AI Agents vs Automation vs Workflows

Question 1

Imagine three types of coffee shops:

**Shop A**: You press a button, coffee comes out. Same coffee every time. Fast, predictable.

**Shop B**: A barista takes your order, remembers you like oat milk, notices you look tired, and suggests a double shot today. Adapts to context.

**Shop C**: A team of specialists - one grinds beans, one steams milk, one does latte art - coordinated by a manager who assigns tasks based on the order.

Which shop represents which concept?

Accepted Answer

Shop A = AI Automation, Shop B = AI Agent, Shop C = Agentic Workflow (Multi-Agent System)

Answer

Shop A = AI Agent, Shop B = Workflow, Shop C = Automation

Answer

They're all the same thing with different marketing names

Answer

Shop A = Agentic Workflow, Shop B = Automation, Shop C = AI Agent

Question 2

What is the CORE technical difference between workflow automation and agentic automation?

Accepted Answer

Workflows decide based on predefined conditions (code), agents decide based on real-time predictions (model)

Answer

Workflows are slower, agents are faster

Answer

Workflows use Python, agents use JavaScript

Answer

Agents require more RAM than workflows

Question 3

Your friend shows you their 'AI agent' built in n8n. It:

1. Receives customer emails
2. Uses an LLM to classify sentiment
3. Routes to different Slack channels based on sentiment
4. Sends a templated response

They call it an 'AI agent' but something feels off. What is this actually?

Accepted Answer

This is AI-enhanced automation (workflow with AI steps), not an agent - it follows a predefined linear flow with no autonomous decision-making

Answer

This IS a true AI agent - it uses an LLM!

Answer

This is an agentic workflow with multiple agents

Answer

n8n can't do AI at all

Question 4

Your friend extends their n8n 'agent' to handle complex customer issues:

1. If the first response doesn't resolve the issue, try a different approach
2. Keep trying until the customer is satisfied
3. Use tools to search the knowledge base and check order status

After deployment, they wake up to a $500 OpenAI bill and an n8n crash log showing 'Max iterations exceeded.' What went wrong?

Accepted Answer

n8n's stateless architecture and lack of proper loop control caused the agent to get stuck in an infinite loop, burning through API calls

Answer

OpenAI's API was having issues

Answer

The customer was just really difficult

Answer

They should have used more nodes

Question 5

LangGraph represents agent workflows as graphs with nodes and edges. Why is this architectural choice particularly suited for AI agents?

Accepted Answer

Graphs enable conditional branching, loops, state persistence, and human-in-the-loop interrupts - essential for autonomous reasoning

Answer

Graphs look prettier in documentation

Answer

Python requires graph structures for AI

Answer

LangChain forced them to use graphs

Question 6

Match each use case to the best tool:

**Use Case A**: Sync new Salesforce leads to HubSpot, enrich with Clearbit data, notify Slack

**Use Case B**: Customer support agent that searches knowledge base, checks order status, escalates complex issues, and learns from feedback

**Use Case C**: Code review agent that analyzes PRs, suggests fixes, can revise based on feedback, and coordinates with a testing agent

Accepted Answer

A: n8n, B: LangGraph, C: CrewAI or AutoGen

Answer

A: LangGraph, B: n8n, C: CrewAI

Answer

All three should use n8n because it has 500+ integrations

Answer

All three should use LangGraph because it's the newest

Question 7

You need to build a multi-agent system. When should you choose each framework?

Accepted Answer

LangGraph for regulated/auditable workflows, CrewAI for fast role-based prototyping, AutoGen for code generation and research

Answer

They're interchangeable - pick whichever has better docs

Answer

Always use LangGraph - it's from LangChain

Answer

AutoGen is best because Microsoft made it

Question 8

You're building an expense report processor. Initially it seems simple, but then you learn:
- Policies change monthly
- International receipts have different formats
- Some expenses need manager + finance approval
- Employees often submit incorrect categories

Should you use a workflow or an agent?

Accepted Answer

Agent - the variations and edge cases make a rigid workflow brittle; an agent can adapt to policy changes and unusual situations

Answer

Workflow - expenses are a well-defined business process

Answer

Neither - use a spreadsheet

Answer

Workflow now, rewrite as agent later

Question 9

An 'agentic workflow' is not the same as a pure 'AI agent'. What's the key architectural difference?

Accepted Answer

Agentic workflows insert LLM-powered steps into a controlled pipeline, maintaining predictability; pure agents have full autonomy over their actions

Answer

Agentic workflows use more compute

Answer

There is no difference - they're marketing terms

Answer

Pure agents are always better than agentic workflows

Question 10

A bank wants to automate loan application processing:

- **Happy path (80%)**: Standard applications, clear approval criteria
- **Edge cases (20%)**: Unusual income sources, international applicants, exceptions

What's the recommended architecture?

Accepted Answer

Hybrid: n8n workflow for the happy path (predictable, auditable), LangGraph agent for edge cases (reasoning, adaptation)

Answer

Pure n8n workflow - banks need predictability

Answer

Pure LangGraph agent - let AI handle everything

Answer

No automation - financial services are too risky for AI