Olly dashboard watch
Dashboards are where teams look to check system health, but keeping an eye on them is manual, and traditional alerts often add noise without telling you what actually changed or why. With Olly dashboard watch, you hand a custom dashboard to Olly: it checks the dashboard on a schedule, investigates what it sees, and sends you a short narrative report explaining what changed, the likely cause, and where to drill in.
Dashboard watch is the first Olly Watch experience. You start it from the dashboard itself, tell Olly what matters in a short chat, and Olly takes over from there. Each watch then runs as a scheduled task, so run history, notifications, sharing, and editing all behave the way they do for any other Olly automation.
Turn a dashboard over to Olly in a couple of clicks: Start a watch.
What you need
- Olly turned on for your team. See Enable Olly.
- A custom dashboard to watch.
- A watch belongs to the user who created it and stays private until you share it for read-only access.
Start a watch
You start a watch from the dashboard you want Olly to monitor.
- Open the custom dashboard.
- In the top right of the dashboard, select Olly AI to open the Olly drawer.
- Under Actions Olly can take, select Watch this dashboard. Olly greets you and begins setting up the watch.
When a watch is created, a Watching badge appears beside the dashboard's name.
Actions Olly can take also lets you Build an alert, for when someone needs to be paged the moment a threshold breaks rather than told on a schedule what changed. Below it, Questions you can ask suggests starter prompts for this dashboard, such as a health read of the selected time range, or the spikes and drops that stand out from the baseline. Select a suggestion to ask it straight away, or type your own question instead.
The dashboard travels with the conversation as a Dashboard Page context chip, so you can ask follow-ups without re-explaining what you're looking at. To ask about a single widget rather than the whole dashboard, select Add to Olly context on the widget's toolbar instead. See Ask Olly about a widget.
Tell Olly what to watch
Setting up a watch is a short conversation; you don't write a scheduled prompt yourself. Olly looks at the dashboard and asks you three things:
- What to monitor: describe what matters on this dashboard, such as error spikes on a specific widget or a daily status summary. If you're not sure, ask Olly to investigate the dashboard and suggest what's worth watching.
- When to check: how often Olly should look at the dashboard, and when to start.
- Where to send the results: whether each run's report should be delivered as well as stored on the watch. Destinations are Notification Center connectors, and a scheduled task can send to email, Slack, or Microsoft Teams. See Get notified when a task runs.
You can add detail in the conversation (a threshold, a widget to focus on, the kind of report you want), but nothing more than the dashboard and a schedule is required. When Olly has enough to go on, it drafts the watch, shows it to you for approval, and adjusts it if you ask. After you approve, Olly confirms that the dashboard is now being watched and gives you a link to manage it.
What each run produces
At the cadence you set, Olly reads the whole dashboard and analyzes it, rather than comparing a single value against a threshold. Each run gets a chat of its own and produces a narrative report. The report opens with a one-line verdict, such as a daily health status, then gives the findings behind it: what moved, by how much, compared with which earlier period, and what Olly thinks caused it. Figures are quoted inline, Olly builds dynamic tables for the breakdowns it used, and the chat keeps the reasoning trace so you can expand it to see which queries ran.
Keep two things in mind as you read a report. Olly re-derives the data with its own queries rather than reading the dashboard's rendered widgets, so a figure can differ from the widget you remember when the two cover different time ranges or groupings. And Olly infers the explanations from that data rather than recording an actual cause, so treat the numbers as observed and the "why" as a lead to confirm.
A dashboard can also go out on a schedule as a PDF, through Schedule a report in the dashboard's more actions menu. The two features look similar and answer different questions: a watch sends Olly's reading of what changed and why, while a scheduled report sends the dashboard as it renders, every time, with no interpretation. A dashboard can have both.
Manage your watches
Manage a watch from the dashboard it watches, or from the Scheduled tasks page.
From the dashboard
A watched dashboard carries a status badge beside its name. The badge appears only once a watch exists, and it tells you whether the watch can run:
| Badge | What it means |
|---|---|
| Watching | The watch is running on its schedule. |
| Paused | The watch is set up but paused. |
| Out of AI units | The watch keeps its schedule but can't run until your team has AI units again. |
| AI consent off | Your team turned off AI-Powered Capabilities. The watch keeps its schedule but can't run until they're turned back on. |
Select a Watching or Paused badge to open the watch panel. Headed Olly Watch, the panel shows when the next run is due, includes the active toggle, and lists the watch's recent runs with the time each one started. Select a run to open its chat and read the report. Before the first run, the panel says No runs yet and tells you when that run is due.
The other two badges open an explanation instead of the panel. Hover over the badge to see what's blocking the watch, along with a way to act on it: Open AI settings for a team that turned AI off, and View plan for a team that's out of AI units.
From the panel you can also pause and resume the watch with the toggle in its header, and, from the menu, select Manage Schedule to open the watch on the Scheduled tasks page, or Delete configuration to remove it. Coralogix asks you to confirm a deletion, because it permanently deletes the dashboard's scheduled task. Chats from past runs stay in your history.
From the Scheduled tasks page
To see all of your watches in one place, in Coralogix select Olly AI in the navigation bar, then Scheduled tasks, and open the Dashboard watch tab. Your watches live here, separate from your other scheduled tasks, and both tabs count toward the same limit. To watch another dashboard from this tab, select Add dashboard to watch, then select a dashboard.
Each watch appears as a card showing its name, an Active toggle, when the next run is due, and the name of the dashboard it watches. Select the dashboard name, on the card or on the watch itself, to open that dashboard. Select a card to open the watch, where you can read its prompt, schedule, model, and run history. From the watch, or from the card's more actions menu, you can also:
- Edit: change the name, what Olly monitors, the model and reasoning effort it runs with, how often it runs, and where its reports go. Select Edit manually to fill in the form, or Edit with Olly to describe the change in chat. See Edit a task.
- Share: give your whole team, or specific teammates, read-only access to the watch, its run history, and the chat from every run. Sharing never lets anyone else edit, run, or delete it. See Share a scheduled task.
- Run now: run the watch immediately, outside its schedule. This doesn't affect the next scheduled run.
- Delete: remove a watch you no longer need. Chats from past runs stay in your history.
Limits
- Each dashboard can have one watch. Olly tells you if the dashboard you pick is already being watched.
- A watch stays bound to the dashboard you started it from. To watch a different dashboard, start a new watch.
- A watch counts toward your scheduled task limit, which covers watches and other scheduled tasks together.
- Each watch run consumes AI units, like any other Olly activity. If your team runs out of AI units, watches stop running until more units are available. See Usage Management.
Next steps
Manage the runs, history, and schedule of every watch alongside your other automations in Scheduled tasks.