One page that reads every voice room for you: a live timeline, plain-English summaries of what each voice room is saying, and the signals that need a look — with the actions to deal with them.
On a busy day nobody can listen to every voice room at once. Monitoring listens for you. Open Incidents → Monitoring and one page shows what is happening across the whole workspace right now: how much each voice room is carrying, a running summary of what each one is saying, the public reports and keyword alerts arriving alongside, and — when several of those start pointing at the same thing — a signal that says so, before anybody has declared anything. The page keeps itself up to date; you never need to refresh it.
The idea underneath it is simple: things that line up happened together. If the Gate C stewards start talking about a queue, then Control starts talking about Gate C, then two members of the public report pushing at Gate C, that is one situation forming — and it is visible on this page as three lanes lighting up at the same moment, minutes before it becomes an incident.
monitoring.view, which every role that can view Security Operations holds, along with workspace admins. Acknowledging a signal, raising an issue, declaring an incident, dismissing a signal and standing a voice room down need security_ops.manage. The voice room summaries and signals are written by an AI model from the voice rooms' transcripts: they are something to check, not something to trust. Every summary carries a View conversation control that shows what was actually said.The top of the page is one line: a word for how the workspace is, a sentence saying why, and the figures behind it. The word is one of four, and it only takes a colour when there is something to look at — a calm workspace is a quiet grey line.
The sentence after the word is the situation summary: a few lines written from everything on the page — every voice room's summary, the open issues, the newest reports and alerts — and rewritten whenever any of them changes. For example: "Gate C is under crowd pressure: radio traffic in three voice rooms and two public reports describe pushing against the barrier. A request to open the relief gate has been outstanding for several minutes." The figures after it read left to right: open signals, voice rooms talking, voice rooms of concern, incidents, reports to triage, open issues, new keyword alerts. A figure takes a colour only when its number needs attention.
The pulse is the page's timeline. Every voice room is drawn as a stepped line over the last hour, three hours or six: each step is one minute, and taller is busier. The line is one colour, so height is the only thing it says. The thin coloured strip under a voice room is how concerning its conversation was, from its summaries, in the same blue, amber and red as the badge beside its name. Small pins mark where a keyword alert, a public report or an incident happened: a circle, a square and a triangle. An incident or a public report also draws a faint line down every voice room, so you can see what each voice room was doing at that moment. Each voice room's total for the window sits at its right-hand end, and the right-hand edge is now.
When one voice room stands out — at high or critical — a line above the chart names it as the voice room to look at, with a Declare incident control beside it for people who can declare. The Declare launcher opens prefilled from that voice room's summary, so the record says what prompted it.
View conversation, on the voice room in focus, shows the transcript the summary was written from, right there under it: each message with its time and speaker, the messages the summary relied on marked, and a little either side for context. This is the one place on the page that shows a voice room's conversation itself, it only opens when you click it, and each opening is recorded in the workspace's audit trail. People who hold Replay access also get Open in Replay to hear it.
A signal is the page saying look at this. One is raised when the same subject is coming up in more than one voice room, when radio traffic matches an open issue or a public report, or when any single topic reaches high or critical. Each signal has a title, a sentence or two on what it is and why it matters, a severity from low to critical, the voice rooms it was heard in (each a way into the voice room in focus) and any issue it is tied to. Signals sit in a strip under the condition band, most severe first, and the strip is simply absent when there are none.
A signal is kept up to date rather than raised again: as the situation changes its detail and severity move with it, and if it is raised to a higher severity you are told. When nothing supports it any more — the voice rooms have moved on, the issue is resolved — it clears by itself. A signal that goes unseen for about half an hour clears too, so the page returns to calm without anyone tidying it.
A voice room's summary is rewritten whenever the voice room says something new — and only then. So a voice room that fell silent after a worrying reading would, left alone, look worrying long after the matter was over. Two rules keep the page honest. First, a reading counts as the voice room's current concern for fifteen minutes. After that it is shown as history — High · earlier, greyed, with its age on hover — and it no longer drives the condition, the figures or the lane. The next thing the voice room says is read afresh.
Second, a voice room can be stood down. The situation summary sees every voice room at once, so when other voice rooms make clear that the matter a voice room's reading described is over — Control says the casualty is with Medical and the gate is flowing — it stands that voice room's reading down in the same pass, and the voice room shows as Stood down. You can do the same yourself: Stand down on the voice room in focus says this has been dealt with, is recorded with your name, and re-runs the situation summary so any signal resting on the voice room can clear. Either way, standing a voice room down never silences it: its next reading counts as normal.
Wallboard, in the page header, puts the board over the whole screen for a control-room wall: no menus, the same board a little larger, and a clock in place of the title. While nobody has chosen a voice room, the voice room in focus walks through the voice rooms of concern every twenty seconds, most concerning first; choose one and it stays. Press Esc, or the button again, to leave.
Voice room summaries are on for a workspace unless you turn them off, and the switch is yours: Security Ops → Settings → Monitoring holds the workspace switch and a switch per voice room, so a voice room whose conversation should not be summarised can be left out while the rest carry on. Direct calls between two people are never summarised, nor are the private lines to outside callers, nor voice rooms shared in from another workspace — those belong to their owner's page. A summary says what is being said, how concerning it is, and who reported it or is dealing with it: what people said and did, never what they are like. The language model never sees a name. Each speaker reaches it as a label, and the name is added when the summary is shown, as a full name or a first name: the same settings page holds that choice, and it applies to summaries already written as well. Signals name nobody, because they are sent on as notifications and copied into Issues. Summaries and signals are kept for a limited time, and a summary is deleted as soon as the conversation it was written from is.