What Is Robotic Process Automation (RPA)?
RPA uses software bots to click and type through your existing applications the way a person would. Here's how it works, where it fits and where it doesn't.
By Matthew Clarkson · June 14, 2024 · Updated November 17, 2026

Robotic process automation (RPA) is software that operates your other software. A bot logs in, clicks, types, copies and pastes across the same screens a person uses, following a fixed set of steps. There's no physical robot involved. It's closer to a macro that can work across several applications at once.
We build automations for Australian businesses, and RPA is one of the tools we use. It's useful, with a narrower range of good uses than the vendor marketing suggests. This guide covers what it is, how it works, where it fits and where it doesn't, and how it relates to business process automation and AI agents.
What RPA actually does
An RPA bot works on the user interface: the windows, buttons and text fields you see on screen. It doesn't need the application to have an API or any other way in for developers. If a person can do the task with a keyboard and mouse, a bot can usually be built to do it too.
A typical job looks like this. Every morning a staff member logs into a supplier portal, downloads a report, opens the spreadsheet and keys each row into the accounting system. A bot can do the same thing on a schedule: open the browser, log in, download the file, read each row and enter it into the right fields. It runs the same steps the same way each time, at any hour.
That consistency is the point. A bot doesn't get tired or skip a row on a Friday afternoon. It also doesn't think. It follows its script exactly, so it will repeat a wrong step as faithfully as a right one.
How it works
Most RPA platforms have three parts.
- Studio. The design tool where a developer builds the workflow, usually by dragging steps such as "open application", "click" and "type into" into order. Some tools can record a person doing the task and turn the recording into a starting workflow, though recordings nearly always need cleaning up.
- Bot (or runner). The software that carries out the workflow on a computer or virtual machine.
- Orchestrator. The control centre where you schedule bots, watch them run, deal with failures and read the logs.
Bots run in one of two modes. An attended bot runs on a staff member's desktop and is started by them, for example to fill in a form while they're on a call. An unattended bot runs on its own on a server, on a schedule or when a file lands in a folder.
Every action a bot takes is logged. That gives you an audit trail of what was done and when, which helps with compliance and with working out what went wrong when a run fails.
The common platforms are UiPath, Automation Anywhere, SS&C Blue Prism and Microsoft Power Automate (its desktop flows). We also build RPA-style automation in n8n, calling a headless browser for the screens that have no API. There's more on that in our guide to RPA with n8n.
Where RPA fits
RPA suits work that is:
- Rules-based. You could write the steps as a flowchart with clear "if this, then that" decisions.
- Repetitive and high volume. The same task done many times a day or week, so the time saved adds up.
- Built on structured digital data. Spreadsheets, form fields, database records and reports in a fixed format.
- Stable. The process and the screens it touches don't change often.
It's strongest where there's no better way in. Older desktop applications, government and supplier portals, and systems whose vendor charges heavily for API access are all good candidates. In those cases RPA can connect systems that would otherwise need someone to re-key data between them.
Common examples we see:
- Accounts payable: reading supplier invoices from an inbox, matching them to purchase orders and entering them for payment.
- HR: updating staff details across payroll and rostering systems when someone changes role.
- Logistics: checking carrier portals for shipment status and updating the order system.
- IT support: routine account changes such as password resets.
Where it doesn't fit
RPA struggles with anything that needs judgement. If a step depends on reading a messy email and deciding what the sender wants, a standard bot can't do it.
It also breaks when screens change. If a vendor moves a button or adds a login step, the bot stops at the step it can no longer find. Bots need ongoing maintenance, and a process that touches several frequently updated web apps can cost more to maintain than it saves.
A few more cases where we'd steer away from RPA:
- The system has a decent API. An API integration is faster and far less fragile than a bot clicking through screens. Use it.
- The process has lots of exceptions. If people step in on every third item, the bot won't save much.
- The process is broken. Automating a messy process gives you the same mess, faster. Fix the process first.
- The inputs are on paper. Bots need digital inputs, so scanning and extraction have to come first.
RPA, BPA and AI agents
These three terms often get used as if they mean the same thing. They're different.
Business process automation (BPA) is the broader practice of automating a whole process end to end: an order coming in, stock being checked, an invoice going out. It usually connects systems through APIs and a workflow engine. RPA is one technique inside BPA, used for the steps where the only way into a system is through its screens.
An AI agent uses a language model to decide what to do next, calls tools to act, looks at the result and repeats until the job is done. RPA follows a fixed script, while an agent chooses its own path. That makes agents useful for unstructured work, such as reading an email and working out what the customer wants, and less predictable than a bot.
In practice they're often combined. An AI step reads the incoming document and pulls out the fields, then a workflow or bot enters them into the system. Use the simplest option that does the job. A fixed script is cheaper to run and easier to test than an agent.
How to pick your first process
Start small. The first automation should be low-risk, easy to measure and something the team wants gone.
- Ask the people who do the work. They know which tasks are tedious and where the exceptions hide.
- Check it against the list above. Rules-based, repetitive, structured and stable. If it needs judgement or has frequent exceptions, pick something else.
- Check for an API first. If the systems involved have one, a normal integration is likely the better build.
- Write the steps down. If you can't map the process as a flowchart, it isn't ready to automate.
- Measure the baseline. Record how long the task takes and how often it goes wrong now, so you can show what changed.
- Plan for maintenance. Decide who watches the bot and who fixes it when a screen changes.
A small, well-chosen first project proves the approach and tells you what the second one should be. If you want a hand finding it, our business process audit maps how the work actually happens and ranks what's worth automating.
Frequently asked questions
Is RPA the same as AI?
No. RPA follows a fixed set of steps and can't make judgement calls. AI, including the language models behind AI agents, can read unstructured text and make decisions. The two are often combined, with AI doing the reading and deciding and RPA doing the clicking and typing.
Will RPA replace jobs?
It usually replaces tasks. The work that goes is re-keying data between systems and running the same report every week. Staff then spend that time on the parts of the job that need a person, such as exceptions, customers and judgement calls. Be open with the team about what's being automated and why, and involve them early. They know the process better than anyone.
How much maintenance do RPA bots need?
More than most people expect. A bot depends on the screens it uses, so an update to any of those applications can break it. Someone needs to monitor runs, read failure logs and adjust the bot when things change. Budget for that from the start.
Do I need an RPA platform to do RPA?
Not always. The large platforms suit organisations running many bots with a dedicated team. For smaller workloads, an integration tool such as n8n, with browser automation for the few screens that have no API, is often cheaper and easier to maintain.
If you're weighing up RPA for a specific process, see our robotic process automation service or book a 15-minute call.
Last updated on November 17, 2026
Free tool · 2 minutes
Automation ROI Calculator
See what a manual process is really costing you, and the payback on automating it.
Continue Reading
View all
Robotic Process Automation Solutions for a More Efficient Business
Imagine you have a new team member who loves doing all the boring, repetitive tasks you hate. They can copy-paste data, fill out forms, and run reports all day long. The best part? They work 24/7, never need a coffee break, and never make a single typo. That’s pretty much what robotic process automation is… […]
Dec 13, 2025

RPA for Finance: Streamline Financial Operations in 2025
Discover how RPA for finance transforms operations. This guide explores use cases, ROI, and implementation steps for Australian financial teams.
Jul 25, 2025

Robotic Process Automation for Managed IT Services
Explore how Robotic Process Automation (RPA) is transforming managed IT services, improving efficiency, reducing errors, and enhancing customer satisfaction.
Sep 4, 2024
Sick of reading about automation?
Book a free 15-minute intro call. We’ll talk through what you’re trying to automate and whether we’re a good fit.