Every morning, before the editorial team starts their day, I do one job: decide what we should write today.
For the past six months, this took me 30 to 60 minutes daily. Now it takes 5. Here’s what changed, and what didn’t.
The old workflow
My morning routine looked like this:
- Open Google Analytics and check the top-performing pages.
- Open Search Console and check the top search queries, top pages and Discover performance.
- Open AdSense and see which pages are actually making money.
- Check what competitors are publishing.
- Bring everything into my Claude thread, add my own observations, and ask for article ideas.
- Go through 25+ suggestions and pick the best 5–10.
- Send them to the editorial team on WhatsApp.
Over time, I built a memory bank for this. I taught Claude what to look for, what outcome I needed, and how I make decisions. I gave it my insights every day based on my own reading of the data, along with a set of decision models and strategies I’ve built over the years. So the analysis part got sharper every week.
But the work around the analysis stayed the same.
What was actually eating the time
The thinking was never the problem. The clicking was.
5 browser tabs. Obsidian. WhatsApp. Changing date ranges. Copying tables. Pasting them into Claude. Formatting. 50 to 100 clicks every single morning.
None of this needed my brain, but all of it needed my time.
The new setup
I split the work into two parts and gave each part to the right tool.
Manual actions → Open Bot with a local model
Opening reports, changing date ranges, copying the top pages, queries and revenue data, then pasting it all into my Claude thread. This is repetitive, low-thinking work, and a local model handles it fine.
Analysis → Claude Opus
The same Claude thread I’ve been training for months. It already knows my decision models, so it reads the fresh data and gives me 10 article ideas for the day, each with a Tamil title and a one-line reason why it should be written now.
Delivery → Telegram
The final list arrives on my phone, neatly formatted. I read it, filter it, and forward the best ideas to the editorial team.
No browser tabs. No copy-paste. No clicks.
What I kept for myself
The final decision.
The bot collects the data. Claude does the analysis. But I still pick what goes to the editorial team. That’s the part that needs editorial sense, knowledge of the audience and a feel for what’s right for the publication, and I don’t want to automate that.
The result
From 60 minutes to 5 minutes, every day.
That’s roughly 4 to 5 hours saved every week, and I get the ideas earlier in the morning, so the team can start sooner.
What I’d tell other publishers
Look at your daily routine and separate two things: the thinking and the clicking.
Automate the clicking completely. Use a cheaper or local model for it. Use your best model only where real analysis is needed. And keep the final judgment with a human.
That’s where the real time savings are.
Want to automate your workflow?
If you’re spending hours every day on repetitive data work, or you want a clear content strategy for your publication, mail me at thamiz@iniyan.in. Let’s find what you can automate and what you should keep for yourself.