Appearance
Jobs
Jobs combine reusable tasks into a workflow. Use endpoint jobs for device maintenance and Infrastructure Jobs for central work or coordination.
Use Job Designer to author workflows and Job Operator to run and review them. Some operations also need a specialist role, such as AMT or Hyper-V management; see Access Control.
Create and run a job
- Open Operations → Endpoint Jobs and create a job.
- Add the tasks needed for the outcome, using Available Tasks as a reference.
- Configure requirements, timeouts, and error behavior for the workflow.
- Save it and test it on representative endpoints.
- Assign it to the intended scope or run it for selected endpoints. Add automations when it should run regularly or in response to events.
For a workflow involving both a Konnektor and card terminal, follow SMC-B verification.
How Targeting Works
Endpoint jobs can be assigned to tenants, folders, groups, or individual endpoints. Review the resolved targets before disruptive work. Requirements restrict which devices can run the job; a target that does not meet them is skipped.
Use endpoint concurrency to limit how many devices run at once. For per-device exceptions, disable the job on that endpoint.
Validity windows
Validity windows restrict when new runs may start. They use the server timezone and do not pause a run already in progress. Leave them empty for unrestricted start times.
Use automation timezones to schedule work and account for the server timezone when combining schedules with validity windows.
Step error behavior
Choose the response appropriate to each step:
| Behavior | Result |
|---|---|
| Stop and fail | Stops later steps in the current section and keeps the failure. |
| Continue and fail | Runs later steps in the section while keeping its failed result. |
| Continue and allow success | Allows completion with warnings if no other step blocks success. |
These behaviors apply separately to Tasks and Cleanup. Success-based follow-up automations also run for Succeeded with warnings. Use that choice only when the failed step is optional.
Cleanup
Add tasks to Cleanup to release resources or collect diagnostics after the main work ends, even when it fails, is cancelled, or times out. Cleanup is available for endpoint and infrastructure jobs, including templates.
Cleanup is skipped when a job is aborted before work requiring cleanup has run.
Cleanup can use the job's workspace and earlier task results. Use an If task to release a resource only when the corresponding setup succeeded. See Variables for conditions based on the main result.
Cleanup preserves an unsuccessful main result. Cleanup errors follow the selected step error behavior.
Allow enough time for Cleanup in the job's timeout settings. See Job Runs when you need to stop a run.
Pauses and endpoint availability
Sleep steps and timed postponement in Tasks let other jobs use the endpoint. The run resumes after the pause when the endpoint is available, so later tasks should check any device state they depend on.
The device stays occupied until Cleanup finishes, even while waiting.
Timeout
Tasks and Cleanup have separate timeouts. Set them in the job overview; the defaults are 24 hours for Tasks and 5 minutes for Cleanup. For infrastructure jobs, the Tasks timeout pauses while the job is queued waiting for Runner capacity or inputs. Endpoint job timeouts and Cleanup timeouts include waiting time.
Individual tasks can have shorter timeouts. Set these to cover the expected work; for OS deployment, allow time for restarts and reconnection. A timeout stops the section regardless of the step's error behavior, but active work can take time to stop.
Cancellation cannot undo work already accepted by Windows servicing or another system. Review a timed-out task before retrying.
Durable Result Spooling
Enable durable result spooling for supported maintenance tasks that can interrupt Agent connectivity before reporting a result. It applies to endpoint scripts, APT updates, and Windows update installation tasks.
Set its recovery timeout to cover the expected interruption. It preserves result delivery, but does not rerun failed scripts or guarantee that arbitrary Windows processes survive an Agent restart. The job's overall timeout still applies.
APT maintenance on marked Ordyn infrastructure hosts enables this protection automatically. See infrastructure roles.
Desktop execution
Interactive Windows work needs a suitable execution context. Keep desktop execution enabled for GUI replay or desktop-control tasks.
With Manage desktop preparation enabled, GUI replay prepares and cleans up its desktop automatically. Use manual desktop steps only when the workflow also provides cleanup after failure. A manual input lock needs an unlock step, and a managed logon needs a logoff step unless the session is intentionally left active. GUI replay cannot run inside loops.
User interaction
For disruptive Windows work, configure notification, confirmation, and postponement according to your maintenance policy. Explain the expected interruption in the prompt so the person can make an informed choice.
Variables and branching
Use Job Variables for saved overrides and optional launch inputs. Use Variables for inheritance, expressions, and runtime values. Branches and loops let later tasks respond to earlier results.
Files, scripts, and templates
Use Library Files and Scripts for reusable inputs. Use Job Templates when several jobs should share one maintained workflow, or Composing Jobs to coordinate independent jobs.
Exporting And Importing Jobs
Export a job definition for reuse or a complete job bundle to include portable dependencies. Import explains dependency mapping and what must be configured in the destination.
Imported jobs are standalone. Review assignments, schedules, and environment-specific values before running them.
Job runs and history
Endpoints must meet the job requirements before their work starts. An eligible endpoint can wait for earlier work to finish; if it becomes ineligible while waiting, its target is skipped with a reason. Jobs that require a connection skip disconnected targets unless Power on endpoints before running is waiting for them to connect.
Review a run's task results and artifacts to confirm its outcome. A skipped target means the work did not run; if all targets are skipped by requirements, the run fails. Use Job Runs for cross-job investigation.
Disabling jobs on one endpoint
Disable a job on an endpoint when that device must be excluded without removing the job's shared assignments. Targets queued behind earlier work are skipped; work already in progress continues. Enable it again when the exception ends.
AI-agent execution
Enable Allow AI-agent execution in a job's settings to let an AI agent run it. The agent also needs management access to run that job on its selected targets. The flag is disabled by default and is separate from self-service availability. See AI-agent management access for scoped grants and independent operation.
Self-service jobs
Enable selected jobs in the Self-Service Portal when Windows users should start approved work themselves.
Power on endpoints before running
Use Power on endpoints before running when offline targets should be started before work. Configure the power-on method and prerequisites and allow enough time for boot and Agent connection.