One editorial calendar. Four channels writing themselves.
ITTWeb publishes to a blog, a newsletter, a personal LinkedIn profile and a company LinkedIn page. Four audiences, four voices, one marketer. We built the pipeline that turns a row in their editorial calendar into drafts on all four, ready to review.
- Sector
- Web services and digital marketing
- Built as
- CTO, Prisma Group
- Shipped
- June 2026 · with a written platform guide
- Channels
- Blog · newsletter · two LinkedIn voices
- Their site
- ittweb.net
The same idea, rewritten four times, by one person.
An editorial plan existed and was good. The trouble was everything downstream of it. One topic had to become a blog article, a newsletter, a personal LinkedIn post in one voice and a company LinkedIn post in another — four different lengths, four different registers, four different places to publish it.
Done properly that is most of a day per topic. Done in the time actually available, three of the four get written badly and the fourth does not get written at all.
Four voices, not one voice pasted four times.
The obvious build is one writer that shortens its output for each channel. It produces four texts that are recognisably the same text, which is exactly what an audience following two of the channels will notice.
So each channel got its own writer, with its own instructions, all reading from one shared brand knowledge base. The facts are common; the voice is not. The company LinkedIn page and the personal profile in particular are written by different agents, because a person and a company should not sound alike.
We also decided nothing publishes itself. Blog posts arrive in WordPress as drafts. The LinkedIn workflows were delivered switched off, with the account identifiers left as placeholders that have to be filled in deliberately. A content system that can post without a human reading first is a liability, not a feature.
A chat box in front, the work behind.
The marketer types "generate week 1". Thirty to sixty seconds later the whole campaign comes back as cards to review.
- 01
It reads the plan
It takes the week straight out of their editorial calendar in Google Sheets, hands each topic to the right writer, and brings everything back as one set of drafts. The calendar stays exactly where the team already keeps it.
- 02
Four writers
WordPress gets HTML plus SEO metadata plus an image prompt. The newsletter gets a body, a target segment and an image prompt. The two LinkedIn writers get text and hashtags in their own register. All four read the same brand knowledge base.
- 03
Images in parallel
Each card generates its own image while the others do the same, about seventeen seconds and seven cents apiece. Serially this would be the slowest part of the run; in parallel it disappears into the wait.
- 04
Review, edit, approve
Every card can be edited in place before it goes anywhere. Approve is a button per card, not a button for the batch, so a good week with one bad post is still a good week.
- 05
Publishing, per destination
Each channel has its own publish route. WordPress creates a draft. If a destination is not configured the route says so plainly instead of failing silently — a half-configured system should be loud.
Platform guide, June 2026.
The plan did not change. Everything after the plan did.