I built an AI content pipeline that publishes a media accountability channel every day across three platforms. Here’s exactly how it works, and what it took to get there.
The channel is called The Buried Lede. The format is simple: one media outlet, one thing they got wrong, two links, their article and the primary source. Every day. Substack, Instagram, TikTok.
Here’s the system:
**Four automated workflows, one human checkpoint**
Every Monday: An n8n workflow pulls from six RSS sources, cross-references a curated evergreen story bank, checks which content series is active, and sends a structured brief to Claude. Output: 7 stories selected for the week, saved to Airtable, one per day.
Every morning: A daily workflow fires, pulls today’s queued story, and sends me a single Telegram message. I reply GO or SKIP. That’s my editorial checkpoint, 30 seconds, every day, non-negotiable.
On GO: A production workflow runs three sequential Claude API calls, deep-dive research into the primary source gap, multi-platform writing (Substack/Instagram/TikTok) in a defined editorial voice, and an editorial image prompt. Replicate generates the illustration. Everything saves to Airtable. Done in ~3 minutes.
14 days post-publish: A ratings workflow asks me to score the post 1-5. Scores feed a performance weights table that tracks which beats and outlets resonate most. Closes the feedback loop.
**The thing most people get wrong about AI content automation**
They automate volume. They don’t automate judgment.
The Buried Lede works because the editorial architecture came before the automation. A defined voice. A strict format. A 13-series content roadmap spanning 56+ weeks. Rules about sourcing, tone, which topics are off-limits.
The AI executes that architecture. It doesn’t replace it.
If you feed an AI vague instructions and a topic list, you get generic content that sounds like content. If you feed it a precise editorial brief built on genuine creative decisions, you get something that sounds like you.
**The technical decisions that actually matter**
– Use Airtable field IDs not display names, rename columns without breaking workflows
– Handle timezones explicitly, my server runs UTC-4, my editorial day is UTC+7, that 11-hour gap will destroy your daily workflow if you ignore it
– Guard empty array PATCH calls, Airtable returns 422 errors on empty records arrays, an IF node saves you
**Who this applies to**
Solo creators who are burning out on production overhead while trying to maintain daily publishing. Content agencies managing multiple client channels who need a repeatable production architecture. Media companies exploring AI-assisted editorial workflows without sacrificing voice or quality.
## What This Looks Like for Your Channel
The architecture is channel-specific but the pattern is universal. Any editorial operation that publishes regularly, has a defined voice, and needs to maintain quality without a large team can use this approach. What it requires:
– **A defined editorial identity**: voice, format, content rules, what you will and won’t cover
– **A content architecture**: series, themes, rotation, long-term arc
– **A database layer**: Airtable, Notion, or any structured database with an API
– **A workflow engine**: n8n for complex pipelines; Zapier and Make for simpler ones
– **A human checkpoint**: don’t automate yourself out of the editorial seat
For content agencies managing multiple client channels, this scales cleanly, each channel gets its own Airtable base, its own editorial brief, its own series rotation and performance tracking, all running on shared infrastructure.
—
## The Honest Assessment
Building this took time. Debugging the edge cases, timezone handling, empty array errors, Telegram parse mode, malformed JSON from AI responses, took longer. You need patience with the technical side, or a collaborator who understands both the automation tooling and the editorial requirements.
But the output is a channel that publishes every single day, maintains a consistent voice, follows a coherent long-term editorial arc, generates its own images, tracks its own performance, and routes me one decision each morning that takes 30 seconds.
The Buried Lede is a media accountability channel. It also happens to be a proof of concept for what AI-powered content automation looks like when the editorial strategy comes first.
—
*If you’re building something similar, or want help designing this kind of system for your channel or your clients’ channels, reach out. The architecture described here is adaptable, and the lessons from getting it wrong are worth more than the code.*
*The Buried Lede publishes daily on Substack, Instagram, and TikTok*