Skip to main content

​
Overview

Scheduled Jobs are not currently available. The feature is turned off across the platform while it’s being reworked. What follows describes how it’s designed to behave when it returns — you can’t create or run scheduled jobs today.
Scheduled Jobs run Dexter prompts on a recurring or one-off schedule, starting a fresh conversation each time they fire. Use them to automate recurring factory checks, report generation, simulation runs, and data refreshes without anyone having to remember to ask. Jobs are created and managed in conversation with Dexter: describe what you want automated and when, and Dexter sets up the job for you.

​
Creating a Scheduled Job

Tell Dexter what to automate and when, in plain language. Dexter translates the cadence into a schedule, confirms the interpretation with you, then creates the job:
  • “Every weekday at 5am, run the simulation and generate a PDF of the results”
  • “Every Monday morning, refresh the production data”
  • “Run a Monte Carlo batch on the first of every month”
  • “In 2 hours, generate the quarterly throughput report”
Schedules run in your timezone unless you name a different one. Dexter confirms the interpreted schedule before creating the job, so you can correct any misreading.

​
Recurring vs. One-Off

  • Recurring — runs on a cron cadence (daily, weekly, monthly, custom intervals) until you pause or delete it.
  • One-off — fires once at a specific future time: “On April 1st at 8am, run the quarterly analysis.”
For complex procedures, create a custom workflow first and have the job reference it: “Run the weekly quality audit workflow every Monday at 6am.” The job prompt stays clean, and the procedure keeps improving through the workflow’s own reflection loop.

​
How Scheduled Jobs Run

When a job fires:
  1. Dexter opens a new conversation with the job’s prompt as the first message.
  2. Dexter executes the instructions autonomously — running simulations, generating reports, refreshing data, anything it can do interactively.
  3. Outputs are saved exactly as if you’d asked in person: reports land in the Reports tab of the Data page, model changes take effect in the factory.
  4. The conversation is saved to your conversation history, named for the job.
Scheduled runs are ordinary Dexter conversations. Open one from your history to audit exactly what happened — every step, tool call, and output — and follow up in that conversation if something needs attention.

​
Managing Jobs

Ask Dexter to list, inspect, or change your scheduled jobs at any time:
  • Edit — change a job’s schedule or prompt: “Move the daily report to 6am.”
  • Pause — stop the job from firing while keeping its schedule and run history: “Pause the daily report job while the line is down.”
  • Resume — pick the schedule back up. The next run is computed from now; firings missed while paused are skipped, not backfilled.
  • Run now — trigger a job immediately without waiting for its next scheduled slot.
  • Delete — remove the job permanently, along with its run records.

​
Limitations

  • One factory per job. Each job is bound to the factory it was created in. To automate work across factories, create a job in each.
  • No interactive input. Scheduled runs are autonomous — write prompts that carry all the context Dexter needs, because nobody is there to answer clarifying questions mid-run.

​
Best Practices

  • Keep prompts specific. “Generate a PDF report on last week’s throughput for the CNC line, including utilization and cycle-time KPIs” beats “send me an update.”
  • Name the artifacts. Reference specific models, data sources, and naming conventions in the prompt.
  • Lean on the knowledge base. The more context Dexter has about your operation, the closer unattended runs land to what you’d have asked for.
  • Review the first runs. Check the first one or two conversations a new job produces and tune the prompt if anything’s off.
Scheduled jobs execute with your permissions. Anything Dexter does during a scheduled run has the same effect as if you’d asked interactively — changes to models, BOMs, or configurations take effect immediately.