The company page posts, and then a quiet hour follows. Your teammates meant to react; they were in meetings. The usual fix is a pinned Slack message that says "can everyone like this?", which works once and then becomes noise. A better fix is to plan the first hour before the post goes out, with the people who agreed to take part, and to let an assistant do the tedious scheduling while a person keeps the final say.

Plan it as a sequence, not a pile-on

Hypothesis: a burst of identical reactions in the same minute looks like what it is, while a spread of real reactions from real colleagues looks like interest. So plan for a stagger: a few reactions in the first minutes, a comment or two a little later, a repost once the conversation exists.

Build the plan around who will genuinely do each thing. A teammate who knows the feature can write a real comment. Someone who talks to customers can repost with a line about them. Nobody should post something they would not say anyway. The platforms prohibit coordinated inauthentic engagement, and the Field Manual quotes each rule. Your plan is for people who already want to help, at a pace a person would keep.

What contributed profiles are

Teammates and supporters choose to contribute their own profile to your account. It stays theirs. An action runs as them only because they set it up. That consent is the whole design, so ask first, show them what you plan, and let anyone opt out.

Let the agent queue, and a person approve

Turn approval mode on. With it on, whatever an agent queues waits for a person in the app. An agent can never approve its own content (how the MCP server keeps you in control explains the rule), and that is the reason to trust the setup: the thing that wrote the work is never the thing that signs it off.

Start with a read-only check so you know who is available:

output 
Read only: list the profiles contributed to my account and what each is set up
for. Do not queue anything.

Then the plan, as a rehearsal:

output 
Our announcement goes out from the company profile on Thursday at 10:00. Plan
first-hour engagement from the contributed profiles: stagger reactions across
the first 45 minutes, and queue a repost with a short line of my wording from
two teammates after 10:30. Do a dry run, show me each action and its time, and
wait for me to say "go". Do not confirm yet.

Check the dry run before you say go. Is any profile doing more than a person would? Is each action under the right person's profile? Are the times spread, not clustered? Then:

output 
Go. Confirm the plan, then list everything now waiting for approval.

Confirming does not send anything. It places the work in the queue, where it waits for a person to approve it.

Why self-approval being impossible matters

Imagine the alternative: an assistant that drafts, queues and approves in one breath. A misread instruction, a wrong profile or an over-eager retry becomes a live action before anyone sees it. Splitting the roles puts a human check exactly where the risk is. Per-profile risk limits also apply to what agents do, and pause stops everything if the announcement is pulled.

For the approver, a good habit is a two-minute review before the hour: open the waiting items, read the wording, confirm the times, release them.

After the hour

Look at what happened. Reading your results covers comparing the first hour with the full run. And be in the replies yourself; the stagger helps, and a real answer from you helps more.