Choosing What to Automate
Why this matters
Not everything should be automated. Some tasks take 2 minutes and happen once a month. Building an agent (an AI that runs on its own) for that would take 2 hours to set up. That's a terrible trade.
This chapter teaches you to pick the right battles. The difference between someone who saves 10 hours a month with automation and someone who wastes 10 hours building automations they don't need? They picked different tasks to automate.
The Automation Test
Before automating anything, ask yourself these four questions. If a task doesn't pass this test, leave it manual. Seriously.
1. How often do I do this?
Daily = gold. This is your first target. If you do something every single day, even small time savings compound fast.
Weekly = good. Still worth automating if it takes more than a few minutes each time.
Monthly = maybe. Only if it takes a decent chunk of time and follows the same steps every month.
Yearly = don't bother. Just do it manually. The automation will probably break before you use it again anyway.
2. How long does it take each time?
20+ minutes = automate this. That's real time you're losing.
5-20 minutes = worth considering, especially if it happens daily or weekly.
Under 2 minutes = definitely not worth it. The setup time alone will take longer than a year's worth of doing it manually.
3. Is it the same every time?
Yes, identical steps = perfect for automation. Computers love doing the same thing over and over.
Mostly the same with small variations = still good. You can build in flexibility.
Different every time = harder to automate well. You'll spend more time fixing edge cases than you save.
4. What happens if it goes wrong?
Low risk = great candidate. Drafting an email you review before sending? If the draft is wrong, you just don't send it. No harm done.
Medium risk = automate with safeguards. Always include a human review step.
High risk = be very careful. Sending a client the wrong invoice or the wrong amount? That damages trust. Keep high-risk tasks manual or add multiple checkpoints.
Score your tasks
Grab a piece of paper or open a spreadsheet. List your top 10 recurring tasks. Score each one against those four questions. Here's an example:
| Task | Frequency | Time | Same each time? | Risk | Verdict |
|---|---|---|---|---|---|
| Check for overdue invoices | Daily | 15 min | Yes | Low | AUTOMATE |
| Reply to new enquiries | Multiple/week | 10 min each | Mostly | Low | AUTOMATE |
| Write a client proposal | Monthly | 2 hours | No | High | DON'T |
| Reply to a complaint | Rare | 30 min | No | High | DON'T |
| Weekly client update emails | Weekly | 20 min | Mostly | Medium | MAYBE |
The tasks that score "daily or weekly + 15 minutes or more + same every time + low risk" are your first targets. Everything else can wait.
Real examples
You do this daily. It takes 15 minutes of logging into Xero, scanning the list, and figuring out who to chase. It's the same steps every time. And it's low risk because you're drafting reminder emails, not sending them automatically. This is a perfect automation target.
You get several a week. Each one takes 10 minutes to read, think about, and write a reply. The format is similar every time. And if the draft isn't quite right, you just edit it before sending. Low risk, high frequency, similar format. Automate it.
This happens a few times a month. It takes real thought. Every proposal is different because every client is different. And if you get it wrong, you look unprofessional. This is a task where AI can help you (as a tool), but you shouldn't fully automate it.
Complaints are rare, each one is unique, and they need genuine empathy. An automated response to an unhappy customer is the fastest way to lose them forever. Handle these yourself.
The 80/20 rule
Here's the truth that saves you from going down a rabbit hole: 80% of your time savings will come from automating just 2-3 tasks.
That's it. Two or three. Not ten. Not twenty.
Start with the top 2 from your scored list. Get them running perfectly. Live with them for a couple of weeks. Then, and only then, add one more.
People who try to automate everything at once end up with a pile of half-working automations that break constantly. People who automate 2 things well end up with a business that runs smoother every month.
Don't spend 3 hours automating something that takes 5 minutes. That's not saving time, that's a hobby. If the setup takes longer than a month's worth of the manual work, it's not worth automating yet. Do the maths before you start building.
What you just learned
- Not every task should be automated. Score them first.
- The four questions: frequency, time, consistency, and risk.
- Start with 2-3 high-impact tasks, not everything at once.
- The 80/20 rule saves you from over-engineering.
Try it yourself
- List your 10 most recurring business tasks
- Score each one on frequency, time, consistency, and risk
- Pick your top 2-3 automation targets
- Write down what "success" looks like for each one