Not every annoying task is a good candidate
It is tempting to automate whatever irritates you most. The trouble is that irritation is a poor guide. A task can feel tedious yet occur rarely, and another can feel painless yet quietly consume hours. Choosing well means looking at a few plain criteria instead of going by mood.
Step one: find the repetitive work
Before judging candidates, you need a list. Two easy ways to build one:
- Keep a log for a week or two. Whenever you notice yourself doing something for the second or third time, jot a line: what it was, roughly how long it took.
- Review your calendar and sent mail. Recurring meetings, standing reports and repeated replies show where the repetition lives.
Also ask colleagues. Tasks that bother one person are often shared by many, and a shared routine is a stronger candidate because the savings multiply.
Do not trust memory alone. People tend to remember the dramatic frustrations and forget the small chores that happen daily.
Criterion one: frequency
The more often a task happens, the more any improvement is worth. Here is a hypothetical calculation. A task takes 10 minutes and occurs twice a week. That is 20 minutes a week, or 1,040 minutes over 52 weeks, which is about 17 hours and 20 minutes. If the same task takes 10 minutes and occurs four times a year, the total is 40 minutes. The first deserves attention. The second probably deserves only a checklist.
Frequency also interacts with setup time. If building an automation takes three hours, you need a good number of repetitions before it pays back. Keep in mind that maintenance has a cost too, and the true payback period is longer than a simple division suggests.
Criterion two: clear rules
Automation works best on tasks you can describe as “if this, then that.” Ask yourself whether you could explain the task to a new colleague in a short, unambiguous list of steps. If you find yourself saying “it depends” or “I just know,” the task likely involves judgment, and judgment is difficult to automate safely.
A useful test is to count the decision points. A task with a few simple decisions, like “if the amount is over a set limit, flag it,” suits rules. A task whose decisions rely on context, tone or relationships usually does not. Often a mixed approach works: automate the predictable portion and leave the judgment step to a person.
Criterion three: consistency of inputs
Even a task with clear rules can be hard to automate if its inputs vary wildly. Information that arrives in a standard format, like a form submission or a regular export, is easy to handle. Information that shows up as free-form emails, photos of paper, or messages written a dozen different ways is not. Sometimes the best first step is to standardize the input, for example by replacing emailed requests with online forms, so that later steps can be automated at all.
Criterion four: the cost of errors
Think about what happens when the task goes wrong. Automation makes mistakes at speed and in bulk. A rule that misfiles a newsletter costs you a few seconds. A rule that sends an incorrect message to every customer, or changes the wrong records, costs far more.
As a rule of thumb:
- Low error cost, high frequency: excellent candidates.
- High error cost, high frequency: automate with checks, approvals and logs, or automate only the preparation and leave the final action to a person.
- High error cost, low frequency: usually better handled with a careful checklist.
Tasks involving money, legal commitments, health or safety deserve particular caution. For these, automation may help prepare information, but a human should review the result.
Criterion five: stability
If the process is likely to change next month, an automation built around it will need rebuilding. Stable routines such as a weekly report in a fixed format are better candidates than a process still being worked out. When a process is new, run it manually until it settles, writing down the steps as you refine them. Those notes become the specification for any later automation, much like a standard operating procedure.
Criterion six: who benefits
Consider whether the task is boring for you or actually blocks other people. A step that delays a colleague’s work or a customer’s request may be more valuable to fix than a private chore, even if it takes less of your time. Also consider whether removing the task would affect the quality of work. Some tasks that look like pure busywork, such as reading a report, serve a hidden purpose of keeping you informed.
A simple scoring sheet
You can combine the criteria in a small table. For each candidate, give it a rough rating of low, medium or high on each of these:
| Criterion | Good sign |
|---|---|
| Frequency | Weekly or more |
| Clear rules | Steps are the same every time |
| Consistent inputs | Standard format |
| Error cost | Low, or easily caught |
| Stability | Unlikely to change soon |
A task that scores well on most rows is a promising candidate. Treat the scoring as a conversation starter, not a formula. If it points to something you would not have guessed, take a second look, but trust your knowledge of the work.
Ask whether to remove, simplify or automate
Before automating, ask three questions in order:
- Can we stop doing this? Some reports nobody reads, and some approvals add nothing.
- Can we simplify it? Fewer steps, fewer fields, a better form.
- Only then, can we automate it?
Automating a bad process just makes the bad process faster. Many of the best results come from the first two questions.
Match the task to the lowest useful tool
Once a task passes your checks, choose the simplest approach that works. Repeated typing might call for a snippet, a mail sorting problem for an inbox rule, and a copy-and-paste routine between sheets for a formula. Only when those fall short should you consider connecting tools or something heavier. Tasks you have decided against automating are not failures; they are simply better done by a person, and writing that down saves you from reconsidering them every few weeks.