How I Use Claude Code and CoWork to Turn My Calls Into Email Ideas

I use one folder of my calls, videos, writing guides and past emails as the starting point for new email ideas. Claude Code brings the material together; CoWork helps me find recurring problems and angles, and I still choose and write the emails.

The useful part is having my own teaching in one place. The routine gives me source material to work from instead of asking an AI to invent a broadcast from a blank page.

One folder is the foundation

I keep everything in one folder inside my marketing and strategy coaching calls directory. It holds the material that shows how I teach, and Claude can only work with what I put in it.

I treat that folder as the real asset. If something I teach isn't in there, CoWork won't know about it when it suggests ideas.

• YouTube video transcripts

• Writing guides and principles

• HubSpot marketing emails

• Coaching call transcripts

Pulling my calls in with a routine

The first routine I set up grabs full transcripts of my group coaching calls from Fathom. The first time, Claude Code needs your API key so it can reach the platform. After that, the routine runs without you retyping it.

I set the routine to run daily at 5:00 p.m. Eastern, pull only calls I recorded, and save full transcripts as txt files rather than summaries. Each file is named with the meeting date and title and goes into a subfolder called all Daniel calls.

Claude asked me a decision about the existing history. I chose to backfill the last 90 days.

The catch is that it runs on my computer

In the walkthrough I used a local routine. If Claude or my computer is off, that routine does not run. My Mac usually stays open on my desk, so that tradeoff suited the setup I showed; unattended cloud execution would need a different setup.

Permissions are the other snag. Claude asks you to allow each action, and I had to switch the routine from manual to auto. If you don't, it gets stuck waiting on a prompt that nobody is there to answer.

I back up with Time Machine. Many of these routines are essentially Python scripts saved on the computer, so restoring a backup to a new Mac brings the setup back with it.

Asking for angles, not finished emails

Once the folder is populated, I point CoWork at it and ask for something specific. In one example I asked it to scan my past YouTube videos and come up with five email broadcast ideas, with subject line variations and bullet points, but not full emails.

I usually don't want AI writing the full email for me because I often find the result generic. I'd rather get the angle and structure, then write the email myself and borrow the phrasing that works.

That approach keeps me in charge. I still have to decide whether an idea is good enough to send.

What my calls keep revealing

When I had CoWork review recent calls, the recurring problems were pricing structure confusion, offers that can't sell to cold traffic, nobody calling their leads, misreading ad data, over-engineering media buying and beginners applying advanced advice too early.

Those patterns turn into email ideas: for example, comparing a recurring broadcast retainer with a one-time flow build. The suggestions in the walkthrough were drafts to evaluate, rather than evidence that a subject line or an offer had already worked.

I apply the same logic to YouTube. If my students are struggling internally, my audience is probably dealing with it externally too. Some suggestions were useful and others weren't relevant, so I still filter every one.

Turning 965 emails into an index

My HubSpot export includes about 965 emails, covering flows and every broadcast I've written. I treat that archive as raw material for an email marketing course based on what has actually worked for my list.

I can ask CoWork which subject lines, CTA counts, frequency and click-through rates have performed best. I can also ask which emails from a lead magnet or low-ticket product flow could be reused as broadcasts.

This is an index of my own past sends, not a formula that guarantees future results. It's still a much better starting point than trying to remember what worked.

Keeping the master lessons doc current

A scheduled task in CoWork reads everything in the same subfolder and updates a master lessons doc weekly. The doc is a long text file that summarizes my core teachings.

The doc is only as good as what goes into it. If the calls and videos are incomplete, the gaps carry straight into the summary, so I keep the inputs complete.

Where I'd start

Start with one folder and one routine that pulls in something you already produce, like calls or transcripts. Make sure the routine runs on a machine that will actually be on, and set the permissions before you walk away.

Then ask narrow questions about recurring problems, past email performance and video angles. Those are easier to act on than a blank page, and they keep the final decisions with you.

Client Ascension related reading: the email ideas guide discusses client-call transcripts, broadcasts and sequences.

Adapted by Client Ascension with AI assistance from Daniel Fazio's lesson. The first-person narrative follows the source speaker’s account. Watch the original video