Automations & Scheduled Tasks
The agent doesn't only work when you're talking to it. Give it a schedule and it works while you sleep — preparing the morning briefing before you arrive, running the weekly report every Friday, checking the things you'd otherwise have to remember to check.
Because the agent runs on an always-on appliance at your facility, automations don't depend on your laptop being open or your browser running. They just happen.
Creating an automation
There's no scheduling interface to learn — you set automations up in conversation, in plain language:
"Every weekday at 7am, prepare a short briefing: yesterday's attendance vs the same day last week, any offline devices, and anything unusual. Notify me when it's ready."
"On the first of each month, run the end-of-month skill."
"Every hour during opening hours, check whether any door controllers are offline — only tell me if something's wrong."
The agent confirms what it understood — the schedule and the task — and from then on it runs. Times are interpreted in your facility's local time, daylight saving included.
What automations are good at
Three broad patterns cover most of what facilities build:
Briefings and reports. Anything you'd otherwise compile by hand on a rhythm — daily summaries, weekly performance reports, monthly comparisons. The agent pulls live data through its Performance Hub connection, writes the output to its workspace or straight into a notification, and has it waiting for you.
Monitoring and watchdogs. "Check X, tell me only if it's wrong." Device health, failed payments, unusual patterns. The valuable half of this sentence is the second half — a well-built watchdog is silent for weeks and then earns its keep in one push notification.
Housekeeping. Recurring tidy-up of the agent's own world — archiving old workspace files, refreshing a wiki page with current numbers, updating a standing document.
The strongest automations pair a skill (the how, refined until it's right) with a schedule (the when). Get the task perfect interactively first, save it as a skill, then schedule the skill.
Managing your automations
All conversational:
| You want to… | Say something like |
|---|---|
| See what's scheduled | "What automations do you have running?" |
| Change one | "Move the morning briefing to 6:30." |
| Pause or stop one | "Pause the hourly device check until Monday." / "Delete the old weekly report task." |
| Test one now | "Run the morning briefing now so I can see it." |
Scheduled tasks persist on the device, survive software updates and reboots, and are included in backups — a restored or replaced device comes back with its schedule intact.
Automations and spend
Unattended work is the most common source of unattended spend — an automation runs whether or not anyone's watching. Two habits keep it boring:
- Set a monthly cap (and ideally a daily one) under Budgets & limits — a daily cap is precisely the brake that turns a runaway automation from a bill into a blip.
- Right-size the frequency. A device-health check every hour is sensible; every minute is spend without insight.
If the agent hits a spend cap, scheduled tasks that need AI models pause until the window resets — and resume by themselves.
If the timing matters, mind the power
Automations run on the device, so they run only while the device is powered and — for anything needing models or Performance Hub data — online. This is one more reason for the placement advice on the hardware page: constant power, wired network. A brief outage doesn't break an automation's schedule; it just means a run that fell inside the outage is missed.
Related pages
- Skills — packaging the how before you schedule the when
- Notifications — how automations reach you
- Budgets & limits — keeping unattended spend bounded