Skip to content

Flow Control and AI Employee Management Actions

These actions do not touch the applications the AI Employee is automating - they shape the job itself: jumping between actions, grouping and pacing them, reserving work so two AI Employees never collide, and capturing evidence of what happened.

Jumps to another action in the job sequence instead of continuing to the next one. Use it to build loops (jump back to an earlier action) or to skip ahead past steps that do not apply. Key setting: Go To Action - pick the target action from the job’s action list.

A container that organizes related actions under one collapsible heading in the job designer. The group itself performs nothing, but if a group’s run-when condition is not met, every action inside it is skipped - so groups double as conditional blocks: put a branch’s steps in a group and let one condition govern them all.

Pauses the job for a set number of seconds before continuing. Use it to give an application time to finish loading or processing before the next action runs. Key setting: Delay (Seconds).

Pauses the job until the clock reaches a specific time of day, then continues. Use it to hold work until a system becomes available or a business cutoff passes - for example, wait until 1800 before posting the day’s batch. Key setting: Delay Until Time (24 Hour Format) - a time from 0000 to 2359.

Reserves the process instance the job is working on exclusively for this AI Employee, so no other AI Employee picks up work on the same instance at the same time. Use it at the start of a sequence that must not run twice in parallel - for example, before entering an order into an accounting system. The action requires the job to be running against a process instance.

Releases the exclusive reservation taken by Robot - Exclusively Reserve, making the process instance available to other AI Employees again. Place it after the protected steps are complete; the reservation is also released when the job ends.

Shuts down the active AI Employee: it stops accepting new work, and someone will need to start it again before it resumes working. Most often it is used as a stop switch - put it on the branch a job takes when something has gone wrong, so the AI Employee stops rather than carrying on and doing more work on top of a problem. It also works as a controlled stop at the end of a maintenance or end-of-day sequence.

Nothing in the job runs after this action - the AI Employee stops there. Setting the action’s result on success to exit the job is still tidier, because the job then clearly ends at the step you meant it to.

Whenever an AI Employee is stopped by a step, ProcessPlan records what it was doing at that moment and adds a screenshot in front of that step for future runs. See Managing AI Employee Jobs.

Takes a screen capture and saves it to the job run’s audit log, creating a visual record of what was on screen at that moment. Use it at key checkpoints - before and after a critical entry, for example - so you can review or prove what the AI Employee saw. The capture is also available to later actions in the job. Screenshots are skipped when the job runs in test mode.

You already get some of these without adding anything. Every job run saves a screenshot when it starts and another when it ends, so a run that fails always shows you the screen it finished on. Each screenshot also carries the AI Employee’s own log for the stretch of work leading up to it, so you can read what it was doing as well as see where it got to.

Jobs built from a screen recording add a few more on their own. Wherever the job moves somewhere new - focusing a different window, opening a page or a file, starting an application - it takes a screenshot of what it arrived at, named for the step it followed. Those are the moments everything afterwards depends on, so when a later step cannot find what it expects, the screenshots show you whether the AI Employee was even looking at the right screen. You can delete any you do not want, and add your own wherever you want a closer record.

When an AI Employee stops working, it writes down everything it knew. An AI Employee never carries on into the next job after something unexpected - it stops and waits, because running more automated steps against a screen that is not what it expected is how real damage gets done. At that moment it saves a final screenshot named Failure Debrief, and selecting View Audit Log on it shows a full account rather than just the recent log: the job and how far it got, the step it stopped on with everything that step was set up to look for, how many times that step has failed in a row, what was on the machine at that instant, every step this run took in order with its result and how long it took, and the AI Employee’s whole log for the run. It is written by the AI Employee itself, so it appears whether or not the workstation can reach the internet, and anything that looks like a password or a code is removed before it is written down.

Takes a screenshot of the screen and holds it in memory as a file that later actions in the job can reference - for example, attaching it to a process field or including it with an AI file prompt. Unlike Robot - Capture Screen to Audit Log, nothing is written to the audit log; the image only exists for the rest of the job.