Free

818 Automations, 8 Verbs: What People Actually Build

I went through 818 real automations people built and published. Almost all of them do one of 8 things. Once you see the list you'll spot 3 in your own week.

5 min read · Last verified 2026-09-10

"Automation" is one of those words that means nothing until somebody shows you one.

So I went and read 818 of them. Real ones, built and published by real people, sitting in a public library anybody can browse. I wanted to know what people are actually building, not what the AI newsletters say they're building.

Here's what I found, and it made the whole thing a lot less intimidating.

818 automations. Basically 8 jobs.

That's it. Different tools, different industries, wildly different setups, and underneath they're nearly all doing one of 8 things. Once you can name the 8, you stop seeing a scary technical category and start seeing your own Tuesday.

The 8 jobs

1. Watch something and tell me. A folder, an inbox, a form, a price, a search ranking. Something changes, you get told. This is the most common one by a mile and it's the easiest to build.

2. Take what came in and file it. An email arrives, the details end up in a spreadsheet. A form gets filled, a row appears. No retyping, no forgetting.

3. Read the long thing and tell me the short thing. A recording, a PDF, a 40 minute call, a webpage. You get the 6 lines that matter. Enormously popular, and honestly the one that gives you the most time back per hour of setup.

4. Pull the useful bits out of a mess. Invoices, receipts, screenshots, a pile of notes. It finds the numbers and the names and puts them somewhere structured.

5. Write the thing that always needs writing. The follow-up, the description, the post, the reply you type 40 times a week with 3 words different.

6. Send it where it needs to go. Once the thing exists, it posts, emails, texts, or files it. No human clicking publish.

7. Go get information from out there. Reviews, listings, competitor pages, news. It goes and collects, on a schedule, without you.

8. Turn it into a different format. Video to text, text to audio, audio to notes, a doc to a post, a call to a summary.

That's the whole category. Every "AI automation" you've been scrolling past is a chain of 2 or 3 of those.

The part that surprised me

I expected exotic tools. The most common things in the entire library were OpenAI, Google Sheets, ChatGPT, Google Drive, Gmail, Airtable, and Telegram.

Google Sheets shows up in about 1 in 5.

That's it. That's the stack. A spreadsheet, an inbox, a drive folder, and an AI to think about what's in them. You almost certainly already have all of that.

The barrier was never the tools. It was that nobody showed you the 8 verbs.

Rest of it's yours

Drop your email and the whole thing opens up right here. No waiting on a confirmation, no download, nothing to install.

You'll also get the new ones as I make them. Leave whenever, no hard feelings.

Finding yours

Don't go looking for a cool automation. Go looking for an annoying Tuesday.

Answer these 3 and you'll have your first one:

What do I retype? Same information, moving from one place to another by hand. Job 2.

What do I read that I resent reading? Long documents, recordings, threads where 5% matters. Job 3.

What do I check? A thing you open 6 times a day to see if something changed. Job 1.

Whichever answer made you sigh, that's the one. Start there.

Build it in the chat first

Here's the part almost nobody tells beginners: you don't need automation software to start.

Before you go sign up for anything, do the job manually in Claude 3 times. If it can't do it in a chat window with you watching, it definitely can't do it unattended at 4am.

I want to automate part of my week and I want to test it by
hand first.

Here's the job: [describe the annoying thing in plain words]
Here's a real example of the input: [paste an actual one]

Do the job on this example now, and show me the output exactly
how I'd want it to look.

Then tell me: what would go wrong if this ran automatically
100 times without me watching?

That last question is the important one, and it's the step people skip. It'll tell you honestly where it breaks. Now you know what you're building instead of finding out after it's live.

Run it on 3 real examples. If all 3 come back right, you have something worth automating. If 2 of 3 are wrong, fix the instructions before you build anything, because automating a bad process just gets you wrong answers faster.

Then make it run without you

Once it works by hand, you need 2 things: a trigger and the steps.

The trigger is what starts it. New email. New file. Every morning at 7. A form submission.

The steps are the job you already tested, in order.

Tools that do this without code: Zapier (easiest, most expensive), Make (cheaper, more powerful, steeper), n8n (free if you self-host, the one most of that 818-automation library is built on). Claude can also run things directly through connectors, which for a lot of jobs is enough on its own. That's here.

Pick one and stay in it. They all do the same 8 jobs, and switching tools 3 times is how a Saturday disappears.

Where people go wrong

I read a lot of these. The pattern in the ones that clearly got abandoned:

Too big for a first build. They tried to automate their entire sales process instead of one email. Build the smallest useful piece. One trigger, 2 steps. Get that running, then add to it.

No human check on anything that leaves the building. Automations that quietly send things to customers are how you find out about a bug from a customer. Put a draft step in. Let it prepare everything and let you press send, at least for the first few weeks.

Automating something they didn't actually do. If you weren't following up with old customers before, an automation doesn't create the habit. It just fails on a schedule. Automate the thing you already do and resent.

The honest bit

Your first one will take way longer than you think. Mine took an evening for something that saves maybe 20 minutes a week.

That math looks bad until you notice the second one took an hour, and the fourth one took 15 minutes. You're not building an automation, you're learning a shape. Everything after the first is fast, and that's the actual reason to do it.

Start with the thing you check 6 times a day. That's job 1, it's the easiest, and you'll have it working tonight.

← Back to all the free stuff