Automated Content & Blog Post Copy Writer

AI Slop vs Useful Articles

Everybody knows you can use an LLM to generate slop in the form of blog posts and other text that nobody wants to read. The question is, can you use LLMs to help generate compelling content written by humans using their own words instead of machine extruded filler?

Content marketing can take many hours of human effort to come up with new stories, research them, and write them up. What if you could use existing content already produced by the people in an organisation and create fresh new blog articles grounded in real human discussions, solutions, and original words?

For this project we took an existing content source and used LLMs to repurpose those human-authored words into articles that flowed naturally in the new format. The source here was informational videos but they could be any type of internal company material such as wiki entries, work logs, research papers, product manuals, or even existing articles that need curating for the less formal public blog format.

The Content Pipeline

The goal was regular blog posts on a weekly schedule. We set up an automated pipeline that ran a weekly job mining existing content for juicy concise blog posts and then turning those into compelling, well-formatted articles that are posted to the website on a regular basis.

• Phase 1: The agent engages in a discovery process looking at newly published internal video transcripts, chat comments, wiki entries, documents, etc. It puts together a candidate list based on these and ranks them.
• Phase 2: It then collates all of the material to do with a particular subject from new and historical sources across the full data set. It compiles these into a structured source document.
• Phase 3: The agent creates an editorial draft using a professional tone without superlatives or clickbait, prohibiting fake first-person claims, and using the original human text wherever possible with only light editing for form and style to tie everything together.
• Phase 4: The posts are scheduled for release on a regular cadence (e.g. weekly) and placed in a review queue for a human to read and provide any feedback before they go out. This final human step is critical for making sure the posts accurately represent the external face of the organisation.

The result is a company blog that runs itself. Each article is compiled, researched, drafted and placed in a human review queue. Once a human has read and approved it, the post is then automatically published on the website when the publishing date rolls around.

Results

In the end this pipeline saves hours of human work trawling through internal resources and repurposing it in the form of blog posts. The content creation overhead is reduced to a brief weekly review, and the posts provide a great way to increase SEO keyword surface. The organisation now has a vibrant blog with compelling, up to date content published on a regular cadence and written in the words of the organisation themselves.

If you're interested in similar content automation workflows, please feel free to get in touch.