Building AI Second Brain: How I Turn Voice Memos Into Substack Notes and Business Ideas

The ultimate guide for building an AI voice-powered second brain.

I talk to myself in public constantly during afternoon walks.

People think I’m crazy. I can see it in their faces when they pass me: this guy muttering about AI workflows and newsletter strategies like he’s having an argument with himself.

But here’s what’s actually crazy:

I spent months perfecting my voice memo capture system: automatic transcription through Wisprflow, instant save to Apple Notes, a smooth way to catch every thought before it slipped away.

Only to create a graveyard of 100+ good ideas I’d never touch again.

The irony wasn’t lost on me. I’d solved capturing ideas so well that I made the real problem worse.

Capturing ideas was never the problem. Knowing what to do with them was.

Every voice memo I recorded added to an invisible pile of “I’ll process this later.” And “later” meant opening each transcript, copying it into ChatGPT, asking AI to extract insights, then manually formatting and saving everything to Notion.

15 minutes per voice memo. Just to organize my own thinking.

When you’re recording 3-5 thoughts daily, that’s over an hour of manual processing. And even then, I’d only get one output. Maybe a Substack note. Maybe a rough idea. Nothing systematic.

So those 100+ voice memos just sat there:

• Brilliant observations about AI automations buried in memo #12

• Content angles that felt electric when I captured them, forgotten in memo #23

• Business ideas I was genuinely excited about, rotting in memo #34

The system I’d perfected was sabotaging me. More in, nothing out, guilt piling up with every note.

I needed the change to be as effortless as the capture process. Otherwise, what’s the point?

If I could capture thoughts easily but couldn’t turn them into content, I wasn’t building a second brain. I was building a productivity graveyard disguised as organization.

Turning AI voice memo into my second brain

So I built an AI automation that processes voice memos automatically:

I record a thought on Wisprflow during my walk. It transcribes instantly and emails the transcript to my automation. AI analyzes it and transforms it based on what I need:

1. “Substack” in the subject line → 10 strategic Substack Notes, each with different engagement angles and formats

2. “Newsletter” in the subject line → Complete newsletter brief with hook, angle, and why-it-matters structure

3. “Business” in the subject line → 3-5 validated business concepts with implementation frameworks and next steps

And regardless of what the transcription contains, it organizes thoughts into eight categories, structured, and ready for whatever I need later.

By the time I sit down at my laptop, everything’s already organized in my Notion database. No manual processing. No formatting. No “I’ll deal with this later.”

It’s ready-to-publish content waiting for me.

The system runs itself.

What used to take me an hour per voice memo now happens automatically in minutes, running in the background without me noticing.

More importantly: there’s no friction between capture and creation anymore. The system handles everything between “I have this thought” and “this is ready to publish.”

Here’s proof it actually works:

A few weeks ago, I was walking and rambled this into my phone:

“I think more and more people should learn how to automate their job using AI because they will learn that automating their job is much harder than actually doing the job. You need to understand the thought behind it, the things that you care the most, the quality that you really want. It’s much harder to discern this type of thinking rather than just doing the thing compared to doing the thing. Does it make sense?”

Messy. Unclear. Classic thinking-out-loud voice memo.

By the time I got home, the system had already processed it into 10 Substack Notes.

Here’s one:

So yeah, I still talk to myself in public. People still judge me.

But now those conversations actually turn into content.

This guide shows you how to build the exact system I use every day: a complete Make.com blueprint, all three AI prompts (Substack Notes, Business Ideas, Newsletter Brief), AI-generated memo organization, Notion database structure, and the Wisprflow workflow.

By the end of this post, you’ll have a second brain that actually works instead of another productivity graveyard.

---

🚨 Before we move on, we’ll keep using Make.com to build this AI automation because it’s reliable and cost-effective (FREE or cheaper than alternatives). If this is your first time with Make.com, check my previous posts—linked “here” and “here”—for basic setup and configuration. I won’t repeat the module steps covered there, so we can focus on what matters next.

What you’ll need before you can run this automation:

• Wisprflow account (free tier works)

• A Notion page + 4 databases (I’ll show you the format)

• OpenAI API key with credits worth of $5

• Make.com account (free tier works fine for this)

---

I hope you don’t get overwhelmed by the workflow, because it’s a repeated module throughout. Once you’ve nailed one workflow, it’s easier to keep going.

I’m going to explain this from the top branch all the way to the bottom.

Let’s dive in.

1. Install your AI voice transcriber app

I wish Wisprflow had promoted this post, but this has been my go-to voice transcriber app for almost six months. Feel free to try other options such as Voicenotes or Monologue by Every.

Separately, Voicenotes has webhook integration, where you can generate voice notes and connect them with Make.com automation directly without mailhook. You might want to check out Jonas Braadbaart’s viral post on building his “second brain.”

Since Wisprflow doesn’t support webhook integration (cmiiw) yet, we need Mailhook to activate this automation. I believe they’ll ship webhook support soon, so we won’t have to rely on Mailhook.

2. Set up your first module: Custom Mailhook

Go to Make.com and set up your first module: Custom Mailhook. If this is your first time, you might want to check out this post to learn more.

This automation starts by sending your voice transcription via email.

1. Open Gmail

2. Compose a new message

3. Use Wisprflow from your keyboard to create the content

4. Set the subject line, enter the mailhook address as the recipient, and send.

Once sent, the automation will run end to end.

3. Activate your router

What is Router and why do we need one?

Quick note before we dive in: this AI second brain has four outputs: Substack notes, newsletter brief, business ideas, and thoughts organization. And each goes to a different Notion database.

To keep it all running, we set up four OpenAI modules and four Notion databases.

The router simply sends each use case to the right place so the output is processed as intended.

You should know we have control over which outputs we want to produce with AI. If your voice memo fits “business ideas” but not “newsletter brief,” you can tell the Router through the email subject line which one you want AI to proceed with.

In the Router module, we can set up a filter with condition based on the email subject line to choose any of them—or all of them—by mentioning: substack, business, newsletter, or memo.

If you write anything in the subject line that doesn’t match those keywords, it will fall back to “memo.”

4. OpenAI module for Substack Notes

Now that we need to branch out to multiple outputs, let’s tackle the Substack Notes router first.

Here’s the system prompt you can copy and paste: