Robotic process automation (RPA) is software that uses “bots” to carry out repetitive, rule-based tasks in computer applications the way a person would: clicking buttons, entering data, copying information from one system to another, generating reports or moving files. The bots follow defined rules and recorded steps rather than making judgment calls. RPA is most often used in back-office work such as finance, HR and customer administration, especially where systems can’t easily be connected any other way.
At a glance
- RPA bots mimic the clicks and keystrokes a person makes in applications, following set rules.
- It suits high-volume, repetitive tasks with clear rules and structured inputs.
- It can automate work in older applications that have no API, which is a common reason to choose it.
- Bots can break when applications or screens change, so they need ongoing maintenance and monitoring.
- RPA is often combined with AI, such as document processing, to handle less structured work.
What problem it solves
Many organizations have people spending hours on work that is entirely predictable: copying order details from email into an ERP, checking invoices against purchase orders, updating customer records in three systems, or pulling the same report every morning. The work is slow, tedious and prone to typing errors, and it grows with the business.
The ideal fix would be to integrate the systems directly, but that isn’t always possible. Older applications may have no API, third-party portals may only offer a web screen, and integration projects can take months. RPA works through the same screens a person uses, so it can automate the task without changing the underlying systems. Done well, it frees staff for work that needs judgment and makes routine processes faster and more consistent.
How it works
Choosing the process. Teams start by identifying tasks that are rule-based, frequent and stable. Some platforms include process or task mining tools that analyze system logs or desktop activity to find candidates.
Building the bot. Most RPA platforms provide a visual designer where steps are recorded or assembled from drag-and-drop actions: open this application, read this field, paste it there, click submit. Business users can build simple bots on some platforms; complex ones usually need trained developers.
Running the bot. Attended bots run on an employee’s computer and are started by that person. Unattended bots run on their own on servers or virtual machines, on a schedule or triggered by events. A central console schedules bots, stores credentials, logs activity and reports exceptions. Robotic desktop automation (RDA) is a term sometimes used for the attended, desktop-level form.
Handling exceptions. When a bot meets something it wasn’t built for, such as a missing field or an unexpected pop-up, it should stop and hand the case to a person rather than guess.
Adding intelligence. Many platforms now add intelligent document processing to read invoices and forms, machine learning to classify requests, or AI agents to handle less predictable steps. The rule-following bot remains the part that does the clicking.
Security. Each bot needs accounts and permissions in the systems it uses. These non-human identities should get only the access they need, with credentials stored centrally and activity logged, like any other privileged account.
When it matters for buyers
- When the systems involved have no API. RPA is often the quickest way to automate across legacy or third-party applications.
- When comparing automation with outsourcing. RPA keeps the work in-house; business process outsourcing (BPO) hands it to a provider, and some BPO providers use RPA themselves.
- When estimating cost. Licenses, development, infrastructure and ongoing bot maintenance all count. Savings depend on how stable the process and applications are.
- When applications are about to change. A system upgrade or migration can break existing bots, so include them in change planning.
Our Business Process Outsourcing guide covers when to automate in-house and when to hand a process to a provider.
Questions to ask vendors
- How is the platform licensed: per bot, per attended user, by run time or by process?
- Which applications and environments are supported, including virtual desktops and older systems?
- How do bots handle exceptions, and how are failures reported?
- How are bot credentials stored and rotated, and what audit logs are kept?
- What happens to our bots when a target application’s screens change, and what tools help find and fix breaks?
- Which AI or document-processing features are included, and which cost extra?
- Who builds and maintains bots: our staff, your services team or a partner?
How it differs from hyperautomation
RPA is a specific technology: bots that follow rules to carry out steps in applications. Hyperautomation is a broader approach that uses RPA as one tool among several, adding process discovery, document processing, AI, integration and governance to automate whole processes. A company can use RPA on its own for a handful of tasks; a hyperautomation program usually includes RPA or similar task automation as one part.
Problems this helps with
- Staff retyping the same data between systems all day
- Old applications with no API to integrate with
- Back-office work that grows faster than headcount
- Bots that break every time a screen changes
