Automating the boring parts without building a monster

Automation is worth doing where a task is frequent, identical every time, and currently done by a person copying from one screen to another. Outside those conditions it usually creates a fragile system that someone has to babysit.

The five that reliably pay off

1. Form to record. An enquiry that creates a client record rather than only an email. Removes retyping and means nothing is lost in an inbox.
2. Order to document. Generating the invoice or offer letter from data you already hold. This is usually the largest single time saving in a small services business.
3. Scheduled reminders. Unpaid invoice at day 14, follow-up on a quote at day 7. Chasing is unpleasant and therefore skipped, which is exactly why automating it makes money.
4. Backups and monitoring. Not glamorous, and the highest-value automation you will ever set up.
5. Report assembly. The monthly numbers pulled together automatically, even if a person still interprets them.

The ones that cost more than they save

• Automated personal outreach. Recipients can tell, and it damages the thing it is trying to build.
• Complex conditional workflows across five tools. Every integration is a thing that breaks silently, and debugging a six-step chain takes longer than the task did.
• Automating a process nobody has written down. You will encode the confusion.
• Anything used once a month. Ten minutes monthly is two hours a year. It is not worth a week of engineering.

The rule

Automate the copying, not the thinking. Moving data between systems is a machine's job; deciding what to charge, what to say and who to chase is not.

Start with the failure case

Before building anything, ask what happens when it breaks — because it will. Does someone find out? Is there a manual fallback? An automation that fails silently is worse than the manual process it replaced, because the manual process at least had a person noticing.