Every newsroom of any size has a style guide, and every one of them is boring on purpose. It exists because the alternative is fourteen writers each independently deciding how to spell email, whether the Oxford comma lives or dies, and how many exclamation points a sentence can carry before it stops being journalism.
The guide isn’t there to make anybody a better writer. It’s there so the paper sounds like one paper.
Most people using AI for social content have fourteen writers and no guide. They just haven’t noticed yet, because the fourteen writers all look like the same chat window.
I sat down with Loren Bartley of Impactiv8 to walk through how she actually produces her social content, and the thing worth stealing isn’t a prompt. It’s the shape.
The problem isn’t the output, it’s that nothing holds the standard
Loren’s diagnosis is sharper than most, and it comes in two parts.
The first is that a fresh chat has no memory of what came before, so you get repetition, or you get contradiction, and contradiction is the expensive one. You argue a position in March and the opposite in April, and because you didn’t write either post, you never catch it. Her framing on why that matters: authority isn’t one clever post, it’s a body of work that’s clearly going somewhere.
The second part is about the prompt itself. You type a good one, it works, and then it’s gone. Next time you’re reconstructing it from memory or digging through a prompt library you can’t face opening. The inputs drift, so the outputs drift.
Her distinction is the load-bearing one for everything below. A prompt carries the instruction. A skill carries the standard, and it runs the same way every time. It’s the difference between telling a new hire what to do this morning and writing them a job description.
Which means the improvement loop changes shape. When Loren notices a certain kind of hook landing, she doesn’t file that away to remember next week. She opens Hugo, the skill that writes her hooks, and updates him. The learning sticks to the system instead of to her memory.
The chain, in order
Here’s what she’s built, and the reason each link exists.
Constance runs about once a quarter. She interviews Loren at length and produces a reference document covering cadence, operating rhythm, content mix and ratios, platform standards, calls to action, offer mapping, repurposing rules, guardrails and language conventions. That document is the standard everything downstream answers to.
Edison takes that standard and produces a cycle brief, usually weekly. What period are we in, what are we building toward, who’s it for, what has to shift in someone’s mind, what the calls to action are. Loren can push her out to eight weeks and deliberately doesn’t, because she says she isn’t organized enough to plan that far, which is a more honest reason than most planning advice gives you. The brief takes about eight minutes of her attention. She said she cleaned her office while it ran.
Connie is the expediter. She reads Edison’s brief and assigns the work, Caprice on captions, Vivian on long-form video scripts, Hugo on hooks, without Loren deciding who does what. That’s the step that killed the old routine of calling each skill by hand and building posts one at a time.
The hub is where it lands, in her case Notion, connected directly so drafts arrive as editable rows rather than text trapped in a chat. This solves the failure she describes as the embarrassing one: content that got made and then never got published, because it was stuck somewhere she couldn’t find and couldn’t edit without retyping. Each row carries the platform variants, alternates she can promote if she prefers them, the image brief, the status and the scheduled date.
The gate is Loren. She edits, attaches assets, and marks rows approved. Nothing moves without that.
Agatha takes what’s approved and stages it into the scheduler. And she’s built to argue.
Build the argument in on purpose
This is the part I’d copy first, before any of the rest of it.
Agatha doesn’t simply do as she’s told. She checks the approved rows against their own briefs and reports what doesn’t line up. On camera she caught that a post briefed as video had a still image attached, which was true, and Loren said she loved being caught. Another time Agatha flagged an image as wrong for the post and Loren overruled her, because that time it was fine.
A check that can be overruled is still a check. A workflow with no check is a conveyor belt, and the thing about a conveyor belt is it moves your mistakes at exactly the same speed as your good work.
One caveat worth naming
Loren works in Claude, and on camera I made the point that projects and skills are different things doing different jobs. Projects are the binder that holds your files and context; skills are the reusable method. As of that recording, Claude was the only one of the major tools letting you run a custom assistant inside a project, where a custom GPT, a Gem or a Copilot agent sits outside the notebook.
Check that before you rearrange anything around it. That’s the sort of gap that closes without an announcement, and the shape she’s built matters more than the platform underneath it. Stations, a standard, an expediter, a place where work becomes editable, and a human who can still say no.
Watch the full walkthrough: Loren’s workspace on YouTube.
Chapters
- 1:43 – content that got made and never shipped
- 2:53 – why a fresh chat breaks at volume
- 4:19 – a skill holds your standards
- 10:35 – projects versus skills
- 11:24 – Edison and the weekly brief
- 16:52 – Connie as the conductor
- 17:57 – the content hub
- 24:37 – human in the loop
- 26:17 – caught making a mistake
- 29:50 – time back versus energy back
Find Loren: Impactiv8. Her editorial skill, Edison, is available at aimegasuccess.com/edison.
11 Techniques to Repurpose Long-Form Blog Content into a Month of Social Posts