dispatch.xonik.com

Dispatch — by XONIK

Every claim, with the source still attached.

Dispatch schedules and publishes LinkedIn posts for agencies and in-house teams. What makes it different is what travels with the post: the figure, the publisher, the date, the sentence it was taken from — and a check that keeps running after the post is live.

We took apart 35 tools before writing a line: 16 social suites and 19 LinkedIn-specific products. Not one attaches a source to a claim. Their approval workflows approve copy. Nobody approves the evidence.

See how it works What it costs

Rae WhitfieldOperations Director · posted 12 days ago

Recruitment, not demand, is what is holding output back. Last quarter 61% of manufacturers named hiring difficulty as their single largest constraint — and productivity moved just 0.4% over the year.

Sources carried with this post

61%Make UK — Manufacturing Outlook 2026 Q2, 2 June 2026. “61% of manufacturers said recruitment difficulty was the single largest constraint on output in the quarter.”

0.4% Office for National Statistics — Labour productivity, UK, 14 July 2026. “Output per hour worked rose by 0.4% in the year to Quarter 1 2026.”

▲ This source has moved. The passage now reads 0.2%. It said 0.4% when the post was written, and the post is still live. Checked 4 hours ago.

The demonstration account, rendered exactly as Dispatch renders it. The figures and the flag are seeded demo data, not a claim about any quarter.

What it does

A scheduler that can answer where the number came from.

Three things happen to a post here that happen nowhere else. The rest — calendar, queue, approvals, client links, analytics — works the way you already expect it to.

Research before writing

Sources are gathered and read first: publisher, title, date of publication and the exact sentence a figure came from. The post is written against them, not decorated with them afterwards.

Evidence, approved with the copy

A figure in the body with nothing behind it is flagged before anyone is asked to sign off. Approval needs a named person and is bound to a fingerprint of the exact text — edit an approved post and the approval falls away, including when the edit puts the words back.

Watched after publishing

Cited pages are re-fetched on a schedule and compared with the sentence recorded at the time. When a page changes, the live post resting on it is named, with the old figure, the new one and the date of the check.

How this differs

One column of the comparison came back empty.

Every feature we expected to find, we found: white-label domains, multi-stage approvals, no-login client review, scheduled first comments, bulk import, best time to post. Most of what a scheduler does is table stakes and we say so. Then there is this.

Capability	16 social suites	19 LinkedIn tools	Dispatch
A research step before writing	none	none	yes
A citation attached to a claim	none	none	yes
Verification of a statistic	none	none	yes
A currency check — this figure was superseded	none	none	yes
A check the cited page still says it	none	none	yes
A claim-level risk flag: opinion or evidence	none	none	yes
Slop detection — while all 35 ship AI generators	none	none	yes

XONIK teardown, September 2026: 35 products checked feature by feature — 16 social management suites and 19 LinkedIn-specific tools. Every row above was absent from all of them.

LinkedIn has already made this turn.

In July 2026 LinkedIn discontinued its own “enhance your post” AI writer and shipped a proofreader in its place, beside a reader-facing “seems like AI slop” control — over a million uses in three weeks, and roughly 40% fewer views on the posts it flags. The platform moved from generation to assistance. Thirty-five third-party products did not.

Scheduling is 5% of the hours and 100% of the trust.

In agency retainers scheduling takes one to two hours of a seventeen to twenty-eight hour month; research and content take ten to fifteen. And clients who leave name delivery and value, not price — while three quarters of agencies believe they left over budget. A tool should defend the ten hours, not automate the one.

Screens

The application, not a mockup.

Photographed against the demonstration account while it was running.

calendar screen of the application — **Calendar** — The month across every brand in the account. A post sits in a slot, and the slot carries the destination's own posting rules.

composer screen of the application — **Composer** — Body, image and alt text, link, first comment and sources in one window — with unsupported figures called out before sign-off.

sources screen of the application — **Sources** — The evidence shelf: publisher, date, the figure and the sentence it came from, when it was last checked — and one source marked as moved.

review screen of the application — **Client review** — The client reviews without an account: a signed, expiring link scoped to one brand, showing the post as it will appear with its sources beneath it.

connections screen of the application — **Connections** — One LinkedIn application, and an honest account of it: the scopes actually granted, and for anything out of reach, which of the three reasons it is.

Who it is for

Three people open this and see three different things.

The agency

Ten client brands in one account, each isolated: its own destinations, voice, sources and media, with nothing shared by accident. A monthly report that shows the work rather than describing it.

The company posting for itself

One brand, its page and its founder's profile. Sign-off sits with a named person. When a board, a customer or a regulator asks where a figure came from, the answer is one click and it is dated.

The client being posted for

Opens a link, sees the post as it will appear with its sources underneath, approves it or sends it back with a reason. No account, no password, no seat to pay for.

Pricing

What an account gets.

Limits are counted per account rather than per brand, enforced by the software, and refunded when a scheduled post fails.

Trial

Free

no card

1 brand
1 destination
5 posts a month
the evidence layer in full

One brand

$960

per year

1 brand
3 destinations
40 posts a month
client review links
monthly report

Agency

$2,400

per year

10 brands
30 destinations
400 posts a month
per-brand isolation
operator console

Custom

Talk

self-hosted or above

unlimited brands
your own infrastructure
retention set by you
data processing addendum

Standing

Where it stands, written down.

Built

The application and the evidence layer, proved by automated checks that include the panels driven in a real browser.

Submitted

The LinkedIn application: one developer application, verified against the company page.

On approval

Development tier while under review — 500 calls a day for the application, 100 per member. The software budgets for that tier rather than the one after it.

Official API only

No scraping, no automated connection requests, no automated messages, no scraped lead database.

Not claimed

No customer results, case studies or ROI figures. There are none yet, and a product about evidence cannot open with an unsourced number.

Questions

Before you ask.

Do I need LinkedIn API approval to start?

No. Composing, sources, evidence, approvals, client review links and reports all work today. Publishing is the one part that waits for LinkedIn to approve the application.

What happens when a source I cited changes after the post is live?

Dispatch re-fetches the page on a schedule and compares it with the sentence recorded at the time. When it differs, the live post resting on it is named, with the old figure, the new one and the date of the check. It flags; it never edits or deletes a published post.

Does my client need an account to approve a post?

No. They open a signed, expiring link scoped to one brand, see the post as it will appear with its sources underneath, and approve it or send it back with a reason.

Can one agency see another agency's brands?

No. Every row carries the brand it belongs to and every query is filtered by it. Cross-account reads are the failure most likely to end a company, so isolation is proved by a dedicated suite of automated checks rather than by convention.

Is anything published automatically?

Never without an approval carrying a name. Approval is bound to a fingerprint of the exact text, so editing an approved post drops the approval — including when the edit puts the words back.

Do you scrape LinkedIn or automate connections and messages?

No. Official API only. No scraping, no automated connection requests, no automated messages, and no scraped database of people.

What does it cost, and how are the limits counted?

Plans run from a free trial to an agency plan. Limits are counted per account rather than per brand, enforced by the software, and refunded when a scheduled post fails.

Who writes the posts?

You do. Research briefs and drafts are assembled with the help of a language model, and everything generated is labelled as such and waits for a person to approve it.

Start with one brand.

The trial takes an email address and posts five times. If the evidence layer is not worth it by the fifth post, nothing has been spent.

Open an account Read the trust page