not work, but because the first time it gets something wrong, it gets it wrong in public.

The metric worth watching

Not how many processes are running. How many hours you no longer spend, and how many errors did not happen. If those two did not improve, the automation is just another system to maintain.