Spotting Automation Opportunities

The "I do this every week" test

You probably already know which tasks are eating your time. The report you format every Monday. The files you rename after every client call. The data you clean up before every monthly review.

If you catch yourself thinking "I just did this last week," that's a signal. The task might be worth automating.

But not every repetitive task is a good candidate. Some are too complex. Some happen too rarely to justify the setup. Some require judgment calls that only you can make.

This page gives you a practical way to decide what to automate, what to leave alone, and how Claude Code changes the math.

Four signs a task is worth automating

A good automation candidate usually has four qualities:

1. It follows a pattern. You do roughly the same steps each time. Rename files, clean up a column, generate a report from a template. If you could describe the process to a colleague and they could follow it without asking questions, it's a pattern.

2. It has clear inputs and outputs. You start with something (a folder of files, a CSV export, a set of text documents) and you end with something (organized folders, a clean spreadsheet, a summary document). If you can point to the "before" and the "after," Claude Code can fill in the middle.

3. It happens more than once. A task you'll never do again isn't worth automating. A task you do every week is. The more often it recurs, the more time you save.

4. It doesn't require much judgment. If you need to read each item carefully, weigh competing factors, and make a nuanced decision, that's your job, not something to hand off. But if the decisions are straightforward ("bug report goes here, feature request goes there"), Claude Code can handle it.

Not every task needs all four. But the more of these it has, the stronger the case for automating it.

When NOT to automate

Some tasks look like automation candidates but aren't:

One-time tasks. If you'll never do it again, just do it. Spending 20 minutes automating a task that takes 10 minutes and never recurs is a net loss.

High-judgment tasks. Drafting a strategy document, choosing which customer complaints need escalation, deciding how to handle an edge case. These require your expertise and context. Claude Code can help you do them faster, but automating them away usually produces bad results.

High-stakes tasks you can't verify. If the output goes directly to a client, a regulator, or a public audience, and you can't easily check every detail, don't automate it. Use Claude Code as an assistant on these tasks, not an autopilot.

Tasks that change every time. If the process is different each time (different inputs, different rules, different exceptions), there's no pattern to automate. You're better off using Claude Code interactively, the way you learned in Module 2.

How Claude Code changes the math

There's a well-known framework for deciding whether automation is worth the effort. It comes from the webcomic xkcd and boils down to a trade-off: how much time does the automation save per occurrence, and how often does the task happen?

If you save five minutes on a task you do daily, that's about 25 minutes per week. Over a year, that's over 21 hours. If setting up the automation takes less than that, it's worth doing.

Traditionally, automating a task like this meant either learning to write a script or asking a developer to build one for you. Either way, hours of work even for a small task. The break-even point was high. Only the most frequent, most tedious tasks were worth the effort.

Claude Code changes that equation.

Setting up an automation with Claude Code takes minutes, not hours. You describe what you want in plain language, Claude Code builds it, and you review the result. If a task saves you five minutes and you do it weekly, spending 15 minutes to automate it pays for itself in three weeks.

This means tasks that were never worth automating before suddenly become good candidates. The monthly report formatting. The biweekly file cleanup. The occasional batch rename. The weekly status update that's always the same format.

The threshold has shifted. If a task takes you more than five minutes and you do it more than twice a month, it's probably worth automating with Claude Code.

A practical way to find your automation candidates

Here's a concrete exercise you can do right now, before you even open Claude Code.

Step 1: List your recurring tasks. Think about your last two weeks of work. What did you do more than once? What felt tedious? What made you think "there has to be a better way"?

Write down five to ten tasks. Don't filter yet.

Some common ones across non-developer roles:

• Reformatting a report before sending it out
• Renaming files to follow a naming convention
• Pulling data from one system and pasting it into another
• Cleaning up a spreadsheet export (fixing dates, removing duplicates, standardizing names)
• Summarizing a batch of feedback, tickets, or survey responses
• Generating a weekly or monthly status update
• Sorting files into folders by date, client, or type
• Converting documents between formats

Step 2: Score each task. For each task, ask yourself:

Step 3: Pick one. Start with the task that scores best: frequent, patterned, clear, low-judgment. That's your first automation candidate. You'll build it in the next page.

Tip: Teresa Torres, author of Continuous Discovery Habits, suggests a useful test for each task: "Can Claude just do this for me? Or should Claude be helping me do this?" Tasks where the answer is "just do it for me" are your automation candidates. Tasks where the answer is "help me" are better as interactive conversations.

The automation trap

One warning before you start.

Automation can become its own time sink. You set out to automate a 10-minute task and spend an hour tweaking the output format, handling edge cases, and adding features nobody asked for. There's even a famous xkcd comic about this (#1319): a developer sets out to automate a small task and ends up spending more time on the automation than the task ever took.

It happens with Claude Code too. The setup that was supposed to save time starts costing more time than the original task.

Guard against this with a simple rule: set a time limit. If the automation isn't working after 15-20 minutes, stop. Either the task is more complex than you thought, or you need a different approach. You can always come back to it later, or do it interactively a few more times to understand the pattern better before trying again.

The goal isn't to automate everything. It's to automate the right things, the ones that free up your time for work that actually needs your brain.

What's next

You've identified your first automation candidate. In the next page, you'll take that task and build your first automation with Claude Code, going from a one-off interactive task to something you can run again whenever you need it.