The five-times rule.

If a workflow in your business runs more than five times a day by hand, it should run itself. How to spot the candidates, the honest payback maths, and why every automation needs a fail-safe.

The simplest rule we use to find automation candidates in a new client’s business is this: any workflow that runs more than five times a day, every day, by hand, should run itself. That is the bar. It is deliberately low.

Most owners react to the rule by saying their business does not have such workflows. Then we walk the floor for an afternoon and count fifteen. Sending the booking confirmation. Sending the follow-up two days later. Updating the spreadsheet that tracks the cohort. Generating the invoice. Marking the invoice paid. Notifying the customer the invoice is paid. Reconciling the bank statement to the invoice. Copying the new client from the contract platform into the CRM. Each of these takes a minute. Each of them runs every day. Each of them is paid for in someone’s hours.

Automation is the cheapest leverage in software. It does not need AI. It does not need a custom database. It does not need a new tool. Most of the work is wiring up systems your business already pays for — Zapier and Make are good, Trigger.dev is better for anything that involves real logic, webhooks are best when both endpoints support them.

The maths is easy and almost always favours doing it. Twelve workflows in your business, each running ten times a day, each consuming one minute of someone’s time, is twenty hours a week. Twenty hours a week is half a salary. The build cost for a typical workflow is three to eight hours of engineering time. Most workflows pay themselves back in under a month.

What automation is good at: deterministic transitions between known states. A new order arrives → send a confirmation. An invoice is created → send a notification. A booking is confirmed → block the calendar. These have one right answer and they happen often.

What automation is bad at: judgement. A customer complaint needs a human. A pricing decision needs a human. An exception that breaks the assumed flow needs a human. Automation is reliable to the degree the flow is predictable; the moment a real human has to intervene, the automation should be smart enough to escalate rather than guess.

The wrong question to ask first is “what should I automate?” The right question is “which workflows in my business have I been meaning to fix for more than a year?” Those are the candidates. The reason they have not been fixed is that no one has time. Automating them is the way they get fixed.

The honest payback warning. Automation pays back, but it also creates a maintenance surface. A webhook that breaks silently is worse than no webhook. Every automation needs a fail-safe: an alert when it stops running, a fallback when an external service is down, a log that can be read after the fact. This is not glamorous. It is the difference between an automation that earns its keep and one that costs you the trust of your team.

→ Book a Discovery. We will count the workflows in your business that have crossed the five-times rule and price what closing them is worth.