Automate the Boring, Guard the Important

Automation is a sharp knife.

Great for onions.

Terrible for fingers.

The sharp kitchen knife

A sharp knife makes cooking faster.

Chopping onions takes seconds.

But nobody juggles a sharp knife.

And nobody uses one without looking.

The sharper the tool, the more careful you become.

AI automation is a very sharp knife.

The simple rule

I sort every task into two boxes.

Reversible: if it goes wrong, you can undo it.

• Drafting an email
• Summarising a document
• Creating a report
• Preparing a spreadsheet

Irreversible: once it's done, it's done.

• Sending an email
• Posting publicly
• Paying money
• Deleting files

Automate the first box freely.

Guard the second box with a human "yes".

Why this matters

Imagine an automation that replies to every customer email.

Most days it's fine.

Then one day it misreads an angry complaint and replies,

"Thank you so much for your kind feedback! 😊"

You can't unsend that.

Write the guardrail into the instruction

Bad:
Reply to customer emails automatically.
 
Good:
Every morning, read new customer emails and
draft replies. Do not send anything. Put the
drafts in a list for me to review, and flag
any email that sounds angry or mentions a refund.

The AI still does 90% of the work.

You do the 10% that could go wrong.

Good assistants ask first

A well-designed assistant should ask before it sends, posts, or deletes.

That isn't a weakness.

It's like a junior colleague saying, "I've drafted this. Shall I send it?"

That's exactly the colleague you want.

Final thoughts

The goal isn't maximum automation.

It's safe automation.

Let AI chop the onions.

Keep your hand on anything that can't be undone.

Your fingers, and your customers, will thank you.