The object campaigns are most often confused with: what a Routine is, its four states, its triggers and approval policy, and how to decide which one you actually need.
Jump to module
A Routine is a saved, repeatable arrangement: this work, on this trigger, with this approval policy, indefinitely.
Three weekly posts. A performance report every Monday. A reply pass over yesterday's comments. None of those have a subject and none of them have an ending — they are things you want to keep being true.
Routines are organised by what they accomplish:
| Category | Produces |
|---|---|
| Social posts | Scheduled content |
| SEO articles | Long-form for your site |
| Engagement replies | Responses to conversations |
| Repurposing | New pieces from existing ones |
| Performance reporting | Analytics summaries |
| Audience research | Findings about who you reach |
You will also see them merchandised by goal — build in public, stay consistent, grow engagement, repurpose content, launch something, thought leadership, analyse performance. That is a browsing axis, not a second taxonomy: one routine template can serve several goals while producing exactly one kind of thing. "Weekly social posts" is how you stay consistent and how you build in public, but it only ever produces posts.
| State | Meaning |
|---|---|
| Draft | Saved, never armed. Triggers nothing, consumes nothing. |
| Active | Armed. Triggers fire and runs dispatch. |
| Paused | Definition intact, triggers suppressed. Reversible. |
| Archived | Retired. Hidden from the list, history preserved. |
Pause is the reversible stop, and it is what the system itself uses when something goes wrong — a disconnected source or a plan limit pauses a routine rather than deleting it.
There is no destructive delete. The audit trail behind a routine — its runs, artifacts, approvals and published posts — outlives the routine itself. Archiving is terminal, and an archived routine is restored by cloning rather than by re-arming a definition whose sources may have gone stale.
If that sounds familiar, it should: campaigns follow the same rule for the same reason (Module 11).
A routine's approval policy can be inherit, review every artifact, or auto.
The rule that governs it is the same one governing agents: a routine may only ever tighten oversight, never loosen it. Setting a routine to auto is a request, honoured only if your workspace and agent autonomy already permit it. It is not a way around workspace policy.
This is the sharpest difference from a campaign. A routine can be set to publish without review, if your workspace allows it. A campaign cannot, ever.
A routine's runs report status — running, awaiting approval, partial, scheduled, completed, failed, cancelled — and none of those are stored. They are computed from the underlying records every time they are read, exactly as campaign results are. Adding a customer-facing object never adds a second version of the truth about what happened.
Ask three questions. Any single "no" points at a routine.
1. Can you name what the content is about? A subject in one phrase — "the new pricing", "spring intake", "the conference". If the honest answer is "our business", that is a routine.
2. Can you name when it ends? If it should still be running in six months, it is a routine. Campaigns are one to four weeks by construction.
3. Do you want a verdict? A campaign exists to answer "did this angle work?" If you are not going to read the result and decide something, the finite window is buying you nothing.
| You want | Use |
|---|---|
| Three posts a week, forever | Routine |
| Two weeks of content about the launch | Campaign |
| A weekly performance summary | Routine |
| A promotion that ends on the 30th | Campaign |
| Replies to every comment | Routine (or an Engager agent) |
| A push to test whether teaching content lands | Campaign |
| An article a week for SEO | Routine |
| Four weeks of thought leadership before a funding announcement | Campaign |
Using a campaign as a schedule. You end up recreating one every month, never letting any of them finish cleanly, and getting no readable result from any of them. The four-week ceiling exists to make this uncomfortable on purpose.
Using a routine for a launch. The content goes out, but nothing gathers it into a result. There is no window, no angle applied consistently, and nothing at the end that says whether it worked. You get posts; you do not get an answer.
The strongest setup runs both, with the routine as the floor and campaigns as the peaks.
The floor. A routine producing your baseline — enough to keep you present in the feed on the weeks when nothing special is happening. Modest volume. Consistency is the whole point.
The peaks. Campaigns layered on top when there is something to say. A launch. A promotion with a real deadline. A four-week push on a single angle you want to test.
Two practical notes when both are running:
Watch the combined review load. A routine and a campaign both put drafts in the same Approve queue. Three routine posts plus a five-piece campaign week is eight decisions, not five. If the queue becomes unmanageable, cut the campaign mix before you cut the routine — the routine is your floor.
Do not double-book the same subject. If your routine is already posting about the launch because you told the agent to, and a campaign is also producing launch content, you get two uncoordinated versions of the same message. Let the campaign own the subject for its window, and keep the routine on general presence.