I85PREP
⌘ K
AA Advanced RPA study materials
AA Advanced RPA study resources

AA Advanced RPA practice questions

One question at a time. Build your AA Advanced RPA understanding.

First 50 of 200 questions. Each one lists the choices, the correct answer, and an explanation. The checker at the top of this page saves progress in this browser when it is available.

Continue in the practice tool above to load more questions, 50 at a time.

Question 1Automation and AI fundamentals

Which component of an RPA platform executes the individual steps that a developer has designed in a workflow?

  1. The orchestration queue
  2. The bot runtime agent Correct answer.
  3. The process mining engine
  4. The credential vault

Explanation

The runtime agent is the software component installed on a machine that interprets and performs the workflow steps. Other platform services support design, scheduling, or security but do not execute steps themselves.

Question 2Automation and AI fundamentals

In RPA terminology, what does an 'attended' bot primarily do?

  1. Runs only inside a virtual desktop infrastructure
  2. Runs on a schedule without any human present
  3. Performs only API-based integrations
  4. Executes alongside a human user who triggers or supervises it Correct answer.

Explanation

Attended bots are designed to work in real time with a human operator, often triggered from the user's desktop. Unattended bots run without a person present, typically on a schedule or queue trigger.

Question 3Automation and AI fundamentals

Which AI capability allows an RPA bot to extract the vendor name and total amount from a scanned PDF invoice?

  1. Optical character recognition combined with document understanding Correct answer.
  2. Reinforcement learning
  3. Speech synthesis
  4. Sentiment analysis

Explanation

Document understanding pipelines use OCR to convert image text into machine-readable characters and ML models to classify and extract fields. Sentiment, speech synthesis, and reinforcement learning address unrelated problem types.

Question 4Automation and AI fundamentals

What is the primary purpose of a queue in an RPA orchestration platform?

  1. To render the developer's workflow canvas
  2. To store the bot's compiled source code
  3. To distribute work items across bots and track their status Correct answer.
  4. To encrypt credentials used at runtime

Explanation

Queues hold discrete work items so multiple bots can process them in parallel while status and retries are tracked centrally. They are a workload-distribution mechanism, not a code repository or security store.

Question 5Automation and AI fundamentals

Which statement best describes the relationship between RPA and artificial intelligence in a modern automation program?

  1. They are identical technologies marketed under different names
  2. RPA replaces AI because rules cover all business cases
  3. RPA handles deterministic steps while AI adds judgment for unstructured inputs Correct answer.
  4. AI replaces RPA because models can act on any interface

Explanation

RPA is strongest for deterministic, rule-based interactions with systems, while AI models add classification, extraction, and prediction for unstructured data. Combining them lets a workflow handle both structured and ambiguous inputs.

Question 6Automation and AI fundamentals

A developer needs a bot to read a value from a desktop application whose interface exposes no API. Which approach is most appropriate?

  1. UI automation of the application's controls Correct answer.
  2. Manual rekeying by the end user
  3. Waiting for the vendor to publish an API
  4. Direct SQL query against the application database

Explanation

UI automation interacts with the application's presentation layer through its accessible controls, which is the standard RPA technique when no API exists. Querying the database directly bypasses business logic and may violate policy.

Question 7Automation and AI fundamentals

Which metric best indicates how much human effort an unattended bot has removed from a process?

  1. Lines of code in the automation package
  2. Hours of manual work eliminated per period Correct answer.
  3. Number of workflow activities in the bot
  4. Count of developers on the automation team

Explanation

Effort displacement is measured in time saved, such as manual hours eliminated over a period. Activity counts and code size describe the solution's construction, not the value delivered.

Question 8Automation and AI fundamentals

What does a digital worker typically refer to in an enterprise automation context?

  1. A software bot that performs business tasks like a human employee Correct answer.
  2. A dashboard that visualizes process metrics
  3. A chatbot limited to answering FAQs
  4. A physical robot on a factory floor

Explanation

Digital worker is an enterprise term for a software bot that is managed, scheduled, and measured much like a human worker performing business tasks. It is not hardware, a narrow FAQ chatbot, or a reporting dashboard.

Question 9Automation and AI fundamentals

A bot must select the correct option from a dropdown whose position shifts between sessions. Which selector strategy is most robust?

  1. Fixed pixel offsets from the window edge
  2. Absolute screen coordinates
  3. The order of controls in the tab sequence
  4. Accessibility attributes such as control name and ID Correct answer.

Explanation

Accessibility attributes identify a control by its semantic identity, so the bot still finds it when layout shifts. Coordinate-based and positional strategies break whenever the interface changes.

Question 10Automation and AI fundamentals

What distinguishes a supervised machine learning model from an unsupervised one?

  1. Unsupervised models cannot process text
  2. Supervised models learn from labeled examples Correct answer.
  3. Supervised models require no training data
  4. Unsupervised models always achieve higher accuracy

Explanation

Supervised learning trains on inputs paired with known correct outputs, while unsupervised learning finds structure in unlabeled data. Accuracy depends on the problem, not the learning paradigm alone.

Question 11Automation and AI fundamentals

An automation processes invoices and routes any item whose confidence score falls below a threshold to a human reviewer. What is this pattern called?

  1. Full straight-through processing
  2. Human-in-the-loop exception handling Correct answer.
  3. Load balancing
  4. Batch reconciliation

Explanation

Routing low-confidence cases to a person keeps a human in control of ambiguous decisions while the bot handles clear cases. Straight-through processing would push everything without review, and the other options describe unrelated techniques.

Question 12Automation and AI fundamentals

Why is it important for an RPA bot to use a dedicated service account rather than a personal user login?

  1. It allows the bot to bypass audit logging
  2. It keeps credentials stable and attributable when staff change Correct answer.
  3. It makes the workflow run faster
  4. It reduces the license cost of the target application

Explanation

A dedicated identity survives personnel changes and makes bot activity clearly attributable in audit logs. Bypassing logging is a control failure, and identity choice does not determine runtime speed.

Question 13Automation and AI fundamentals

What is the role of a reference in a workflow design tool?

  1. To schedule the bot on a calendar
  2. To compile the workflow into an executable
  3. To store a reusable pointer to an application window or UI element Correct answer.
  4. To encrypt data at rest

Explanation

References capture reusable pointers to UI elements or applications so multiple actions can target them consistently. Compilation, scheduling, and encryption are handled by other platform components.

Question 14Automation and AI fundamentals

A process has high transaction volume but many undocumented variations and frequent system changes. What is the most prudent assessment before automating it?

  1. Standardize and document the process first, then automate stable paths Correct answer.
  2. Skip automation entirely
  3. Automate immediately because volume guarantees return
  4. Automate only the variation-heavy paths

Explanation

Automating an unstable, undocumented process bakes in rework because the bot must be rebuilt with each change. Stabilizing and documenting the process first yields automatable, maintainable paths.

Question 15Automation and AI fundamentals

Which technique best handles a web page that loads its data asynchronously after the initial page render?

  1. Increasing the browser zoom level
  2. Disabling JavaScript in the browser
  3. Polling for the target element until it appears or a timeout elapses Correct answer.
  4. Fixed sleep of ten seconds before every action

Explanation

Waiting dynamically for the element to become available adapts to variable load times while bounding the wait with a timeout. Fixed sleeps are either too short or waste time, and disabling JavaScript breaks the page.

Question 16Automation and AI fundamentals

In a natural language processing pipeline, what is tokenization?

  1. Encrypting text before transmission
  2. Ranking documents by relevance
  3. Translating text between languages
  4. Splitting text into smaller units such as words or subwords Correct answer.

Explanation

Tokenization divides raw text into the units a model processes, such as words or subword pieces. Encryption, translation, and ranking are separate operations performed elsewhere in a pipeline.

Question 17Automation and AI fundamentals

A bot must log into a system whose password rotates every 30 days. Where should the password be stored?

  1. In a plain text file on the bot machine
  2. In the platform's credential vault Correct answer.
  3. In the process documentation wiki
  4. Hard-coded in the workflow

Explanation

A credential vault stores secrets centrally and can be updated without editing workflows, keeping them out of code and documents. Hard-coding or plain files expose secrets to anyone with repository or filesystem access.

Question 18Automation and AI fundamentals

What is the main advantage of using an API integration over UI automation when both are available?

  1. APIs eliminate the need for error handling
  2. APIs require no authentication
  3. APIs are always faster to build initially
  4. APIs are less affected by changes to the user interface Correct answer.

Explanation

API integrations depend on a stable interface contract rather than screen layout, so UI redesigns do not break them. They still require authentication and error handling and may take longer to build than a simple screen interaction.

Question 19Automation and AI fundamentals

A bot reads a spreadsheet, and one row contains a blank required field. What should the workflow do to follow robust exception-handling practice?

  1. Log the exception, route the row for review, and continue processing other rows Correct answer.
  2. Stop the entire process permanently
  3. Fill the blank with an arbitrary placeholder value
  4. Ignore the row silently and continue

Explanation

Recording the exception and isolating the problematic item preserves throughput while keeping the failure visible for resolution. Silent skipping hides errors, and fabricating data corrupts downstream results.

Question 20Automation and AI fundamentals

Why should a bot's workflow be decomposed into reusable components?

  1. It guarantees the bot will never fail
  2. It reduces duplication and simplifies maintenance and testing Correct answer.
  3. It removes the need for version control
  4. It eliminates the need for logging

Explanation

Reusable components centralize logic so fixes and tests apply everywhere they are used, reducing duplication. Modularity does not guarantee failure-free execution and does not replace version control or logging.

Question 21Automation and AI fundamentals

An organization wants to measure bot performance in production. Which combination best supports operational monitoring?

  1. Developer satisfaction survey scores
  2. Transaction success rate, average handling time, and exception counts Correct answer.
  3. Count of installed browsers on bot machines
  4. Number of design documents and meeting minutes

Explanation

Operational health is reflected in how often transactions succeed, how long they take, and how many exceptions occur. The other options measure documentation, morale, or environment inventory rather than runtime performance.

Question 22Automation and AI fundamentals

A model used to classify support tickets was trained on data from one region and performs poorly in another region with different language patterns. What is the most likely cause?

  1. The model has too few parameters to run
  2. The tickets are too short to classify
  3. The training data does not represent the target population Correct answer.
  4. The inference server is underpowered

Explanation

Performance degrades when the training distribution differs from the production distribution, a data representativeness problem. Model size, server capacity, and ticket length are separate concerns that do not explain a region-specific drop.

Question 23Automation and AI fundamentals

In a hybrid workflow, an ML model returns a probability score for each document class. How should the automation use that score to balance accuracy and throughput?

  1. Always send every document to a human regardless of score
  2. Treat every score as equally reliable and skip review
  3. Apply a confidence threshold so low-confidence documents are reviewed by a human Correct answer.
  4. Discard any document whose score is below one hundred percent

Explanation

A threshold routes only uncertain cases to review, preserving automation throughput while protecting accuracy on ambiguous inputs. Ignoring confidence risks errors, and reviewing everything forfeits the automation benefit.

Question 24Automation and AI fundamentals

A team observes that a bot works in development but fails in production because a target application runs at a different display scaling. Which design principle was most likely violated?

  1. Idempotent transaction design
  2. Environment independence in UI interaction Correct answer.
  3. Least-privilege access
  4. Separation of concerns

Explanation

UI interactions that depend on display scaling are environment-dependent and break when the runtime environment differs. Designing selectors and interactions to be resolution and scaling independent prevents this class of failure.

Question 25Automation and AI fundamentals

When would reinforcement learning be a poor fit compared with supervised learning for an automation task?

  1. When the agent can safely explore during training
  2. When the environment provides a clear reward signal
  3. When the task requires sequential decision making
  4. When abundant labeled examples of the correct outcome already exist Correct answer.

Explanation

If labeled examples of correct outcomes are plentiful, supervised learning is simpler and more sample-efficient than reinforcement learning. Reinforcement learning is better suited to sequential decisions with reward signals and safe exploration.

Question 26Automation and AI fundamentals

Which statement best describes robotic process automation as a technology category?

  1. It requires changes to the backend database schema before deployment.
  2. It rewrites the target application's source code to expose new functions.
  3. It mimics user actions at the interface layer to execute repetitive tasks. Correct answer.
  4. It replaces the business rules engine with a trained neural network.

Explanation

RPA operates at the presentation layer, driving applications the way a person would through clicks, keystrokes, and UI element reads. This lets it automate without altering underlying systems.

Question 27Automation and AI fundamentals

What is the primary purpose of an orchestrator in an enterprise RPA platform?

  1. To schedule, dispatch, and monitor bots across machines from a central point Correct answer.
  2. To compile automation scripts into native machine code for faster execution
  3. To provide OCR capability for scanned document processing
  4. To host the relational database used by the target business application

Explanation

An orchestrator is the control plane: it queues work, assigns it to available robots, and tracks status and logs centrally. Execution of individual steps still happens on the robot machine.

Question 28Automation and AI fundamentals

Which capability most clearly distinguishes intelligent document processing from plain OCR?

  1. It classifies document types and extracts structured fields with confidence scores. Correct answer.
  2. It extracts text while preserving the original font and layout.
  3. It converts scanned pages into images at a higher resolution.
  4. It compresses documents so they transfer faster between servers.

Explanation

IDP layers classification and field-level extraction on top of OCR, returning structured data plus confidence so downstream logic can act on it. OCR alone only produces raw text.

Question 29Automation and AI fundamentals

In a supervised classification task, what does the training label represent?

  1. The feature vector the model will receive at inference time
  2. The known correct output associated with a training example Correct answer.
  3. The learning rate applied during gradient descent
  4. The number of layers configured in the network

Explanation

A label is the ground-truth target paired with an input example, which the model learns to predict. Features, hyperparameters, and architecture are separate concepts.

Question 30Automation and AI fundamentals

What does an unattended robot typically do when it finishes its assigned queue items?

  1. It becomes idle and waits for the orchestrator to assign more work. Correct answer.
  2. It escalates every completed item to a human for confirmation.
  3. It shuts down the orchestrator so no further jobs can be scheduled.
  4. It keeps the interactive desktop session open for the next employee.

Explanation

Unattended robots run without a human at the console and are driven by the orchestrator's queue; when work is exhausted they idle until new items arrive.

Question 31Automation and AI fundamentals

Which statement correctly describes the relationship between a process and a task in automation design?

  1. A process is executed only by humans, while a task is executed only by bots.
  2. A process is a single UI action, and a task is the application it targets.
  3. A process and a task are synonyms in every automation methodology.
  4. A process is the end-to-end business flow, and tasks are the discrete steps within it. Correct answer.

Explanation

Process decomposition breaks an end-to-end business flow into discrete tasks that can each be automated or assigned to a person. This separation supports selective automation.

Question 32Automation and AI fundamentals

What is the main reason an automation team builds a library of reusable components?

  1. It removes the need to test individual workflows before release.
  2. It reduces duplication and keeps behavior consistent across automations. Correct answer.
  3. It eliminates the need for version control in the repository.
  4. It guarantees the bot will never encounter an unexpected error.

Explanation

Shared components centralize logic so a fix or improvement propagates to every consumer, reducing duplication and drift. Testing and version control remain necessary.

Question 33Automation and AI fundamentals

Which outcome is the clearest sign that a candidate process is a strong RPA fit?

  1. Low-volume work requiring frequent subjective judgment
  2. Work that changes its rules every week without documentation
  3. A process with no defined inputs or outputs
  4. High-volume, rule-based, stable steps with digital inputs Correct answer.

Explanation

RPA excels where steps are deterministic, repeatable, high in volume, and driven by digital data. Subjective judgment and unstable rules erode the return.

Question 34Automation and AI fundamentals

A bot must fill a form in a legacy Windows application that has no API. Which automation layer is appropriate?

  1. Direct database inserts into the application's backend tables
  2. A REST call to an endpoint the vendor does not publish
  3. Manual entry performed by a human operator on every transaction
  4. UI automation against the application's window controls Correct answer.

Explanation

When no programmatic interface exists, UI automation drives the application through its rendered controls, which is the supported path for legacy software.

Question 35Automation and AI fundamentals

An automation must handle a business rule that says refunds above 500 require manager approval. Where should this threshold live?

  1. In the comments section of the workflow file
  2. In a configuration asset that the workflow reads at runtime Correct answer.
  3. In the job title of the robot's service account
  4. Hard-coded inside the workflow that processes each refund

Explanation

Business thresholds change over time, so externalizing them into a configuration asset lets operations adjust behavior without redeploying the workflow.

Question 36Automation and AI fundamentals

A workflow processes a queue of items and one item fails with a transient network error. What is the most appropriate handling pattern?

  1. Ignore the error and mark the item as successfully processed
  2. Pause the queue indefinitely until an administrator restarts the machine
  3. Retry the item a bounded number of times, then mark it failed for review Correct answer.
  4. Terminate the entire job and discard all remaining queue items

Explanation

Transient failures are best handled with bounded retries so a temporary condition can clear, after which the item is flagged rather than silently lost. Unbounded retries or indefinite pauses risk stalling the queue.

Question 37Automation and AI fundamentals

Which practice best supports a bot that must interact with a web application whose layout changes between releases?

  1. Disabling all logging so selector failures are not reported
  2. Increasing the robot's screen resolution to the maximum available
  3. Using stable attributes and resilient selectors with validation checks Correct answer.
  4. Relying on absolute screen coordinates captured once during development

Explanation

Stable attributes such as IDs and accessible names survive many layout changes, and validation checks surface breakage early. Coordinate-based targeting breaks whenever elements move.

Question 38Automation and AI fundamentals

What is the primary role of a human-in-the-loop step in an automation that uses machine learning?

  1. To replace the model with deterministic rules at runtime
  2. To slow the process so auditors can observe each step
  3. To review low-confidence or exception cases the model cannot settle Correct answer.
  4. To retrain the model from scratch on every transaction

Explanation

Human-in-the-loop routes uncertain or exceptional cases to a person, preserving quality while automation handles the confident majority. It is not a mechanism for continuous retraining or deliberate slowdown.

Question 39Automation and AI fundamentals

A team wants to compare two candidate ML models for a document classification task. Which evaluation approach is most appropriate?

  1. Test both on a held-out labeled dataset and compare precision and recall Correct answer.
  2. Ask the developer which model felt more accurate during testing
  3. Choose the model with the largest parameter count
  4. Select the model that trained fastest on the development machine

Explanation

Held-out labeled data gives an unbiased estimate of generalization, and precision and recall reflect the tradeoffs that matter for classification. Model size or training speed does not indicate accuracy.

Question 40Automation and AI fundamentals

How does a large language model differ from a traditional rules-based chatbot in handling user input?

  1. It requires every possible user phrase to be enumerated in advance
  2. It bypasses the need for any prompt or context when responding
  3. It can only respond to inputs that exactly match a stored template
  4. It generates responses from learned patterns rather than matching fixed rules Correct answer.

Explanation

Language models produce responses by generalizing from training data instead of matching enumerated rules, which lets them handle phrasing never explicitly scripted. They still depend on the prompt and context supplied.

Question 41Automation and AI fundamentals

An automation sends customer data to a third-party AI service for classification. Which control is most directly relevant to protecting that data?

  1. Confirming the provider's data retention and processing terms before sending Correct answer.
  2. Increasing the batch size of requests to reduce API call count
  3. Caching responses locally in an unencrypted text file
  4. Logging the full request payload at the debug level in production

Explanation

Sending personal data to an external service requires verifying how the provider stores and processes it, so the organization can meet its privacy obligations. The other options either ignore or worsen exposure.

Question 42Automation and AI fundamentals

Which statement accurately describes the tradeoff between an API integration and UI automation for the same business function?

  1. API integration is generally more stable and faster but depends on the endpoint being available and supported Correct answer.
  2. UI automation is immune to changes in the target application's layout
  3. UI automation is always faster because it avoids network calls entirely
  4. API integration cannot handle authentication of any kind

Explanation

APIs offer structured, stable contracts and better throughput when available and supported. UI automation is a fallback that is inherently more brittle to interface changes.

Question 43Automation and AI fundamentals

A bot must extract a total amount from an invoice whose format varies across vendors. Which approach is most robust?

  1. Reading the value from the invoice's filename
  2. Splitting the page text on a fixed character offset
  3. Using an ML-based extraction model trained on varied invoice layouts Correct answer.
  4. Assuming the total always appears on the last line of the page

Explanation

Layout variance defeats positional heuristics, so a model trained on diverse formats generalizes better to unseen vendors. Fixed offsets and positional assumptions fail as soon as the template changes.

Question 44Automation and AI fundamentals

Why is a confidence threshold useful when combining an ML classifier with an RPA workflow?

  1. It lets the workflow route uncertain predictions to a different handling path Correct answer.
  2. It removes the need to validate the model on a test dataset
  3. It guarantees the model is correct on every prediction above the threshold
  4. It converts an unsupervised model into a supervised one

Explanation

A threshold partitions predictions into those the workflow can auto-process and those needing review or alternate handling. It manages risk but does not make predictions infallible.

Question 45Automation and AI fundamentals

A team is deciding whether to automate a process with RPA or to reengineer it as an API-first service. Which factor most strongly favors the API-first option?

  1. The underlying application is scheduled for decommissioning next month
  2. The process is used only once per year by a single employee
  3. The process is strategic, high-volume, and expected to persist for years Correct answer.
  4. The process requires a human to judge ambiguous customer complaints

Explanation

For long-lived, high-volume strategic processes, investing in a proper service interface avoids the maintenance burden of brittle UI automation. Short-lived or low-volume cases rarely justify that investment.

Question 46Automation and AI fundamentals

An ML model performs well overall but poorly on a minority class that carries the highest business risk. Which action best addresses this?

  1. Remove the minority class from the dataset to simplify training
  2. Report only overall accuracy to stakeholders and proceed
  3. Adjust the decision threshold or rebalance training data to improve minority-class recall Correct answer.
  4. Increase the model's inference batch size

Explanation

Overall accuracy can mask poor performance on rare but important classes. Threshold tuning or data rebalancing targets the class that matters most operationally.

Question 47Automation and AI fundamentals

Which scenario is the strongest argument against using a generative model to produce legally binding customer communications without review?

  1. The model requires a GPU to run inference
  2. The model's responses can be fluent yet factually incorrect in ways that are hard to detect Correct answer.
  3. The model was trained on text written by humans
  4. The model's output length can exceed a page

Explanation

Fluent but incorrect output is the central risk of generative models in high-stakes communication, since plausibility can mask error. Compute needs, length, and human-authored training data are not the decisive concern.

Question 48Automation and AI fundamentals

An automation must choose among several actions in a simulated environment where rewards are delayed and the environment is fully known. Which learning paradigm is designed for this setting?

  1. Reinforcement learning through trial-and-error interaction Correct answer.
  2. Unsupervised clustering of unlabeled records
  3. Supervised learning on labeled input-output pairs
  4. Rule-based decision tables authored by an analyst

Explanation

Reinforcement learning is built for sequential decision-making where an agent learns from delayed reward signals through interaction. Supervised and unsupervised methods assume labeled or unlabeled static datasets rather than reward-driven interaction.

Question 49Automation and AI fundamentals

A process owner wants to estimate the payback period of an automation investment. Which combination of inputs is most directly required?

  1. The number of developers on the team and their office locations
  2. The count of queue items processed during the pilot week alone
  3. Development and run cost versus the value of labor hours and errors removed Correct answer.
  4. The robot's screen resolution and operating system version

Explanation

Payback compares the total cost of building and running the automation against the quantified benefit it delivers, such as saved labor and reduced errors. Technical or staffing details alone do not yield a payback figure.

Question 50Automation and AI fundamentals

When is a deterministic rules engine preferable to an ML model for a decision step?

  1. When the decision depends on subtle patterns too complex to express as rules
  2. When historical labeled outcomes are abundant and the rules are unknown
  3. When the input is unstructured free text with no consistent format
  4. When the decision boundary is well documented and must be auditable and reproducible Correct answer.

Explanation

Documented, auditable decisions are best served by explicit rules, which are transparent and reproducible by construction. ML is preferable when patterns are too complex or unknown to codify.