Alerts
The Alerts page (owner-only) is a live log of every alert that has fired in your organisation, plus the settings that decide when alerts fire.
Three kinds of alert
| Kind | Source |
|---|---|
| AI | The AI noticed something matching one of your alert prompts. |
| Suspicious match | The desktop tracker matched a rule in your suspicious app / window-title list and produced a short evidence clip. |
| Direct send | You — or another owner — pushed an ad-hoc nudge / warning to a specific employee from the dashboard. |
Every alert has an employee, a timestamp, and (for AI and suspicious rows) a short evidence clip you can play back inline.
The log
The top of the page lists every alert chronologically, newest first. For each row you can:
- Play the evidence clip (if any) without leaving the page.
- Open the detail panel to see the AI's reason, the matched rule, or the message that was sent.
- Jump into the employee's timeline at the alert's timestamp.
Send a direct alert
The Send direct alert button opens a modal where you can:
- Pick an employee.
- Write a short message (it shows up as a desktop notification on their machine, surfaced by the tracker).
- Optionally tag the message as a nudge (gentle) or a warning (stronger).
- Send.
Direct alerts are logged in the same Alerts feed.
Configuring alert rules
The Settings subpage (top-right button) lets you configure:
- Alert prompts — free-text instructions you give the AI for what to watch for. Common examples: "abusive language toward customers", "browsing job-hunting sites", "extended distraction blocks".
- Suppression rules — quiet down alerts that fire too often (per employee, per app, or globally).
- Desktop notification surfaces — whether the employee sees a popup when an alert about them fires, and what it says.
Defaults are conservative: out of the box, no alert prompts are active. You opt in to each surface deliberately so employees aren't showered with notifications.
Limits
Configuration of suspicious-app / window-title rules themselves lives under Sensitive-app exclusions in the desktop policy, since the same lists feed both the censoring behaviour and the alert evidence-clip behaviour.