It leaves the gap rather than fill it
When nothing in the archive honestly supports a line of script, the slot comes back empty and marked, with an explanation. A gap an editor can see is worth more than a shot nobody checked.
Editorial trust
A newsroom tool has to be able to say “I don't know.”
Everything below is a constraint the software enforces, not a value it professes. They are commitments in the same sense that a locked door is a commitment — you can check. Where something is a default rather than a law, we say so, and we say where the setting lives.
- 07
- Refusals it enforces
- Constraints, not values it professes
- 04
- Governance controls
- Access, audit, sessions, location
- 00
- Automatic routes on by default
- Each is switched on deliberately, and recorded
01 / What it refuses to do
Seven refusals.
It rejects footage that contradicts the script
If a line names one person and the footage shows another, the shot is refused outright — not down-weighted, refused. The same applies to a different place or a different kind of event. Person substitution is the most damaging mistake a package can make.
It will not write past its sources
Where a story has a hundred words of fact behind it, you get a hundred words. A draft that cannot be supported is refused with an explanation rather than padded out to look like an article. Every quotation is traced back to a filing that contains it, and anything untraceable is flagged before publication rather than after a complaint.
It suggests. You decide.
AI can suggest or carry out trimming, reel assembly, headline creation and publishing. Your team switches automation on or off for each step and decides where review is required. With automation off, proposals wait for a person; with it on, the configured step proceeds automatically. The record remains available in either mode.
Nothing publishes by accident
By default everything the platform produces stops one step short of an audience and waits for a person, delivered in the least public state its destination allows. Automatic publishing is available, switched on per destination by somebody who holds that permission, and recorded when it is. The distinction the software enforces is not between manual and automatic; it is between a route you chose and a route that happened. Switching a route on costs you nothing else: the same reasoning travels with the work, the same entry is written to the audit trail, the same person is named as having allowed it, and the route closes again in one action. Automation changes who presses the button, not what you can check afterwards.
It shows its reasoning, and can be overruled
Every match arrives with why it was chosen and every alternative it considered, one click away. Every classification can be opened to see what drove it. A choice you cannot inspect is a choice you cannot overrule.
It says when it does not know
An unrecognised face is an unnamed recurring person, not its closest match. Where the platform will not do something, it says so rather than producing something plausible. Where a score is based on general practice rather than your own audience data, it says so on screen.
1 / 03
Everything holds, by default
A finished piece of work does not go anywhere on its own. It waits for a person, and the person is named in the record. This is the state every route starts in, on every destination, on the day you switch the platform on.
2 / 03
Automatic is a route you switch on
Publishing without a person is available, per destination and per category. It is off until somebody with the permission turns it on, and turning it on is itself written to the record. Nothing about it happens quietly.
3 / 03
Two routes, one record.
A newsroom that automates nothing will not keep up, and one that automates everything will eventually publish something nobody read. The setting exists so a desk can move the line where it wants it — per destination, per category, once it has watched the output long enough to know. What stays fixed is that the line is somewhere you put it, and that the record shows which side of it every published item came from.
Why this is the product
Confidence is the cheap part.
It costs nothing to build a system that answers every question. Every automated newsroom tool can produce a headline, a matched shot and a caption for anything you give it, and most of them do. What costs something is the refusal: admitting the archive fell short, handing a problem back to a producer who is already short of time, saying plainly where the platform will not go. A newsroom that cannot trust the empty slots cannot trust the filled ones either. It is also why automation here is something you switch on rather than something you switch off — a desk that has watched the drafts for a month is in a position to automate a route, and a desk on its first day is not.
02 / Governance
Who did what, and when.
The editorial commitments above only mean something if you can see who exercised them. Access, attribution and history are treated as part of the editorial product rather than as an administrative afterthought.
Five levels of access
What somebody can see and do is set by their role, and the roles map onto how a newsroom actually works — a trainee can build a package all evening without being able to send it anywhere.
An audit trail you cannot edit
Who wrote, who corrected, who reclassified, who released — recorded permanently with a name and a time. An archive without a history of its own edits is an archive nobody can audit.
Sessions you can end instantly
Every active session is visible with where it is signed in from, and any of them can be ended immediately. When somebody leaves, their access ends when you say it does.
It stays in your building
The archive, the recordings and the transcripts live on your own installation. A closed briefing does not leave the premises to be turned into text.
See it against your own archive.
The useful demonstration is the one that runs on your footage and your wire feeds. Tell us how your desk works and we will show you the parts that matter — and say so where we do not help.