Drafting and repurposing
Handled by the AI Employee, from your source material and your voice examples, on the rhythm you set.
How to Run a Content Calendar With an AI Employee
How-to — — by Mahmoud Zalt
Build a content calendar around format return instead of post count, and hand the drafting and scheduling to an AI Employee you can actually review.
Why most content calendars fail in week six
Because they are built as a schedule of slots rather than a plan of formats. A grid that says Monday, Wednesday, Friday tells you when something must exist but nothing about what it should be, so the slot gets filled with whatever is fastest to produce on the day. Six weeks in, the calendar is technically full and doing nothing.
The second failure is that the plan assumes a version of you who has time. Every calendar looks achievable when you build it on a quiet Sunday. It meets a launch, a sick week or a big customer and quietly dies, because the constraint was never planning, it was production capacity on your worst week.
Both problems have the same fix, and it is not more discipline. Choose formats by what they actually return, then remove production from your critical path so a bad week does not empty the queue. That is the whole method, and the rest of this is how to run it.
Pick formats by return, not by effort
Start from the data rather than from habit. In one published analysis of monthly volume and engagement rate by format, the ranking of what brands publish most is close to the inverse of what performs best. On LinkedIn, native documents lead engagement at 6.90 percent and go out about twice a month, while images at 5.00 percent go out around seven times.
| Platform | Format | Posts per month | Engagement rate |
|---|---|---|---|
| Native documents | 2 | 6.90% | |
| Multi-image | 1 | 6.80% | |
| Video | 4 | 5.90% | |
| Images | 7 | 5.00% | |
| Carousels | 5 | 0.52% | |
| Reels | 10 | 0.50% | |
| Images | 8 | 0.35% | |
| Reels | 7 | 0.20% | |
| Albums | 6 | 0.17% | |
| Links | 10 | 0.05% |
The pattern repeats on every platform in that data. Instagram carousels lead at 0.52 percent while being posted about five times a month against ten Reels. On Facebook, links are the most published format and the weakest by a wide margin at 0.05 percent, well behind Reels at 0.20 percent.
There is a simple reason for the inversion. The formats that perform take longer to make, so they are the first thing dropped when the week gets tight. Your calendar is not really a record of what works, it is a record of what you had the energy for. Fix the energy constraint and the mix changes on its own.
Build the calendar in four passes
Work from source material outward rather than from dates inward. One substantial thing you know well can populate a month if you plan the reshaping deliberately instead of hoping you will get around to it.
The four passes
- Pass one: pick two formats and one channel — Choose the two highest-return formats on the single channel that matters most to you. Two is enough. A third arrives once the first two run for a month unattended.
- Pass two: name four source assets for the month — A customer problem you solved, a decision you made and why, something you changed your mind about, and one piece of real data from your own work.
- Pass three: map each source to formats — One source becomes the heavy format and two or three lighter follow-ups. Repurposing is planned here, not improvised later.
- Pass four: set the rhythm and the gates — Decide the publishing days, then decide which formats publish on release and which always wait for you. Write that down before anything is connected.
Four sources and two formats is a deliberately small month. It will feel thin next to advice telling you to post daily. It is also a month you will actually finish, and finishing is the variable that separates a channel that grows from one that restarts every quarter.
Once the shape exists, the calendar becomes a briefing rather than a to-do list. You hand an AI Employee the four sources, the two formats, the voice examples and the rules, and it drafts on the rhythm you set. That is the shape we build around at Sistava. Employees keep memory across runs, so the corrections you make in the first fortnight keep shaping drafts months later, and the calendar gets more accurate to your voice rather than drifting away from it.
Who does what once it is running
The division is fixed and worth stating plainly, because getting it wrong is how people end up in trouble with a platform. The AI Employee owns production. You own decisions and every human interaction. Publishing runs through official platform APIs, never through anything that imitates you clicking.
Benefits
Scheduling and publishing
Through each platform's official API, with the formats you flagged held for your release.
Engagement
Yours, always. LinkedIn prohibits bots that comment, like, share or message, and accounts using them risk restriction.
Claims and sensitive topics
Behind a permanent approval gate. Anything touching pricing, customers or legal waits for a person, every time.
Reporting
Assembled for you into one readable summary so the monthly review takes minutes rather than an afternoon.
The mix decision
Yours. The employee shows what performed, you decide whether it means anything and what to do about it.
Two platform details are worth building into the plan rather than discovering later. Instagram limits accounts to 100 API-published posts in a rolling 24 hours, where a carousel counts as one, and needs a professional account connected to a Facebook Page. TikTok requires an app audit, and until it passes, all content posted by unaudited clients is restricted to private viewing mode.
Neither of those will trouble a sane calendar. They matter because they tell you what kind of tool you should be using at all: one that publishes on the official path and expects a human to hold the permissions, rather than one promising to act as you invisibly.
The weekly review is the one habit you cannot delegate, and it takes about fifteen minutes. Open the activity feed, which records every action with a screenshot, skim what was drafted and why, correct anything that reads slightly off, and release what is waiting. On Sistava a gated post waits as a request instead of going out and being walked back afterwards, so the same fifteen minutes covers both the review and the release. Do that and the corrections compound instead of repeating.
Frequently asked questions
FAQ
How do I make a content calendar with AI?
Choose one channel and two high-return formats, name four source assets for the month, map each source to a heavy format plus a few lighter follow-ups, then set the publishing rhythm and decide which formats need approval. Hand that brief plus your voice examples to an AI Employee to draft against, and review the output weekly rather than post by post.
How often should I post on social media?
Less often than most advice suggests, in better formats. In one published analysis, LinkedIn native documents led engagement at 6.90 percent while going out roughly twice a month, and Instagram carousels led at 0.52 percent on about five posts a month. Volume and return are not correlated in that data, so pick the format first and let the cadence follow what you can sustain.
Can AI schedule and publish posts automatically?
Yes, through each platform's official API with your authorisation. Instagram allows up to 100 API-published posts in a rolling 24 hours and requires a professional account connected to a Facebook Page. TikTok requires an app audit, and unaudited apps have their posted content restricted to private viewing. Automating engagement is a separate thing and is prohibited on LinkedIn.
How far ahead should a content calendar be planned?
A month is the sweet spot for most small teams. Further ahead and the sources go stale or the plan stops matching what is happening in the business. Closer in and you are improvising. Plan the month in one sitting, let drafting run on a weekly rhythm, and keep one flexible slot for something timely.
How do I repurpose one asset across formats?
Decide the reshaping in the planning pass, not afterwards. One substantial source becomes the heavy format on your main channel, then two or three lighter derivatives that each stand alone rather than teasing the original. Handing all of those to one AI Employee in a single brief keeps the voice consistent across them, which manual repurposing rarely achieves.
What should never be published without my approval?
Anything touching pricing, named customers, legal or regulatory claims, competitor comparisons, or a live controversy. Attach those as plain-English rules to the publishing tool so they bind on every run rather than living in a prompt, and keep the approval gate on them permanently, even after you relax the gate on routine formats.
A content calendar is not a discipline problem. It is a capacity problem wearing a planning costume, which is why buying a better template never fixes it. Take production off your critical path and the calendar stops depending on how your month went.
If you want to feel the drafting quality before you hire anyone, the free tools on Sistava will write a single social caption or a set of blog post ideas with no account, which is a cheap way to see whether the voice is close enough to be worth correcting.
Pick the formats that actually return something, plan a small honest month, let an AI Employee do the drafting and the scheduling on the official path, and keep the decisions and the conversations for yourself. That is a calendar that is still running in six months, which is the only measure of one that ever mattered.