Unattended automation is a mode of robotic process automation in which software robots run end-to-end processes without a human present, typically on dedicated servers or virtual machines, and are started by schedules, work queues or system events rather than by a user. In RPA, it contrasts with attended automation, where a robot works on an employee’s desktop at that person’s request.
Unattended robots are managed centrally by an orchestrator, which assigns jobs, stores credentials, balances workload across machines and records every run. This mode suits high-volume, rule-based back-office processes such as invoice entry, report generation, data reconciliation and account provisioning.
How unattended automation runs without a person
An unattended robot executes on a runtime machine that nobody uses interactively. Because most RPA tools drive applications through their user interface, the machine must open a real Windows session, usually through a remote desktop or console login created by the robot service with stored credentials.
- Trigger: the orchestrator starts a job on a schedule, when items arrive in a queue, or through an API call.
- Allocation: the job is assigned to a free robot on a machine with the required applications.
- Session start: the robot logs in with a dedicated service identity and opens a desktop session.
- Execution: the workflow processes transactions through user interfaces, APIs, files and databases.
- Reporting: results, logs and screenshots of failures are sent back to the orchestrator.
- Cleanup: applications are closed and the session is logged off to leave the machine in a known state.
The defining property is independence from human timing. The process can run at night, on weekends or in parallel on many machines, which is where most of the capacity gain of RPA comes from.
Unattended, attended and hybrid automation compared
The choice between modes depends on whether a decision or input from a person is needed during the process. The table compares unattended automation with its neighbors in the RPA topic.
| Mode | Who starts it | Where it runs | Typical process |
|---|---|---|---|
| Unattended automation | Schedule, queue or event | Server or virtual machine | Bulk invoice posting overnight |
| Attended automation | Employee on demand | Employee desktop | Pre-filling forms during a customer call |
| Hybrid automation | Robot hands over to a person and back | Both | Exception approval inside a batch process |
| API-based automation | Event or schedule | Integration platform or service | System-to-system data sync |
Hybrid designs are common in mature programs. An unattended robot processes the straight-through cases, places exceptions in a human task queue and resumes once a person has resolved them.
Infrastructure, licensing and operating parameters
Running robots without supervision shifts responsibility from the user to the operations team. Machines, sessions, credentials and capacity must be planned like any other production service.
| Parameter | Typical choice | Why it matters |
|---|---|---|
| Runtime environment | Virtual machines or terminal server sessions | Isolation and scalability |
| Screen resolution | Fixed, for example 1920 by 1080 | Stable UI element detection |
| Robots per machine | One per VM, several on server editions | Cost versus interference |
| Credentials | Stored in an orchestrator or credential vault | No passwords in workflow code |
| Run window | Off-peak hours or continuous queue processing | Load on target systems |
| Licensing unit | Usually per unattended robot or runtime | Unattended licenses cost more than attended ones |
Capacity is planned from transaction volume, average handling time and service level. If each case takes about 90 seconds and 4,000 cases must be finished within an eight-hour window, the workload needs roughly 100 robot-hours, so at least 13 parallel robots are required before any allowance for retries.
Choosing processes for unattended automation
Good candidates are stable, rule-based and high in volume, with digital inputs and clear outcomes. Processes that need frequent judgment or depend on unstructured documents without reliable extraction are better handled in attended or hybrid mode.
| Situation | Recommended mode | Reason |
|---|---|---|
| High volume, fixed rules, digital input | Unattended | Full capacity gain without supervision |
| Agent needs data during a live call | Attended | Robot must react to the user in real time |
| Mostly rules with some exceptions | Hybrid | People handle only the exceptions |
| Target system offers a stable API | API integration instead of UI robot | Faster and less fragile |
| Frequent application UI changes | Defer or redesign | Maintenance cost exceeds benefit |
Risks and common mistakes
Because no one watches the screen, failures must be detected by design. A robot that stalls on an unexpected pop-up can block a queue for hours unless timeouts, alerts and recovery routines are in place.
- Shared human accounts: robots that log in as employees break audit trails and violate segregation of duties.
- Locked or expired passwords: rotation policies stop all robots at once unless managed in a vault.
- Missing recovery logic: a failed transaction should be marked, logged and skipped, not stop the entire batch.
- Screen-dependent selectors: changes in resolution, language or application version cause silent misclicks.
- Unmonitored queues: without alerts on queue age, backlogs grow unnoticed over weekends.
Governance completes the picture: every unattended process needs an owner, a runbook and an audit trail showing what each robot did, under which identity and with what result.
Understanding it simply
Imagine an office worker who comes in at night, works through a stack of forms and leaves a report on your desk by morning. Unattended automation is the software version of that night shift. A program clicks, types and copies data on a computer that nobody is sitting at.
People set the rules and check the results. They do not need to press a button each time. The robot works on its own schedule and hands over only the cases it cannot solve.
Applied in practice
The finance department of a regional utility receives about 6,000 supplier invoices a month as structured electronic files. Clerks had spent most of each morning entering them into the accounting system. The team deploys four unattended robots on virtual machines that start at 10 p.m., read new files from a shared folder, match each invoice to its purchase order and post it.
Invoices with price differences above 2 percent or missing order numbers go into a human review queue. Credentials are stored in the orchestrator, and each robot uses its own service account. Within a quarter, about 85 percent of invoices post without human touch, clerks start the day with only the exceptions, and the audit team can trace every posting to a specific robot run.