How-To Guide
How do you run a social media approval workflow for agency clients?
Run agency approvals as five named gates with one reviewer each, an expiry clock, and a rule that any edit resets the yes. Includes a review-window template.
Run every post through the same five gates: an agreed brief, an internal review, a client sign-off on the exact version that will publish, a dated slot, and a record you can retrieve. Apply the process to each client; what changes is who holds each gate, not the sequence. Two rules keep the workflow honest. The client approves the version that publishes, so every network's version is visible at the yes. Any edit to words, media or timing after approval voids the yes until someone re-approves the new version. Tools can enforce parts of this chain. The habit is yours.
- Five gates, one order: brief, internal review, client sign-off, schedule, record.
- The client approves the posting version for each network, not a master caption the scheduler will rewrite.
- Write the review window down before the month starts: one working day for routine posts, two for high-risk, and drafts landing at least three days before publish.
The five gates a client approval workflow needs
An approval workflow is the set of decisions a post passes between being written and going live. Agencies that run it well use the same five gates for every client, because a gate that moves or disappears is a gate nobody trusts.
- Brief. The client has already agreed what the month is for, so a reviewer argues about execution, not direction.
- Internal review. Someone at the agency checks facts, names, links, crops and voice before the client spends attention on the post.
- Client sign-off. The client sees each network's version as it will publish and gives one decision: approve, or request a change with a note.
- Schedule. The approved version moves to a dated slot. Approval and scheduling stay two recorded steps.
- Record. The version, the approver, the time and any note are stored where you can find them when the post is questioned.
| Gate | Question it answers | Evidence it leaves |
|---|---|---|
| Brief | What did we agree to make? | An approved brief or content plan |
| Internal review | Is the work correct and on-brand? | A reviewer note or a cleared status |
| Client sign-off | Does the client approve this version? | Approver name, time and version |
| Schedule | When and where does it publish? | A dated slot on the approved version |
| Record | Can we prove what was approved? | Version history and an activity log |
The sequence is the same for an agency with several clients and for a single brand, and that is the point. What changes per client is who holds each gate and how long each one takes. An agency that lets a client skip straight to the calendar, or lets an account manager approve on the client's behalf without saying so, loses the only thing the workflow produces: a yes that means something.
Planable splits approval into four types: none, optional, required and multi-level. Those map onto gate strength. Optional records an opinion. Required blocks publishing until at least one approver signs off. Multi-level chains stages where each one gates the next.
Original diagram of the five-gate approval sequence. Gate names are editorial; the internal-then-client order follows Kontentino's 2026 roundup and the status-plus-log record shape follows ZoomSphere's documented approval flow, both read 15 November 2026.
Make it work for your brand
Find it. Remake it. Post it. Superdirector turns the viral formats in your niche into your brand's carousels, X posts and video scripts, and keeps the calendar full.
Set the internal gate before the client sees anything
The internal gate exists so the client reviews a post, not a draft of a post. Copy errors, a wrong handle, a crop that cuts the product, or a claim legal will reject are the agency's job to catch. Each error the client catches costs credibility and a review round.
Kontentino's roundup describes the standard chain as internal review first and client sign-off second. ZoomSphere's per-client flow keeps a client discussion and a separate internal thread the client never sees, which is a practical way to run the internal gate inside the same tool the client uses.
- Facts and claims. Names, prices, dates, links and disclosures.
- Format. Each network's crop, caption length and the version the scheduler will send.
- Voice. Does this sound like the client, or like a template?
Staff the gate with one or two named reviewers. Postprism's guidance is that two approvers per workspace is usually enough, and a longer list means nobody owns the decision. Do the same on the client side: name the person whose yes counts, and name a backup for holidays.
What the client signs off, version by version
A client approval is a decision about a version, not a mood about an idea. If the approver sees one master caption and the scheduler publishes a different version on another network, the client approved something that never went out. Show every network's version at the point of approval, including the image crop and the opening lines of the caption.
Planable's own FAQ is precise about the click path: a public share link lets anyone view and comment without an account, but approving requires an Approver account, which the vendor makes free and which does not count toward paid seats. That model works if the client accepts an account and fails if they will not. Cadenus makes the same point from the buyer's side: the first thing a sign-off needs is a frictionless yes, and every login you add slows it down.
Batch the review. ZoomSphere sends a week or a month of posts in one email with real platform previews, and the client approves or comments post by post. A monthly batch respects a busy marketing director's calendar better than a stream of separate notifications, and it keeps the decision in one sitting.
Check the bill at the same time you check the click. Kontentino's rule is that a tool forcing every client contact into a full paid seat just to approve has turned pricing into a security feature. Cadenus found that Loomly counts an approving client as a user, so the fourth participant pushes the agency to a higher plan, and that Later's account-free link allows comments but not approval. Ask each vendor for the exact path before promising a client a no-login yes.
What happens when a client edits an approved post
The failure mode is quiet. A post is approved, someone fixes a typo or swaps an image later, and the version that publishes is not the version the client saw. Postprism names it directly: an approved post can quietly change before it goes out unless the tool resets approval on an edit. Sequence's test is the one to run in a trial. Change a caption after approval and inspect what the software does.
Planable's answer is a lock. Its Lock after Approval setting makes an approved post read-only, covering text, media, date and order, and anyone who needs a change must first unapprove the post. Copy that rule even on a spreadsheet: the yes belongs to a version, and editing the version revokes it.
Approval is not scheduling. Sequence documents final approval as a move to Ready to post, with a person still confirming the destination and time, and Planable makes auto-scheduling a separate setting you can leave off. Treat the approved timestamp as the boundary of the review work, and the slot as a second decision with its own record.
- Freeze the approved version and mark it read-only where the tool allows it.
- Log any change and name who made it. If the change touches a claim, a price, a date, a person or the media, void the approval.
- Send the new version back through the same gate, with the previous decision visible, and get a fresh yes before the slot.
- If the slot is too close for another round, move the slot. Publishing an unapproved version is the failure the workflow exists to prevent.
Original diagram of what an edit does to an approval. The read-only lock and the unapprove-first step follow Planable's approval FAQ; the quietly-changed warning follows Postprism's 2026 workflow guide, both read 15 November 2026.
Write the review window before the month starts
Approval stalls on silence more often than disagreement, and a deadline you never wrote down is the easiest thing for a client to miss. Put the window in the monthly plan: when drafts land, when the client reviews, when the slot is.
Postprism's working standard is one working day for routine posts and two for high-risk ones, and its rule of thumb is that if approvals can take two days, drafts must land at least three days before publish time. Use that as a starting point and adjust it from evidence in your own queue.
| Risk tier | What goes in it | Who approves | Target turnaround |
|---|---|---|---|
| Routine | Evergreen tips and reposts of the client's own content | One team lead | Same or next working day |
| Medium | Offers, customer mentions, event promotion, product features | A team lead or the client | One working day |
| High risk | Pricing, launches, legal or health claims, partnerships, sensitive topics | Two reviewers in order, a lead then the client or legal | Two working days |
Keep the window in one place the client can see. Planable sends approval requests by email, Slack, mobile push or an in-app inbox, and automatic reminders chase approvers who are late; most tools have some version of that. Use batch review for the month, name a backup approver, and look once a month at how many posts came back and why.
Original diagram of a written review window. Tier names, approver counts and turnaround figures follow Postprism's 2026 turnaround standard, read 15 November 2026; treat them as a starting point and adjust from your own queue.
Keep a record you can retrieve per client
The record is the part that saves the relationship. When a client says the caption was never approved, you need the version, the name, the time and the note, not a search through email. ZoomSphere logs every approval, status change and copy edit in an activity log. Planable logs every edit, comment and approval with a timestamp and the user who made it. Cadenus's second test is exactly this: who approved which version, and when.
Store the record next to the post, not in a second system. An approval recorded in email while the post lives in a scheduler is the separation Kontentino warns about, and the handoff between the two is where the wrong version goes out. When a client's legal team wants a document, ZoomSphere's PDF export is one legitimate way to hand over the batch.
Test the paths you rely on, and do not read a missing feature row as proof. Sequence's comparison says an unlisted feature is not a verified absence, so a vendor page that never mentions approval tells you to ask, not to conclude. Run four checks in a trial: the client's link, an edit after approval, a post left pending at its slot, and the record of who said yes.
Keep the record honest about publishing limits. Postprism notes that LinkedIn's Posts API lets an integration update only a short list of fields after a post is created, so a wrong image has to be deleted and re-posted. That makes the pre-publish gate the only cheap correction you get.
Where Superdirector fits
Superdirector does not run this workflow for you. It publishes Instagram carousels and X text posts for a single account under account-level standing posting permission granted by the account owner. Growth is $199 a month for one AI-run operated account. Starter is $29 a month for MCP access with hands-on creation, where you work from Claude or ChatGPT. It is not a client review portal and does not replace an approval tool.
Where it changes the arithmetic: if your approval problem is a queue that never fills, the publishing side can run from that account-level permission. The client gates above still belong to your process. Decide who approves and what voids a yes before you automate anything downstream.