Attended or Unattended RPA: Where a Software Robot Should Work
Some RPA bots run on an employee's desktop and wait for a click. Others work alone on a server through the night. The choice shapes which tasks speed up, what can go wrong and who keeps watch.

Robotic process automation has nothing to do with physical robots. The term, usually shortened to RPA, describes software that operates other programs the way a person does. It clicks buttons, copies values between screens, fills in forms and reads files. Its appeal is that it works with existing systems, including old ones without a modern interface.
Discussions about RPA often skip a basic design question: where the bot works and who is present when it does. Some bots live on an employee's desktop and start when that person asks. Others run alone on a server, often at night, and work through a queue of cases without anyone watching. The industry calls these attended and unattended automation.
The difference sounds technical, but it decides which tasks get faster, how errors surface and what the setup costs. This article compares the two modes and shows how to choose between them.
How a software bot does its work
An RPA bot follows a script built in a design tool. The script describes steps such as opening an application, finding a field, reading a value, applying a rule and writing the result somewhere else. Modern tools identify screen elements by their properties rather than by pixel position. That makes bots less fragile than early screen scrapers. They still break when an application changes its layout or a pop-up appears that the script did not expect.
Bots are fast and tireless at the steps they know. They are poor at anything outside the script. That simple fact explains most of the difference between attended and unattended use.
Attended bots: help at the desk
An attended bot runs on the same computer as the employee, using that person's login. The employee starts it with a click or a hotkey, or it starts when a certain screen opens. It takes over a stretch of tedious work and hands control back when a decision is needed.
A typical case is a service desk. During a call, an agent needs the customer's contract, open invoices and delivery status from three systems. A bot can gather all three and show them in one window while the agent keeps talking. Another case is a clerk entering an order, where the bot fills the routine fields and the clerk checks the unusual ones.
The efficiency gain here is measured in seconds and minutes per case, spread across many employees. The human stays in charge, so the bot can be simpler. When something unexpected happens, the person sees it at once. The downside is that the bot's pace is tied to the person's. It cannot work through a backlog while the office is closed.
Unattended bots: working through the queue
An unattended bot runs on a server or virtual machine with its own account. A scheduler or a trigger starts it, for example when new files arrive in a folder or a batch of orders lands in a queue. It processes case after case without anyone at a screen.
This mode fits work that is high in volume, follows clear rules and arrives in a structured form. Examples include matching payments to invoices, moving records from an old system to a new one, creating standard reports or checking supplier data against a register. A single bot can work around the clock, and several can share a queue when volume rises.
The price of this independence is design effort. Since nobody watches, every likely problem has to be anticipated. What happens when a value is missing? When a system is down? When the same case appears twice? A well-built unattended bot sets such cases aside for people, logs what it did and alerts someone when its error rate climbs.
An attended bot fails in front of a person. An unattended bot fails in a log file, unless someone designed it to speak up.
Choosing the right mode for a task
A few questions usually settle the choice.
- Does the task need a human decision partway through? If so, attended automation or a hybrid is the better fit.
- Does the work arrive in batches or a queue, independent of when staff are present? That points toward unattended bots.
- How clean are the inputs? Unattended bots need predictable data, while attended bots can rely on a person to correct odd cases.
- How many people perform the task? Attended automation pays off when many employees repeat the same small steps every day.
- What does a mistake cost? The more expensive an error, the more checks an unattended bot needs before it writes anything.
Many organizations end up with a mix. In a hybrid design, an unattended bot handles the routine cases and passes the rest to an employee, who resolves them with help from an attended bot. That split often brings the largest gain, because the queue keeps moving while people focus on the exceptions.
Security, licenses and upkeep
The two modes raise different security questions. An attended bot acts with the rights of the person using it, so it can do anything that person can do. Organizations should control which bots employees may run and keep a record of their actions. An unattended bot needs its own account, with only the permissions the task requires. Its passwords belong in a protected vault, not in the script.
Licensing models often differ as well. Attended bots are commonly priced per user, unattended bots per robot or by capacity. Servers, monitoring and a support team add to the bill in both cases.
Upkeep is the cost most often underestimated. Every time a target application gets an update, a bot may need adjusting. Attended bots then break in front of many users at once, which is visible but disruptive. Unattended bots can break quietly and leave a growing backlog. Either way, someone must own each bot and test it whenever the systems around it change.
Where RPA fits in the longer run
RPA is often described as a bridge. It connects systems that lack proper interfaces and buys time until a better integration exists. That is a fair role, as long as everyone understands it. When a vendor later offers an interface for direct data exchange, the bot that copied values through screens can retire.
Newer tools add AI to RPA, for example to read unstructured documents or to understand an email before a bot acts on it. This widens the range of tasks, especially for unattended bots. It does not remove the basic choice. Someone still has to decide whether a person stays beside the bot or the bot works alone, and then design the process around that answer.
Questions and answers
What is the difference between attended and unattended RPA?
Attended bots run on an employee's computer and start when that person triggers them. Unattended bots run on servers, start on a schedule or an event and work without anyone present. Many organizations use both, sometimes within the same process.
Which mode saves more time?
It depends on the task. Unattended bots clear large queues of routine cases and can run around the clock. Attended bots save smaller amounts per case, but across many employees those minutes add up. The best results often come from combining the two.
Can an attended bot become an unattended one later?
Sometimes, but rarely without rework. An attended bot can rely on a person to handle surprises, so it often lacks the error handling, logging and alerts an unattended bot needs. Plan for that work if a task is likely to move to unattended mode.
Do RPA bots need their own user accounts?
Unattended bots should have them, with only the permissions the task requires and passwords kept in a secure vault. Attended bots usually act under the employee's own login. In both cases, the bot's actions should be logged so that changes can be traced.
Does RPA replace employees?
Unattended bots can take over entire routine tasks, which changes how a team spends its time. Attended bots support people rather than replace them. In practice, staff often move toward handling exceptions, checking results and work that involves customers or suppliers.


