top of page

How I Built an AI Journal That Saves Itself to Google Drive

Writer: Benjamin Tassin
Benjamin Tassin
Oct 20
4 min read


What's happening

On October 6, Anthropic opened Claude for Google Workspace to all paid Claude users as a public beta. It puts Claude in a sidebar inside Google Docs, Sheets, and Slides. It also adds connectors so Claude can work on those files from a normal chat when you paste in a link.

Translation: your AI and your Google Drive are getting a lot closer. That opens the door to small, personal systems that just run in the background. My journal is one of them.

What my Journal skill does

I have wanted to keep a journal for years. The problem was never the stories. It was sitting down to write them.

So I stopped writing. Now I just talk.

I tell Claude about my day (voice mode when I am on the go, or typed if I am somewhere loud, like a plane). When I am done, I say "save this as a journal entry." Claude does the rest:

1. Creates a brand new Google Doc for that one entry.

2. Names it with the date and time, like "2026-10-19 1830 - Journal Entry," so everything sorts in order.

3. Puts the date as a heading, a line with the time and whether it was voice or typed, then my words.

4. Saves it into one Journal folder in my Google Drive.

5. Goes back and checks that the doc actually landed before it tells me it is saved.

The rules are the real product

Anybody can get an AI to make a Google Doc. The rules are what make it trustworthy:

1. Talking is not saving. If I am just telling a story, Claude can ask "Want me to save that?" It never saves without a yes.

2. My words stay raw. No polishing them into tidy AI prose. If voice-to-text gets a word wrong, Claude can point it out, but it does not quietly rewrite it.

3. Claude's notes are labeled. If it adds context, it is tagged "Claude's addition" so I always know which words are mine.

4. Nothing gets overwritten. A correction becomes a new entry that points back to the old one.

5. No duplicates. Before saving, it checks today's entries so the same story does not show up twice.

Why one doc per entry?

When I built this, Claude's Drive connection could create files but not edit what was inside them. Instead of fighting that, I designed around it- one new doc every time.

Turns out that is the better design anyway. Nothing gets overwritten. Every entry has a date. And later I can search the folder ("what was I thinking about when I started AIM?") instead of scrolling through one giant file.

It also costs nothing extra. No Zapier, no paid automation steps. Just Claude and Google Drive.

How to set it up

1. Connect Google Drive to Claude. Look for it in the connectors section of Claude's settings. (This is a paid-plan feature.)

2. Make one folder in Google Drive for your journal. Name it whatever you want.

3. Paste the prompt below into Claude and put your folder name in the brackets. Claude will ask a few questions and build the skill with you.

4. Test it. Tell a short story, say "save this as a journal entry," then open the folder and look.

5. Use it for a week before you tweak anything.

Try this prompt

Help me build a Claude skill called "voice-journal." When I say "save this as a journal entry" (or something close), take the story I just told you, spoken or typed, and save it as a brand new Google Doc in my Google Drive folder named "[YOUR JOURNAL FOLDER NAME]". Follow these rules: 1. Never save unless I say the trigger phrase, or you ask and I say yes. 2. One new doc per entry. Never edit an old one. 3. Name each file "YYYY-MM-DD HHMM - Journal Entry" using my local time. 4. At the top of the doc put the date as a heading, then a line with the time and whether it was voice or typed, then my words. 5. Keep my words exactly as I said them. Do not polish them. 6. If you add anything, label it "Claude's addition." 7. Before saving, check today's entries so you do not save a duplicate. 8. After saving, open the folder and confirm the doc is there before you tell me it worked. Ask me any questions you need, then write the skill.

The AIM take

This is not really about journaling. It is about a pattern I use everywhere: say a trigger phrase, let the AI do the filing, then make it prove the work is done.

As an engineer, I do not trust "done" until I see it. Your AI should not either. That one rule (check before you report) is the difference between a tool you hope worked and a tool you can count on.

The same setup works for job-site notes, client call recaps, inspection logs, or daily crew updates. The tool is the easy part. The rules are the product.

If you have a process living in your head or on sticky notes, I can help you turn it into something like this. Book a free workflow review at intellmanage.com.

 
 
 

Comments


bottom of page