In this article
RPA basics
What is RPA?
Robotic process automation is software that carries out a business task by working through the user interface (UI) of other applications: it clicks, types, copies and pastes by fixed rules. The "robot" is a program, not hardware. It repeats the same steps on every item, so it suits stable, rule-based, high-volume work such as moving data between two systems that have no connection to each other.
Why it is asked: it shows whether you can define the term without vendor jargon. Related: the coding interview guide covers the wider technical-interview context.
Is RPA the same as artificial intelligence?
No. A basic RPA bot is deterministic: given the same input and screen, it follows the same rules and does the same thing. Artificial intelligence (AI) and machine learning (ML) learn patterns from data and can give different answers. The two combine well, for example when a model reads a scanned document and a bot enters the result, but the bot itself is not intelligent.
Why it is asked: it separates people who know what the tool does from people repeating marketing.
When should you use an API instead of RPA?
Use an application programming interface (API) when a stable one exists. An API call talks to the system directly, so it is faster and breaks less often than a bot that depends on how a screen looks. RPA earns its place when there is no API, when building one is too costly, or when the work must go through the user interface for access or audit reasons.
Why it is asked: it tests judgement about brittleness, not enthusiasm for the tool. Related: testing roles also drive user interfaces and weigh the same brittleness.
Which processes are good candidates for RPA?
Good candidates are rule-based, repetitive, high-volume and done on structured digital data, with few exceptions and stable applications. Poor candidates need judgement, change their screens often or handle mostly unstructured input such as free text. A strong answer also mentions checking the volume against the effort to build and maintain the bot.
Why it is asked: choosing the right process matters as much as building it. Related: Java interview questions for automation testing covers the same "what is worth automating" judgement for test code.
What are the benefits and limitations of RPA?
Benefits: it automates repetitive, rule-based work without changing the underlying applications, runs the same way every time and leaves a log of what it did. Limitations: it breaks when screens change, handles unstructured data poorly without extra tools, and adds maintenance work, so an API is better where one exists. A balanced answer names both sides and a case where you would not use RPA.
Why it is asked: it shows whether you see the tool's limits as well as its appeal.
Robots and the platform
What are the main components of an RPA platform?
An RPA platform usually has three parts: a design tool where developers build workflows, a robot that runs them, and a control layer that schedules, monitors and stores shared resources. In UiPath these are Studio, the Robot and Orchestrator; other products name them differently. Knowing which part does what lets you say where a problem sits.
Why it is asked: it is a warm-up that shows you have seen a platform end to end. Related: UiPath Orchestrator guide.
What is the difference between attended and unattended robots?
An attended robot works alongside a person on the same machine and is started by the user or a user event; UiPath's Robot admin guide describes it as the personal assistant of an end user (UiPath Robot admin guide, version 2023.10, October 2026). An unattended robot is a robot that runs without a user to start it, and UiPath calls Orchestrator the central hub for unattended automation (UiPath Orchestrator guide, October 2026). It usually runs on a dedicated machine.
Why it is asked: it checks that you can match a robot mode to a use case. Read the full explanation: UiPath Robot admin guide.
What is Orchestrator, and what does it do?
Orchestrator is UiPath's central server for managing robots. According to UiPath's documentation it launches unattended runs on the spot or through triggers, manages the resources that automation projects use, and can distribute workload to unattended robots to use them efficiently (UiPath Orchestrator guide, October 2026). Other RPA products have an equivalent control layer under another name.
Why it is asked: it shows you know how a bot is run at scale, not only built. Read the full explanation: UiPath Orchestrator guide.
What is a queue, and why use one?
A queue is a shared list of work items that robots take one at a time. It lets several robots work from one pool without handling the same item twice, keeps a status for each item, and lets a failed item be retried or reviewed on its own. A job that loops over a spreadsheet gives none of that.
Why it is asked: queues separate a script from a production process.
Knowing it is half the job. Saying it is the other half.
Interview Clozo is an AI interview copilot for real interview rounds: it hears the question and puts a strong answer on your screen, and you say it in your own words. See the features or how it works.
Start Free Trial →Building a reliable bot
What is a selector?
A selector is a text description of a screen element, built from its attributes such as name, type and parent, which the robot uses to find the element again at run time. A good selector uses stable attributes. A fragile one depends on position, a generated identifier or a window title that changes, so it stops matching when the application is updated.
Why it is asked: selectors are a common cause of bot failures, so interviewers probe them. Related: JavaScript interview questions for the document-structure ideas web selectors rely on.
How do you make a selector less fragile?
Prefer stable attributes, such as a fixed name or identifier, over index numbers and screen position. Replace the changing part with a wildcard or a variable when only one section varies. Check the selector in the tool that highlights the element, and re-test after an application update. When an element has no stable attribute, ask whether another approach, such as an API, is available.
Why it is asked: it separates someone who has maintained a bot from someone who only recorded one. Related: the full stack developer questions explain how a web page's structure changes.
What is the Robotic Enterprise Framework and what are its states?
The Robotic Enterprise Framework is a UiPath project template for transactional processes. Its state machine has four states: Init, Get Transaction Data, Process Transaction and End Process (UiPath Community blog, October 2026); the template names the first state Initialization, and Init is the shorthand used here. Init reads the configuration and starts the applications; End Process closes down. After a success or a business exception the framework returns to Get Transaction Data for the next item. A system exception within the retry limit returns to Init, otherwise to End Process. When no items are left, Get Transaction Data leads to End Process.
Text description of the diagram
- The Robotic Enterprise Framework state machine has four states: Init, Get Transaction Data, Process Transaction and End Process.
- Init reads the configuration and starts the applications.
- After a success or a business exception the framework returns to Get Transaction Data for the next item.
- A system exception within the retry limit returns to Init, otherwise to End Process.
Why it is asked: it is the standard structure for production UiPath work.
What is the difference between a business exception and a system exception?
A business rule exception means the data broke a rule, for example an invoice with no number, so retrying cannot help; the item is marked and the bot moves on. A system exception means something technical failed, such as a timeout or a closed application, so a retry after restarting the application can succeed. In the template, a system error within the retry limit sends the bot back to Init, and a business rule exception is handled and the loop continues (UiPath Community blog, October 2026).
Why it is asked: it tests whether you know which failures deserve a retry.
How do you handle errors in a workflow?
Wrap risky steps in a try and catch, log what failed and which item it was, and decide per error type whether to retry, skip or stop. Take a screenshot on failure so the problem can be seen later. Keep business rules separate from technical failures, and never swallow an error silently, because a bot that fails quietly corrupts downstream data.
Why it is asked: reliability is most of the job. Related: Java interview questions cover try and catch design in a general-purpose language.
How should a bot handle credentials?
Keep passwords and keys out of the workflow file and out of plain-text configuration. Store them in a secure credential store managed by the platform, give the bot's account only the access the task needs, and rotate them under the organisation's policy. A bot account is a real account, so treat it with the same care as a human one.
Why it is asked: bots touch sensitive systems, so security habits count.
How does a bot work with virtual desktops or Citrix?
On a virtual desktop or Citrix session the robot often sees only a streamed picture of the screen, not the elements behind it, so ordinary selectors cannot find anything. Automation then relies on image matching or optical character recognition (OCR) and on screen coordinates. That is slower and more fragile, because a changed resolution, theme or scaling can break it, so prefer an API or a native connection when one exists.
Why it is asked: it tests whether you know where UI automation is weakest.
Running RPA in production
How do you test and debug an RPA bot?
Run the workflow step by step on realistic data, watch variable values, and add log messages at the start and end of each stage. Test the normal path, a business exception, a system failure and an empty queue. Run it in a test environment that matches production, since a screen that differs from production can hide a failing selector.
Why it is asked: it shows you test failure paths, not only the happy path. Related: testing interview questions explain test design for the same cases.
What happens when the application's screen changes?
The selectors stop matching, and the bot raises a system exception or fails to find an element. Detect it through logs and alerts, fix the selector, re-test and redeploy. To reduce the cost, build with stable attributes, keep selectors in one reusable place and agree with the application owner on how changes are announced.
Why it is asked: maintenance cost decides whether automation pays off.
How do you decide whether a process is worth automating?
Compare the time saved per item multiplied by volume with the effort to build, run and maintain the bot. Check that the process is stable, documented and rule-based, and that the data is structured. Ask what happens when it fails. A short, frequent task often pays back faster than a long, rare one, but estimate with your own numbers, not general figures.
Why it is asked: it shows business sense beyond tooling.
What does a good handover of an RPA project include?
A process definition document, the workflow with readable names and comments, a configuration file with no secrets, a list of the applications and accounts used, test cases and results, and a runbook that says how to restart the bot and what each error means. Someone who did not build the bot should be able to operate it.
Why it is asked: handover quality shows production experience. Related: the UiPath Orchestrator guide covers the operating side.
Follow-up questions
What are the common RPA interview questions?
Expect the definition, attended against unattended robots, selectors, queues, the Robotic Enterprise Framework, and business against system exceptions, all above. For the general method of answering a technical question aloud, see the coding interview guide.
What RPA interview questions should an experienced developer prepare for?
Prepare to go deeper on the same topics: how you design for retries, how you keep selectors stable across application updates, how you split work across several robots, and a failure you diagnosed. A strong answer states the trade-off, not only the definition.
Do I need to code for an RPA interview?
Many RPA roles centre on a visual workflow designer, but you should be comfortable with variables, conditions, loops, data tables and error handling. UiPath projects use Visual Basic or C# for expressions, and some roles expect a scripting language as well. Check the job description, and practise explaining logic aloud.
Is RPA the same as test automation?
No. Test automation checks that software behaves correctly and reports pass or fail; RPA performs real business work on live systems. The techniques overlap, since both drive a user interface, which is why the testing interview questions help.
- RPA is deterministic software that operates applications through their user interface; it is not AI.
- Use an API when a stable one exists; use RPA when it does not.
- Attended robots help a person; unattended robots run under Orchestrator.
- Queues and the Robotic Enterprise Framework make a bot process one item at a time and recover from failure.
- Retry system exceptions; do not retry business rule exceptions.
Sources
- Attended vs unattended robots (UiPath Robot admin guide) — UiPath documentation (read 26 October 2026), 26 Oct 2026
- About robots (UiPath Orchestrator user guide) — UiPath documentation (read 26 October 2026), 26 Oct 2026
- How to Use Enhanced REFramework to Implement UiPath Projects — UiPath Community blog (read 26 October 2026), 26 Oct 2026
