I have a confession.
I subscribed to Claude Pro because I was tired.
Not tired of writing. I actually love writing. But I was tired of the gap between having an idea and getting it published. The drafting. The formatting. The thinking about which platform to post it on. The switching between tools. The time it takes before a thought becomes a live article on my blog.
I run mazlanabbas.com. I also post on LinkedIn, Threads, Facebook, and TikTok. Every single day, content needs to go out. And every single day, I am the one doing all of it manually.
There has to be a better way, Mazlan.
That thought had been sitting in my head for weeks before I finally did something about it.
So I subscribed to Claude Pro.
And on day one, I had a very specific mission in mind.
I wanted to build an AI agent that could take a topic, generate a full blog article in my voice, and publish it automatically to mazlanabbas.com. Every day. Without me having to touch every single step manually.
Simple enough, right?
Not quite.
What I Thought Would Happen
I imagined the workflow clearly in my head before I even started.
I give Claude a topic. Claude writes the article in my voice. The agent formats it properly. It gets pushed to my WordPress blog. It goes live. Done.
I even imagined waking up in the morning to find a fresh article already published. My blog running while I sleep.
Mazlan the visionary, I told myself.
The reality was a little more humbling.
What Actually Happened
The first thing I did was describe my vision to Claude. I explained that I wanted a workflow that could generate blog content daily and publish it to my WordPress site automatically.
Claude understood immediately. It walked me through the concept of what such an agent would need. A content generation layer. A formatting layer. A publishing layer that connects to the WordPress REST API. And a trigger mechanism to run the whole thing on a schedule.
I nodded along reading all of this.
Yes, yes, yes. This is exactly what I want. Now let us build it.
Then Claude asked me a question that stopped me.
“What does your voice actually sound like? Can you share some examples of your existing articles?”
I paused.
I had been so focused on the automation that I had skipped the most important ingredient. The voice. My voice. The thing that makes my content mine and not just another generic AI article that nobody wants to read.
That was the first lesson of day one.
Automation without identity is just noise.
So I went back and started properly. I shared examples of my writing from mazlanabbas.com. I described how I write. Short sentences. Self-dialogue. Personal stories that connect to bigger ideas. First person. Honest, not polished.
Claude absorbed it and started generating content that actually sounded closer to me. Not perfect. But closer.
That part worked better than I expected.
Where It Got Complicated
The content generation side was promising. The publishing side was where things got real.
Connecting an AI agent to WordPress through the REST API is not impossible. But it is not a one-click setup either. There are authentication steps. API credentials. Formatting requirements for how WordPress receives a post. Categories, tags, featured images, post status.
I spent a good part of that first day just understanding what the connection would actually require.
Okay Mazlan. This is not a morning project. This is a proper build.
And I had to be honest with myself. I was exploring. I was not building yet. There is a difference.
I had the concept. I had the direction. I had Claude helping me understand the components. But the actual working agent that publishes daily to my blog automatically? That is still a work in progress.
And I think that honesty matters.
What I Did Get Working
By the end of that first day, I had something useful even if it was not the full vision.
I had a content generation workflow inside Claude that could take a topic I give it, write a full blog article in my voice, and format it ready for publishing. That part works.
What I have not yet completed is the automated publishing step. The piece that removes me from the middle and lets the agent push content to mazlanabbas.com on its own.
But I learned something important about that gap.
The gap between “Claude can write it” and “the agent can publish it” is not a technology gap. It is a design gap. I need to think through the full workflow properly. What triggers it. What happens if the content is not good enough. Whether I want a human review step before anything goes live. What happens to the social media versions of the same content across LinkedIn, Threads, and TikTok.
These are not technical questions. They are design questions. And until I answer them clearly, no amount of coding will give me the agent I actually want.
What Day One Really Taught Me
I started the day thinking I would build something.
I ended the day realising I first need to design something.
That is not failure. That is exactly how real building works. You start with a vision, you encounter the real complexity, and you learn what the actual problem is before you can solve it.
For me, the real problem is this. I want an AI-powered content system that generates and publishes daily content in my voice across my blog and social media platforms, automatically and consistently, without sacrificing the quality and authenticity that my audience expects.
That is a bigger and more interesting challenge than I thought on day one.
And I am still in the middle of solving it.
If you are on a similar journey, trying to use AI to simplify your content workflow, I want to hear where you are. What have you tried? What is working? What is still broken for you?
Let us figure this out together.
Discover more from Dr. Mazlan Abbas
Subscribe to get the latest posts sent to your email.
Dr. Mazlan Abbas 