Module 12 — Making It a Habit

🎯 Goal: Turn Claude Code from "a cool thing I learned" into "how I actually get my work done." This module is about adoption, not technology. ⏱️ Time: ~45 min, then ongoing.

The hard part was never the tools — it's the habit. Most people learn a new skill and then quietly drift back to the old way. Here's how not to.

What makes a task worth automating?

Not everything should be automated. The sweet spot is tasks that are boring, repetitive, and rule-based. Score a task against this quick checklist — the more boxes it ticks, the better a candidate it is:

☐ You do it more than once (weekly, or across many files at once)
☐ It follows rules ("rename to this format", "pull these fields")
☐ It's tedious — the kind of thing that makes your eyes glaze over
☐ A mistake is easy to make by hand (miscopying, missing a row)
☐ It doesn't need much judgment — or the judgment is a quick yes/no at the end

Three-plus checks? Automate it. Zero or one? Probably just do it by hand.

Not everything is worth automating. Score a task against four quick tests — does it repeat, follow rules, bore you, and need little judgment? Three or more yeses and it's a great candidate; one or none, just do it by hand.

The five-minute exercise: find your candidates

Do this once, today. It's the most valuable 5 minutes in this whole course.

Open a blank note.
List the document/data tasks you did last week — every one you can recall.
Next to each, mark how often it recurs: Daily / Weekly / Monthly / One-off.
Circle the Daily and Weekly ones that follow rules.

Those circles are your starting backlog. Pick the single most annoying one and automate just that, using the Recipe Cookbook and Module 09. Not your whole job — one task.

Start absurdly small

The classic trap is trying to automate an entire workflow on day one, hitting friction, and giving up. Don't. Automate one step.

Don't automate "my entire weekly report." Automate "merge these 12 PDFs."
Got that working? Next time, add "…and pull the totals into a spreadsheet."
Then "…and format it as the summary."

Each small win builds trust and momentum. In a month you've got the whole report automated — but you got there one safe step at a time.

Build a tiny bit of structure

A little organization makes the habit effortless:

One home for your work: a folder like ~/work-automation/, with a subfolder per recurring task (weekly-report/, intake-cleanup/).
Consistent in/out: each project has an inbox (drop files in) and a results folder. Put it in the project's CLAUDE.md (Module 11).
Save the recipes: anything you'll repeat becomes a note in CLAUDE.md or a slash command, so future-you just says "do the X."

Now the habit is: drop files in inbox → open claude → say the recipe name → check results. Almost no thinking required.

Beat the "it's faster to just do it myself" voice

The first time you automate a task, it is slower than doing it by hand — you're teaching Claude Code the rules. The payoff is the 2nd, 5th, 20th time.

A task you do weekly that takes 30 minutes is 26 hours a year. Spending an hour to automate it is one of the best trades you'll ever make.

So the question isn't "is this faster right now?" It's "will I do this again?" If yes, invest the setup once.

Track the wins (it keeps you going)

Keep a simple running note: "Automated the intake cleanup — was 40 min/week, now 2 min." Two reasons this matters:

Motivation. Seeing the hours add up makes the habit stick.
Visibility. When your company asks how you're using AI (and they will), you'll have concrete numbers — which is great for you at review time.

When not to automate

Good judgment includes knowing where to stop:

True one-offs you'll never repeat — just do them by hand (or have Claude Code do it once, live, without building a reusable tool).
High-judgment decisions — approving a claim, interpreting an ambiguous case. Automate the gathering and formatting; keep the deciding human.
Anything sensitive that breaks the data rules — see Module 10. When unsure, ask, don't automate.

Sharing with your team (carefully)

When a tool saves you hours, you'll want to share it. Great — with two cautions:

Share the recipe, not company data. The prompt and the approach are safe to pass along; the files they ran on are not.
Check policy first before colleagues install tools or before anything touches a shared system. Module 01's "ask IT" applies to them too.

✅ You're done with this module when

You've done the five-minute exercise and have a backlog of candidate tasks.
You've automated one real task from that list, start to finish.
You have a ~/work-automation/ home with at least one organized project.
You've started a running note of time saved.

Next: What is MCP? — optional background for later.