Every agency has an approval process. Most of them are a group chat, a spreadsheet and a person chasing. This is what to replace that with, why the obvious fixes fail, and how to keep sign-off from becoming the slowest part of the month.
Nobody loses a client because a draft was mediocre. They lose one because eleven posts sat unapproved for three weeks, the client saw nothing on their profile, and the invoice arrived anyway. The writing is rarely the bottleneck. The yes is.
The reason is structural rather than personal. Your client did not buy a project-management tool, does not want a login, and is answering you from a phone between meetings. Any process that asks them to remember a password, open a board or scroll a thread will lose to the thing they were already doing. So the workflow has to meet them where they are: one message, one decision, no account.
Below are the four ways this actually gets done in agencies today. Three of them break at a predictable point, and it is worth knowing exactly where.
Each of these works at two clients. None of them survives ten.
Fast to start and genuinely fine for one client. It fails the moment two people reply with different edits, or a “yes” lands under a photo of somebody’s lunch. There is no record of what was approved, so when a post causes a problem, the argument is unwinnable.
Breaks at 2 clientsBetter: comments are threaded and history is kept. But the document is a parallel universe to the thing that actually publishes, so somebody re-types the approved version into the scheduler, and that is where typos and stale edits enter. Nobody notices until it is live.
Breaks at 5 clientsOn paper this is the clean answer. In practice most clients will not maintain a password for a tool they open twice a month, seats cost money, and asking a busy executive to learn your software is a request they are entitled to refuse. Adoption, not capability, is what kills it.
Rarely adoptedThe version that survives a roster has two properties. First, the person approving needs no account — they open a link, read the post as it will appear, and either approve it or say what to change. Second, the thing they approved is the thing that publishes; there is no re-typing step between the yes and the post. Everything else — reminders, roles, audit trail — is detail. Those two properties are the workflow.
Set it up once per client. After that it should run without anyone chasing.
Approvals get slow when the client has to rewrite before they can agree. A draft that already sounds like them gets a yes on the first pass, which is the single largest saving in the whole process.
Not a document, not a screenshot, and not five posts at once. One link showing the post as it will look on LinkedIn, with the scheduled time visible, so the decision is obvious and takes under a minute.
Approve, request a change, or reject. More options invite deliberation. A comment box under “request a change” catches the actual note, and it lands against that post rather than in a thread nobody can find later.
No copy-paste into a scheduler. The approved text is the published text, the timestamp is recorded, and if the post ever causes a problem you can show who agreed to what and when.
This is the case every process eventually meets, and most handle badly. The client has not said no. They have simply not opened the link for nine days, and your calendar is now full of posts that cannot go out.
Do not publish anyway. It is tempting, it is occasionally what they wanted, and it is the fastest route to a conversation you cannot recover from. Publishing without a yes turns a scheduling problem into a trust problem.
Do not stop drafting either. Empty queues are how a paused month becomes a cancelled retainer — when they finally come back you want a fortnight of ready posts, not an apology.
What works is making the silence visible rather than personal. A short status the client can see — four posts waiting, oldest one nine days old — puts the delay where it belongs without anyone having to say so. When it comes up on the monthly call, you are describing a queue rather than making an accusation, and the fix is usually that the wrong person was named as approver in the first place.
If you write for anyone in financial services, healthcare or a listed company, approval is not an efficiency question but a compliance one. Three adjustments cover most of it. Add a second reviewer whose only job is the regulatory read, so the marketing yes and the compliance yes are recorded separately. Never allow silence to approve. And keep the record of the exact published text rather than the draft, since that is the version anyone will ask about.
None of this needs specialist software. It needs a workflow where the approved version and the published version are provably the same text, which is the property most improvised processes quietly lack.
Approval links your client opens without an account, and the approved text is what publishes. Connect one account and watch it work before you move anyone across.
Fourteen days, no card. Connect one account, send one approval link, and see how fast the answer comes back.
Per connected account. Teammates and approvers are free.