Automations and schedules

Your organization's CrewAI automations on AMP are part of what Crewdesk can work with. See Automations for the concepts.

Running one

"Run the weekly sales report for the north region."

The agent finds the automation, reads the inputs it takes, fills them from what you said (and asks for what's missing), and proposes the run on a card. When the run finishes it reports the result; if the automation lost data along the way, it says so instead of presenting a partial answer as complete.

Reading runs and diagnosing

"Why did the invoice reconciliation fail last night?"

The agent reads the run history, and for a failure it calls the automation diagnosis specialist, which investigates the run's trace before suggesting a fix. Your organization's knowledge notes about that automation go with it.

Asking data questions

"How many orders were cancelled in September, by reason?"

When your organization has a query automation, the data analyst specialist answers through it: schema and samples first, then the query, then the answer, with a table or chart.

Scheduling

"Run the inbox digest every weekday at 8:00."

A schedule is a write: you approve it on a card. It runs at that time in your time zone, with the inputs you set. It pauses after three failures in a row, or when the automation's inputs change, and tells you. Removing a schedule asks in every mode.

• /schedules in the app or terminal: your schedules, how each last ran, which are paused.
• crewdesk schedules and crewdesk schedules resume N from the shell.

Creating a new one

"Every Monday, collect last week's support tickets, group them by topic and post a summary to Slack."

For a new recurring job, the agent follows a plan you can see: create the automation, schedule it, and test it with a first run, each step on your yes. If your open folder already holds a CrewAI project, the agent deploys that project rather than creating a new one.