"The automation break-even: when a six-hour setup pays off"
It's easy to spend a weekend automating a task that took ten minutes a month. It's just as easy to keep doing a 30-minute daily chore by hand for years because "setting something up" sounds like a project. The way out is the same in both cases: do the arithmetic before you start.
The three numbers
- Time per run now. Time the task with a stopwatch for a week; guesses are usually low.
- Time per run after automating. Almost nothing is fully hands-off. Count the minutes you'll still spend checking, fixing and approving.
- How often it happens. Per week is easiest.
Then the cost side: set-up hours (be generous; double your first guess) and any monthly tool cost.
The formula
Hours saved per week = (minutes now - minutes after) x times per week / 60
Payback in weeks = set-up hours / (hours saved per week - tool cost expressed in hours)
A worked example: writing up a client summary takes 25 minutes, five times a week. An AI-assisted template brings it to 5 minutes of checking. Set-up: 6 hours. Tool: $20 a month, and your time is worth $40 an hour.
- Saved: 20 min x 5 = 100 minutes, about 1.7 hours a week.
- Tool cost: $20 x 12 / 52 = about $4.60 a week, about 0.1 hours of your time.
- Payback: 6 / 1.6 = under 4 weeks. After that, roughly 80 hours a year back.
Try your own numbers in the break-even calculator.
What the numbers don't show
- Error cost. If a mistake in the automated version reaches a client, how bad is it? High-stakes tasks need a longer checking step, which shrinks the saving.
- Task lifetime. A payback of 30 weeks is fine for a task you'll do for years and a bad deal for one that ends with a single contract.
- Maintenance. Tools change, logins expire, formats shift. Budget an hour a month for upkeep on anything you build.
- Focus. Some tasks are worth automating even with a weak payback because they interrupt deep work. That's a fair reason - just name it.
Where to start
List every task you did more than twice last week. Put the three numbers next to each. Automate the one with the shortest payback first, run it for a month, and only then pick the next. One working automation beats five half-built ones.