AI Automation with n8n interview questions

27 questions from the AI Automation with n8n course, each with the correct answer and an explanation. Try to answer before you open the answer.

n8n Basics

1. What does each node in an n8n workflow represent?

  1. A database server
  2. A whole company
  3. A single step such as an API call or a data change
  4. A user account
Show answer

Answer: A single step such as an API call or a data change

Nodes are individual steps connected on the canvas to form the workflow.

Lesson: What n8n Is and When to Use It

2. Why mount a volume when running n8n in Docker?

  1. So data survives container restarts and removal
  2. To make n8n faster
  3. To open more ports
  4. Docker requires it to start
Show answer

Answer: So data survives container restarts and removal

Workflows, credentials and settings live in the data folder, which must be persisted.

Lesson: Installing and Running n8n

3. What is a trigger node?

  1. A kind of credential
  2. A node that deletes data
  3. The last node in every workflow
  4. A node that starts the workflow
Show answer

Answer: A node that starts the workflow

Triggers such as schedule, webhook or an app event are what cause a workflow to run.

Lesson: Anatomy of a Workflow

4. A node receives 3 items. How many times does it normally run?

  1. Once for all items together
  2. Three times, once per item
  3. Never
  4. Exactly ten times
Show answer

Answer: Three times, once per item

Most nodes run once for each incoming item, though some can process all items together.

Lesson: Items and Expressions

Triggers, APIs and Branching

5. Which URL works only after you activate the workflow?

  1. The test webhook URL
  2. The production webhook URL
  3. Both
  4. Neither
Show answer

Answer: The production webhook URL

Production webhooks are registered on activation; test URLs only listen during manual testing.

Lesson: Choosing a Trigger

6. Where should an API key live in n8n?

  1. In a node field in plain text
  2. In the workflow name
  3. In a Slack message
  4. In n8n Credentials
Show answer

Answer: In n8n Credentials

Credentials are stored encrypted and kept out of exported workflow JSON.

Lesson: HTTP Request and Credentials

7. Why configure a fallback output on a Switch node?

  1. To catch unexpected values so no item is silently lost
  2. Switch cannot run without it
  3. To make workflows slower
  4. To delete the data
Show answer

Answer: To catch unexpected values so no item is silently lost

Without a fallback, an item matching no rule goes nowhere and nobody notices.

Lesson: IF, Switch and Merge

Handling Data

8. What does `$json.address?.city ?? "Unknown"` do when address is missing?

  1. Throws an error
  2. Deletes the item
  3. Returns "Unknown"
  4. Returns the address object
Show answer

Answer: Returns "Unknown"

Optional chaining avoids the error, and the nullish default fills in the fallback value.

Lesson: Edit Fields (Set) and Expressions

9. What must a Code node in "all items" mode return?

  1. A plain string
  2. Nothing
  3. An array of objects with a json property
  4. A number
Show answer

Answer: An array of objects with a json property

n8n expects items in the form { json: {...} }, returned as an array.

Lesson: The Code Node

10. Why process items in batches with a Wait node?

  1. To stay within an API's rate limit
  2. To make data larger
  3. Because n8n forbids loops
  4. To hide errors
Show answer

Answer: To stay within an API's rate limit

Spacing out calls prevents 429 errors from services with request limits.

Lesson: Loops and Batches

11. Which node combines many items into a single list item?

  1. Wait
  2. Error Trigger
  3. Webhook
  4. Aggregate
Show answer

Answer: Aggregate

Aggregate merges the incoming items so the next node, such as an LLM, sees them together.

Lesson: Filter, Sort, Aggregate and Remove Duplicates

AI Nodes in n8n

12. When is a Basic LLM Chain a better fit than an AI Agent?

  1. When many tools must be chosen dynamically
  2. When the task is one prompt and needs no tools
  3. When you want an endless loop
  4. Never
Show answer

Answer: When the task is one prompt and needs no tools

A single prompt step is cheaper and more predictable than an agent loop.

Lesson: Basic LLM Chain and Prompts

13. Why validate data after an Information Extractor step?

  1. Extractors never work
  2. The model might misread values or return null
  3. Validation makes data longer
  4. n8n deletes unvalidated data
Show answer

Answer: The model might misread values or return null

A well-formed JSON object can still contain a wrong number or missing value.

Lesson: Structured Output and Extraction

14. Why include an "other" category in a text classifier?

  1. It hides errors
  2. It speeds up the model
  3. It is required by the licence
  4. So unclear items can be sent to a person instead of being forced into a wrong category
Show answer

Answer: So unclear items can be sent to a person instead of being forced into a wrong category

A safe default prevents confident but wrong routing of odd inputs.

Lesson: Text Classifier and Smart Routing

15. Which sub-node is required to run an AI Agent?

  1. A chat model
  2. A vector store
  3. A Slack node
  4. A Wait node
Show answer

Answer: A chat model

The model provides the reasoning; memory and tools are optional extras.

Lesson: The AI Agent Node

Tools, Memory and RAG

16. What does the model use to decide which tool to call?

  1. The order nodes were created
  2. The tool's colour on the canvas
  3. The tool's name and description
  4. The workflow ID
Show answer

Answer: The tool's name and description

The model only sees the descriptions and input schema, so they must be accurate and specific.

Lesson: Giving an Agent Tools

17. What happens if all users share one memory session ID?

  1. Nothing changes
  2. The model gets smarter
  3. Users see and mix each other's conversation history
  4. n8n crashes immediately
Show answer

Answer: Users see and mix each other's conversation history

Memory is keyed by session ID, so a shared ID leaks context between users.

Lesson: Memory for Chats

18. Why must ingestion and query use the same embedding model?

  1. Vectors from different models are not comparable
  2. It saves disk space only
  3. n8n forbids two models
  4. It is only a naming rule
Show answer

Answer: Vectors from different models are not comparable

Similarity search only works when both sides use the same vector space.

Lesson: A RAG Pipeline in n8n

Reliability and Safety

19. What starts an n8n error workflow?

  1. A Manual Trigger node
  2. A Merge node
  3. A Schedule Trigger only
  4. An Error Trigger node when another workflow fails
Show answer

Answer: An Error Trigger node when another workflow fails

The Error Trigger receives details of the failed execution so you can alert someone.

Lesson: Errors, Retries and Error Workflows

20. What does enabling authentication on a Webhook node prevent?

  1. The workflow from ever running
  2. Anyone with the URL triggering the workflow
  3. The use of JSON
  4. Logging
Show answer

Answer: Anyone with the URL triggering the workflow

Only callers who present valid credentials or signatures can trigger it.

Lesson: Securing Webhooks and Credentials

21. Which is the most reliable defence against prompt injection?

  1. A longer system prompt only
  2. Limiting the agent's tools and requiring approval for risky actions
  3. Hiding the prompt
  4. Using bigger fonts
Show answer

Answer: Limiting the agent's tools and requiring approval for risky actions

Capability limits work even if the model is fooled by injected text.

Lesson: AI Guardrails and Cost Control

Deploying and Operating n8n

22. Why set WEBHOOK_URL when n8n is behind a reverse proxy?

  1. It is only cosmetic
  2. To speed up the database
  3. To disable authentication
  4. So generated webhook URLs use the correct public address
Show answer

Answer: So generated webhook URLs use the correct public address

n8n cannot guess its external address behind a proxy, so you tell it.

Lesson: Self-Hosting with Docker Compose

23. Why split a large workflow into sub-workflows?

  1. Smaller parts are easier to test, reuse and understand
  2. n8n limits workflows to 3 nodes
  3. Sub-workflows are always faster
  4. It removes the need for names
Show answer

Answer: Smaller parts are easier to test, reuse and understand

Modular workflows are maintainable, like functions in code.

Lesson: Managing Workflows Like Code

Putting It Together

24. Why does the case study require approval before sending the reply?

  1. Emails cannot be sent otherwise
  2. Slack requires it
  3. Sending is an outward action that is hard to undo and model output may be wrong
  4. It makes the model faster
Show answer

Answer: Sending is an outward action that is hard to undo and model output may be wrong

A human check catches wrong or risky drafts before a customer sees them.

Lesson: Case Study: AI Support Triage

25. Your webhook works when testing but never fires in production. What is the most likely cause?

  1. The model is too small
  2. JSON is broken
  3. The workflow is not activated or the test URL is being called
  4. Docker is installed
Show answer

Answer: The workflow is not activated or the test URL is being called

Production webhooks run only for active workflows and use the /webhook/ path, not /webhook-test/.

Lesson: Revision: Cheat Sheet and Self-Check

26. You lose N8N_ENCRYPTION_KEY after moving servers. What is the consequence?

  1. Saved credentials cannot be decrypted
  2. Nothing, it is optional
  3. Workflows run twice as fast
  4. The editor gets a new theme
Show answer

Answer: Saved credentials cannot be decrypted

Credentials are encrypted with that key, so without it they are unreadable and must be re-entered.

Lesson: Revision: Cheat Sheet and Self-Check

27. Which step belongs behind human approval in an AI email-reply workflow?

  1. Reading the incoming email
  2. Classifying the email
  3. Logging the token count
  4. Sending the reply to the customer
Show answer

Answer: Sending the reply to the customer

Sending is outward-facing and irreversible; reading and classifying are internal and safe.

Lesson: Revision: Cheat Sheet and Self-Check

← Back to the course