The two-question gut check
For most tasks, two questions settle it:
- 1Do I do this the same way, over and over, every week?
- 2If the computer got it wrong, would it cost me real money or a customer?
A clear yes to the first and a manageable no to the second is your green light. When the answers conflict, run the full score below.
The five-factor score
Score each factor 1 to 3. Add them up.
| Factor | 1 | 2 | 3 |
|---|---|---|---|
| Frequency | rarely | monthly | daily or more |
| Time per run | a few minutes | tens of minutes | an hour or more |
| Error cost | cosmetic, easy to undo | some rework | money, compliance, lost customer |
| Rule-clarity | needs judgment | mostly rules | clear rules, few exceptions |
| Stability | changes weekly | changes a few times a year | unchanged in a year |
- 12 to 15: automate. High-frequency, stable, rule-based work is the textbook case.
- 8 to 11: maybe. Run the break-even math below first. Often the answer is a partial automation with a human checking the output.
- 5 to 7: leave it alone, or hire a person.
The break-even rule
Automate only if the setup time, plus ongoing upkeep, is clearly less than the hours it will save in a year, and ideally it pays back within twelve months.
A few honest anchors:
- Value the time you save at about 1.25 to 1.4 times the wage, to cover benefits and overhead. Then subtract maintenance and any subscription, or you will overstate the win.
- A task done weekly that saves five minutes can justify a lot of one-time setup. A task done once a year almost never does, no matter how annoying it is.
- The first third of the tasks you automate tend to deliver most of the time savings. Start with the highest-frequency, most stable one and stop chasing the long tail.
The "leave it alone" list
Do not automate:
- Rare or one-off tasks. The build costs more than it saves.
- Processes still changing. If it shifts every few weeks, you will spend all your time editing the automation. Wait until it settles.
- High-judgment work. Interpretation, strategy, and creative calls belong to people.
- Relationship and high-stakes moments. Hard customer issues, negotiations, and complaints need a human.
One nuance Auto-Phil stresses: do not automate is not do it yourself forever. A task that is steady and high-volume but needs judgment is often the case to hire a person, not to wire it up. Automation fits steady and rule-based. A hire fits steady and judgment-heavy. Leave-it-alone fits rare or fast-changing.