How does the four-agent content pipeline work?

The pipeline uses Cass as the orchestrator, Scout for research (pulling relevant sources and context), and Quill for drafting the article. A fourth production step handles formatting, metadata, and PR creation. The entire flow from brief to published draft takes around 30 minutes for a standard business article.

What is the human role in an AI-assisted content pipeline?

Humans provide editorial judgment: choosing which ideas are worth pursuing, verifying whether the research is accurate, and deciding whether the draft is good enough to publish. The mechanical work (research retrieval, structure, formatting) is delegated to agents; the quality bar remains with the editor.

Can this content pipeline approach be replicated by other UK businesses?

The specific pipeline uses OpenClaw and custom agents, but the underlying structure (brief, research, draft, production) can be replicated using off-the-shelf tools. The key design principle is separating each stage so that failures are isolated and quality can be reviewed at each handoff rather than discovered at the end.

What types of content work best for AI-assisted pipelines?

Content that follows a consistent structure and is grounded in verifiable sources works best: business analysis, how-to guides, event coverage, and industry commentary. Open-ended opinion pieces or content requiring original reporting benefit less from automation because the research and judgment steps are harder to delegate.

From TikTok to Published: How I Turn Content Ideas Into Business Articles

I watched a TikTok about dirt paths and turned it into a business article in 30 minutes. Here’s how my AI pipeline does the heavy lifting.

The Spark

It started with a 60-second video from @dspacestv. Two point three million views. The concept: desire paths: those worn tracks that form when people ignore official routes and take shortcuts instead.

You have seen them. The muddy track cutting across a park corner. The worn grass between office buildings. The gap in a hedge where the official gate sits fifty metres away. These paths emerge because the planned route doesn’t match how people actually move.

Ohio State University built their campus, waited a year for desire paths to form, then paved them. No guesswork. No “we think people will walk here.” Just observation followed by action.

I saw the business angle immediately. This applies to AI implementation in UK SMEs. Most businesses design processes around how they think work happens. The smart ones find where work actually happens and build there.

I sent the link to Cass, my AI orchestrator. Then I got on with my day.

The Pipeline

Cass took over. First, she resolved the TikTok URL and scraped the video: video, subtitles, full transcript. Everything saved to the knowledge base. This matters because I don’t want to rely on my memory of a video I watched once. I want the source material preserved, searchable, available for future reference.

Then she spawned Scout, my research agent. Scout’s job: find the business angle. He surfaced the design thinking concept “pave the cowpaths” from Steven Champeon in 1999. He found business examples: Twitter hashtags invented by users, not designed by the platform team. Instagram’s pivot from check-ins to photos when they saw how people actually used the app. Slack emerging from an internal gaming tool because the team needed better communication.

He identified the SME angle: shadow IT. Employees using unauthorised tools because official ones don’t match their workflow. The spreadsheet that lives on a USB stick because the ERP system is too slow. The WhatsApp group that handles customer service because the ticketing platform is clunky. These are desire paths in digital form.

Scout delivered a research brief in minutes. Cass reviewed it, then briefed Quill, my content agent.

The Writing

Quill’s brief was specific: “Desire paths as a lens for AI implementation. Find where people actually work, not where you think they work.”

He wrote in my voice. British English, direct, no fluff. He used the research Scout provided: the TikTok, the 2.3M views, the Ohio State example, the business cases. He didn’t invent statistics. He didn’t pad. He wrote what the research supported.

This is where most AI content falls down. People ask ChatGPT to “write a blog post about AI” and get generic mush. The difference is the brief. I don’t ask Quill to write about desire paths. I ask him to write about desire paths as a lens for AI implementation in UK SMEs. Specific angle. Specific audience. Specific purpose.

The draft came back clean. I reviewed it, suggested a few tweaks, and Quill revised. Two rounds. Fifteen minutes total.

The Production

Cass generated a cover image, committed the article to the website repository, and raised a pull request. The whole thing, from TikTok link to PR, took roughly 30 minutes.

I reviewed the preview. The article looked right. I merged.

Done.

Why This Works

The pipeline handles the mechanical work. Scraping. Transcription. Research. First draft. Image generation. Deployment. I handle the judgment: spotting the angle, reviewing the tone, deciding when it’s ready.

This division of labour is deliberate. I’m good at pattern recognition and taste. I’m not good at transcribing videos or formatting markdown. The agents handle what they do best. I handle what I do best.

This isn’t about replacing thinking. It’s about removing friction. I still need to recognise which content has business value. I still need to know what “sounds like me.” But I don’t need to spend hours transcribing videos, hunting for sources, or wrestling with publishing tools.

The agents don’t write instead of me. They write so I can write more.

The Lesson for SMEs

Most UK business owners I speak to think AI will replace their judgment. It won’t. What it replaces is the drudgery that stops you using your judgment.

I see this constantly. Business owners spending hours on tasks that don’t require their expertise. Formatting documents. Chasing information. Wrestling with tools. Meanwhile, the strategic work, the thinking that actually moves the business forward, gets squeezed into whatever time remains.

I could have watched that TikTok, thought “that’s interesting,” and done nothing. Instead, I had a published article 30 minutes later. Not because I’m special. Because I built a system that converts ideas into output without me managing every step.

Desire paths, remember? Find where the work actually happens. Build there.

My work happens in the ideas: spotting connections, deciding what matters, refining the message. The pipeline handles everything else.

That’s the real lesson. AI doesn’t make me a better writer. It makes me a writer who actually publishes.

Sources and References

@dspacestv desire paths TikTok: The video that sparked the article. @dspacestv on TikTok, Designing Spaces channel covering design history and architecture. View count (2.3M) as of article publication date.
Desire paths concept: The term is established in urban planning and landscape architecture. See: Wikipedia: Desire path and 99% Invisible: “Desire Paths” for accessible background.
“Pave the cowpaths”: HTML5 design principle formalised by the W3C HTML Working Group: W3C HTML Design Principles: Pave the Cowpaths. The principle: when a practice is already widespread, consider adopting it rather than forbidding it or inventing something new.
Twitter hashtags, Instagram pivot, Slack origin: These business examples are sourced in full in the companion article: Your Business Is Full of Desire Paths. Are You Building Around Them?
Shadow IT in UK businesses: Enterprise shadow tool usage is well-documented. See: Gartner: Shadow IT Survey (2023): approximately 41% of employees use technology that IT hasn’t approved. UK-specific figures from KPMG UK Tech Insights 2024.