What does the social manager produce?
A planned week of posts, reels and carousels, each built from the catalogue: the product, its real attributes, its price and its availability, with a caption in the brand's voice and generated imagery or video where the brand wants it. The plan shows what goes out on which day and to which platform.
Because the content is generated from catalogue data rather than from a prompt about the brand, a caption that names a material or a price is naming the one in the product record, and an out-of-stock line does not get promoted.
- Posts, reels and carousels built from real catalogue data rather than from a prompt.
- Captions in the brand voice, from the Brand Brain's samples and word rules.
- Generated images, carousels and video where the brand wants them, with the same approval step.
- A planned week showing what goes out on which day to which platform.
- Per-item insights after publication, joined to the product each item featured.
Where does the brand voice come from?
From the Brand Brain: the voice samples the brand approved and the word rules it set, including the words it never uses. The result reads like the brand rather than like a model's idea of a fashion caption, and a reviewer editing a caption is adjusting rather than rewriting.
Where a caption would need a claim the Brand Brain does not hold, the draft flags it instead of inventing one. A social caption is still a public claim about a product.
Why plan and approve a week at a time?
A week is the unit a brand actually thinks in: a launch on Tuesday, a restock on Thursday, a quiet Sunday. Planning a week lets the engine balance the mix and the products, and approving a week takes one sitting rather than seven interruptions.
Each item is still individually editable and individually approvable. An owner can approve five, edit one and reject one, and only the approved items are scheduled.
How are posts published and measured?
Approved items publish to the connected social platforms on their scheduled times. Insights come back from those platforms and are joined to the item, so each post's reach and engagement sit next to the product it featured, and the week's output is measured rather than assumed.
Where a post can be tied to sales through the store's own data, that is shown as what it is: a correlation in the brand's own numbers over the window, with the assumption stated, never a claim that the post caused the sales.
What it will not do
It will not publish without approval unless the owner has explicitly turned on unattended sending for that class, and even then every item is logged and later scored. It will not post a product that is out of stock or a price that has changed since the draft. And it will not invent a claim the Brand Brain does not hold.
It also does not buy reach. Paid media is the performance advertising control room, which is a separate surface with its own money-safety rules.
Which role owns it on ftes.ai?
The Social Officer FTE owns the calendar and is judged on earned reach. It proposes the week, holds the approvals, publishes what was approved, and reports what each item earned. The Content Officer FTE owns the longer-form work the posts often point to.
A brand connecting a social platform can have its first planned week ready the same day, and the honest way to start is with one week, approved item by item, so the voice and the mix can be corrected before the second.
How does a brand keep the voice honest over time?
By correcting the Brand Brain rather than the captions. A caption edited every week is a symptom; the cause is a voice sample or a word rule that does not reflect how the brand now sounds. Updating the source fixes the next fifty captions instead of this one.
The word rules matter most. A short list of words the brand never uses removes the tells that make generated copy recognisable, and it is the single cheapest improvement to the quality of everything the engine drafts.