n8n Good, Bad, and Ugly From External User Interview Synthesis
- Harshal
- 6 days ago
- 5 min read
Customer Discovery Deep-Dive Example For PM Interviews
Here, I show a deep dive into a product memo example I did for n8n a few months ago, when I interviewed with them. Fast-forward, I got a job offer and will join them in Q4 2025. Since this doesn’t have n8n employee inputs, the information is not accurate, but I hope it gives you ideas on the approach.
In earlier posts, I shared a framework to write a case study or product memo when you interview for Product Management roles with companies.
I interviewed 6 users or leads with different backgrounds through my personal network.
These insights may not be relevant anymore, as I researched and wrote these when I was not part of n8n. My goal here is to show you a deep-dive example for your Product Management interviews.
Related:
Whom Did I Interview?
I interviewed six people from my network and replaced their names with persona labels for this blog.
Automation Agency Owner: runs an automation agency, experienced with Zapier, n8n, Make, and CRMs.
Startup Engineer: a programmer who implemented n8n in enterprise workflows and self-hosts it.
Low-Code Builder: agency user familiar with Zapier and Make, tried n8n but found it difficult.
Enterprise AI Consultant: a consultant enabling AI in enterprise workflows.
Enterprise POC Builder: technical business user, used n8n for enterprise proof-of-concepts.
AI Product Manager: tech-savvy business user who built agents on AgentAI.
Questions Asked
Directly or indirectly, I wanted to know these from users and leads:
Can you walk me through how you currently use n8n?
What works well
What tools or services do you connect to n8n?
What’s the most annoying part of using n8n right now?
If you could wave a magic wand and fix or improve one thing about your n8n setup, what would it be?
Anything else?
Can You Walk Me Through How You Currently Use n8n?
Automation Agency Owner:
“I rely on n8n when I need my AI agents to decide which tool to use next.”
AI Agent workflows: Used for automating tasks where LLMs choose tools, compose replies, or maintain context (e.g., Automation Agency Owner, Enterprise AI Consultant).
Data scraping and chatbots: Used for scraping, chatbot workflows, and parsing input triggers (e.g., Startup Engineer).
Multi-agent, V2V automation: Built multi-agent setups for tasks like voice-to-voice orchestration or AI-assisted legacy modernization (e.g., Enterprise AI Consultant).
Internal workflows: Routing form data, Slack automations, conditional sequences (e.g., Enterprise POC Builder).
Enterprise AI Consultant:
“Clients ask us for futuristic use cases like voice-to-voice orchestration.”
What works well
AI Product Manager:
“It feels more like a real development environment than just a no-code tool.”
Startup Engineer:
“Running it on my own server means no lock-in.”
Rapid prototyping: Conditional logic and branching makes n8n flexible to experiment beyond simple automations.
Flexible, modular workflow design: Users consistently appreciate the visual, node-based interface and the ability to build complex automations with conditional logic and branching.
AI Agent Node and AI integration: Automation Agency Owner highlights this as n8n's most unique differentiator, letting LLMs pick tools, maintain memory, and handle decision-making.
Self-hosting and cost control: Startup Engineer values the ability to self-host on a modest machine, reducing reliance on SaaS pricing and gaining full control.
Customization with JavaScript and HTTP: Advanced users insert custom JS logic or hit external APIs, making n8n feel like a low-code development environment.
Reusable subflows: Ability to create modular pieces of workflows and reuse them makes complex automation scalable and maintainable.
Community and Templates: Templates don’t fully meet complex needs but provide a valuable starting point for new users.
Platform reach and integrations: Even users who found it too technical (like Low-Code Builder) acknowledged its broad potential; more powerful than Zapier or Make once you overcome the learning curve.
Partner enablement: Enterprise AI Consultant noted that consulting firms embed n8n into client systems to showcase POCs quickly.
Enterprise POC Builder:
“I build small blocks and reuse them, that saves me hours when projects get bigger.”
Low-Code Builder:
“Zapier is simple but limited. n8n lets me go further, even if it takes longer to learn.”
What Tools Or Services Do You Connect To n8n?
Automation Agency Owner:
“CRMs and email tools are bread-and-butter for agencies.”
CRMs, email platforms, and project management tools (e.g., Automation Agency Owner).
Slack, Google Sheets, Airtable (e.g., Enterprise POC Builder).
Internal APIs and databases (e.g., Startup Engineer).
Zapier and Make migrations (e.g., Low-Code Builder).
Low-Code Builder:
“Clients already use Zapier or Make. I migrate them to n8n when they need more flexibility.
What’s the most annoying part of using n8n right now?
Startup Engineer:
“When something breaks, I’m guessing where.”
Debugging is hard: Trial-and-error debugging with unclear error handling. No obvious visualization of how data passes between nodes.
Expression editor complexity: Not intuitive for non-technical users, especially for basic data references.
AI prompting is fragile: Requires precise prompts for accurate tool invocation and correct decision logic.
Token costs and model selection: LLM-powered nodes introduce cost and latency tradeoffs.
Vector setup and schema mismatches: Setup for retrieval-based AI use cases is non-trivial.
Too technical for some users: One user switched to Zapier or Make due to complexity (Low-Code Builder).
Template recommendation feels rigid: Users feel forced into one template path, even when the actual need is different. Search lacks flexibility.
AI Product Manager:
“Even simple data lookups feel harder than they should be.”
Low-Code Builder:
“Zapier is point and click, with n8n I need to know too much up front.”
If You Could Wave A Magic Wand And Fix Or Improve One Thing About Your n8n Setup, What Would It Be?
Startup Engineer:
“Give me one clear debugging view and I’d save hours each week.”
Low-Code Builder:
“I want my non-technical teammates to build too. Right now, they hit a wall fast.”
Native debugging console with clearer logs (e.g., Startup Engineer).
Easier onboarding for non-technical team members (e.g., Low-Code Builder).
More reliable AI node orchestration with evaluation built-in (e.g., Enterprise AI Consultant).
Templates tailored for enterprise POCs and marketing workflows (e.g., Enterprise POC Builder).
Ability to scale workflows without custom code (AI Product Manager).
Enterprise AI Consultant:
“If evals were built-in, I’d trust AI workflows more.”
Enterprise POC Builder:
“A template that speaks the language of marketing would save us days.”
AI Product Manager:
“Scaling should not require writing scripts, it should be built-in.”
Anything else?
"We want AI but don’t know what to do": Businesses want to benefit from AI, so agency owners must help businesses bridge intent with execution.
Templates help, but fall short: Great for onboarding or inspiration, but not adaptable to specific business logic.
Power users appreciate customization: Self-hosting, JavaScript nodes, and modular subflows enable deep customization but require technical comfort.
AI model orchestration is seen as the future: Multiple users referenced MCP (Model Control Protocol) as a desired direction, natural language control over APIs.
Explosive integration surface is key to agent workflows: Enterprise AI Consultant highlighted that expanding n8n’s integrations is what unlocks most use cases.
Community node friction: Templates using community-contributed nodes often break because the required package isn’t auto-installed. A system like npm install for nodes would solve this.
Rest Of The Product Memo
You can read rest of the product memo here.
Related:
