Ten agents, each built for one part of a channel’s day. They do the repeated work of a traffic and scheduling team inside your rules — and every change waits for a person’s approval before it reaches playout.
What each agent does, the rules it keeps, and when it stops to ask a person.
Builds the next week from the channel’s template and its running orders — next episodes in order, repeats, rotations — then lists what changed from last week.
Keeps: Template blocks, episode order and repeat offsets, air windows and rights dates.
Asks a person: Before the week is published, and before a programme leaves its block.
Times the playlist so fixed starts — the news at 19:00 — land on the exact frame, by balancing segment splits, break lengths and fillers.
Keeps: Fixed starts, shortest and longest breaks, the channel’s frame rate.
Asks a person: Before shortening a programme segment or moving a fixed start.
Places the day’s booked commercials in breaks, gives a reason for every placement, and proposes other breaks for spots that do not fit.
Keeps: Clash codes, break lengths, advertising minutes per hour, air windows and booked dayparts.
Asks a person: Before dropping a booked spot or moving it to another day.
Schedules promos ahead of the programmes they promote and spreads each campaign over its passes.
Keeps: Never inside the promoted programme; campaign dates and passes per day.
Asks a person: Before a promo takes a place meant for a booked spot.
Fills what is left between programmes and in each break with fillers, bumpers and idents, to the frame.
Keeps: Rotation lists, item lengths, no item twice in a row.
Asks a person: Before using an item from outside its rotation list.
When sport overruns or news breaks in, re-flows the rest of the day and moves the spots that no longer fit.
Keeps: Starts that must hold, spots that must air inside their window.
Asks a person: Before any change to the day on air.
Before playout export, runs the channel’s validators on every item — timing, gaps, overlaps, missing media, expired rights — hands each failure to the agent that can fix it, and checks again.
Keeps: Which validators block air and which only warn.
Asks a person: Before export while a warning is still open.
Matches the as-run log to the plan, explains each difference and lists the spots that need a make-good.
Keeps: The channel’s tolerance for timing differences.
Asks a person: Before the day is closed.
How as-run reconciliation worksReads new titles from spreadsheets, e-mails or your media asset management into the catalog, and writes programme guide texts in several languages.
Keeps: Your catalog fields and naming, guide text lengths per platform.
Asks a person: Before new catalog entries are created.
Answers questions in plain words — “When does episode 12 air?”, “Which spots missed air yesterday?” — with links to the airings, orders and items behind the answer.
Keeps: Only what the person asking may see.
Asks a person: Never needs to: it changes nothing.
The agents are not ten separate tools. They share your channel’s data, checks and skills, and hand work to each other.
Validators find a problem, the connected agent fixes it, and the day is checked again until every check passes.
See itBuilt-in checks and your own rules in plain words, chosen per channel.
See itYour channel’s way of working, written once and given to the agents that need it.
See itConnect Claude, ChatGPT, Grok or your own agent through the open Model Context Protocol.
See itNothing reaches playout until a person approves the change.
An agent acts with the roles of the person who runs it, on the channels they may touch.
Each step is logged with its reason, and every change can be undone.
Schedule data stays in your company’s space and is not used to train AI models.
Early access starts with a demo on a channel like yours, then a trial on your own catalog and templates.
They are in preview, as part of early access. Ask in the demo which agents fit your channel.
You choose. The agents can run on the AI provider your company already trusts, and agents you build yourself connect through MCP with only the roles you grant.
Your channel in 9eye — catalog, templates and orders — and the skills you give it: your channel’s rules and way of working, in plain words.
Yes. Each agent works only on the channels and jobs you give it, and only starts when you ask, on a schedule you set, or on the events you choose.
They are part of the Platform + AI agents plan: $2,390 per channel per month, with all ten agents.