Marketing
Picture Muster GmbH, a 30-person German SME with one person who owns marketing alongside three other jobs. Business-Gym gives that person a week’s worth of drafted, on-brand content and a short list of decisions — it does not give them an autopilot.
Every post that leaves the company is approved by a human, one post at a time. Publishing is real — LinkedIn through an official API, Facebook and Instagram through the Meta Graph API — but the publish step only runs after someone says yes. This page walks through a realistic week and marks the parts that are staging only.
What does a marketing week actually look like?
Section titled “What does a marketing week actually look like?”The rhythm is built around one decision on Monday and short approval passes after that, which is roughly the amount of marketing attention a 30-person company can honestly spare.
| When | What happens | Who acts |
|---|---|---|
| Monday 09:00 (Europe/Berlin) | The weekly content sprint runs: it reads your offerings and campaigns, then proposes candidate topics | Nobody yet — it stops and waits |
| Monday morning | You pick the topics and add your angle, an anecdote, and your no-gos | You, at the gate |
| Monday, after your pick | Drafts are written in your voice, one per picked topic, and staged as content plus a pending approval each | Nobody — staging only |
| Tue–Fri | You work the approval queue; each approved post publishes to its channel | You, per post |
| Every night | Ad spend is pulled in, and the Meta budget guard proposes pauses for campaigns that burned money with zero clicks | Nobody — proposals only |
| Any day | You check the outbound ledger for what actually went out and what failed | You, reading |
The sprint stops twice on purpose. It will not draft until you have picked topics, and it will not publish until you have approved a post. Two gates, both human.
Can it post to LinkedIn or Meta on its own?
Section titled “Can it post to LinkedIn or Meta on its own?”No — and the reason is structural rather than a setting you could flip by accident. The drafting step creates a staged send and a pending approval, and it has no path to the publish transport at all.
The publish code runs in exactly one place: after a person decides an approval. That is the same hard floor that governs every external send in the platform, and it does not move as autonomy rises.
Approving the same item twice does nothing. Only a staged send can transition, so a double-click or a second approver cannot double-post.
Which channels really publish today?
Section titled “Which channels really publish today?”Three destinations publish for real once you approve: LinkedIn (through an official LinkedIn Marketing API partner, posting via LinkedIn’s own API — not browser automation), Facebook and Instagram (both through the Meta Graph API).
Google Ads is read-only by construction. Business-Gym reads campaign spend, clicks and conversions and has no code path that could change a campaign, a budget or a bid. That is a design decision, not a gap waiting to be filled.
The full read/write/propose matrix for every connector is on integrations.
How does it stop flooding your feed?
Section titled “How does it stop flooding your feed?”The sprint plans the week before it drafts anything, using a fixed cadence rather than “however much the model felt like writing”. It targets two posts per offering per week, rotated fairly across your offerings, with a hard cap of ten posts per week in total.
Within that plan, topics rotate across your content pillars and across formats — story, how-to, proof, opinion, announcement — choosing whichever format that pillar has used least recently. The variety comes from a deterministic rotation, not from turning up the model’s randomness, so the same inputs give you the same plan.
Per-page voice, pillars and brand kit are configured under Marketing → Channels → Settings. Campaigns live at Marketing, audience segments at Marketing → Audiences, and the full content list at Marketing → Content — where scheduling stages an approval and never publishes.
What happens to your ad budget overnight?
Section titled “What happens to your ad budget overnight?”A nightly guard reads the last seven days of Meta ad spend and proposes pausing any campaign that spent more than the threshold (€50 by default) while getting zero clicks. It stages one pending approval per campaign and stops there.
It never executes. The underlying write actions are deliberately narrow: the only verbs are pause, resume and set daily budget, there is no create or delete path, and a daily-budget change beyond ±50% of the current value is refused with a stated reason rather than quietly clamped.
For Muster GmbH this is the difference between waking up to a €400 hole and waking up to one line in the approval queue that says a campaign is burning money. You still decide.
Where do you check what actually went out?
Section titled “Where do you check what actually went out?”The Outbound surface is the proactive mirror of your inbound queue: an in-flight queue plus a sent ledger you can filter by channel. See the outbound ledger for how to read it.
Send status follows the real publish result. If the platform rejected the post, a token was missing, or the target page is outside your configured estate, the send is marked failed with the specific reason attached — it is never reported as sent. Marketing performance reporting sits under Marketing → Analytics.
Who in a small team can do what?
Section titled “Who in a small team can do what?”Editing channels, dispatching sends and scheduling content are owner or admin only, which matters when marketing is one person’s side job and three colleagues have logins.
An approver or read-only member sees the same content list, the same filters and the same previews, but none of the edit or lifecycle controls. Which roles can approve, and on which surfaces, is covered in who can approve, and where — approvals are worked in the web app.
What is not built yet?
Section titled “What is not built yet?”The marketing pipeline does not learn from your edits. When you rewrite a draft or reject one, that decision is recorded but it does not change what next Monday’s drafts look like. This is the same gap described in the inbound → approve → execute loop.
There is no campaign autopilot. Each post is staged and approved on its own, so a five-post campaign is five approvals rather than one decision that unlocks the rest.
There is no native mobile app and no push notifications. Nothing taps you on the shoulder when an approval lands — you go and look, on the web.
For the mechanics behind all of this — the send gates, the dark-by-design list, and what the approval carries — read marketing in the commercial section. To see the same propose-then-approve pattern applied elsewhere, look at sales, or set the targets these posts are meant to serve in goals and OKRs.