Building an AI agent with n8n without coding: full feedback

Can you build an AI agent without being a specialist? That’s the challenge Clara Aubès, Product Marketing Manager at Thiga, decided to take on. She set out to create an AI agent capable of structuring field feedback, without knowing how to code. Her tool of choice: n8n, a no-code platform widely embraced by the Product community. Here’s her experience, learnings, and advice for anyone who wants to dive in without a technical background.

I’m a Product Marketing Manager, and I built a no-code AI agent with n8n.

Artificial intelligence is everywhere in companies today, and it’s no longer just for tech teams. It has become a real lever to move faster and do better. At Thiga, our Product Marketing team started wondering how AI could help us do our jobs better. Not by producing more content, but by making our feedback more useful and more actionable.

That’s what led me to explore the idea of building an AI agent capable of centralizing and structuring field feedback. The goal? To streamline how feedback is shared, avoid losing it along the way, and enable Product teams to integrate it more effectively into their decision-making. The challenge? I don’t have a particularly technical background. I’ve never learned to code (aside from a few SQL queries) and I’ve never really dealt hands-on with APIs, even though I have a decent grasp of how they work.

Even though n8n doesn’t explicitly market itself as a no-code tool for non-technical profiles, it quickly gained traction within the Product community thanks to its simplified user experience. It makes it possible to create AI-powered automations with minimal technical hurdles.

Here’s my experience, and a few tips to help you get started!

Side note: before diving in, I attended a demo run by a Thiga consultant who’s an n8n expert, and that saved me a lot of time at the start.

Tired of blindly investing in artificial intelligence? Download our free AI Product Canvas, to ask yourself all the right questions before embarking on an AI project.

The upsides of n8n

An intuitive UX that lets you learn while building

n8n’s user experience is particularly smooth and intuitive, which lowers the barrier to entry. The interface is based on a simple visual logic: each step in the workflow is represented by a node, which you can easily connect to others. This graphical representation lets you quickly see the overall structure of your automation, helping you clarify your thinking.

For me, this is one of n8n’s biggest strengths: you don’t need a perfectly clear vision of the end result to get started. The tool fits perfectly with an iterative approach: you build your workflow step by step, refining the logic as you go.

Each node can be tested individually, which makes it easy to validate assumptions, fix mistakes, and learn along the way.

No-code automations within reach

You can absolutely build simple agents without writing a single line of code, especially if you stick to native integrations (Google Sheets, OpenAI, Notion…). In my case, I managed to build a fairly complete AI agent with very little coding. And when I did need some, I generated the missing lines through ChatGPT rather than writing them myself.

An active community

n8n has a lot of momentum, especially among PMs, which means there’s a highly active community, plenty of workflow examples online, responsive forums, and regularly updated tutorials.

A very complete free version

The open-source self-hosted version gives you access to most of the features at no cost. I installed it locally on my computer, and it’s a real asset for testing, experimenting, or building POCs without immediately worrying about pricing limits.

Easy sharing

Another strong point: you can copy-paste a flow to share it with a colleague or post it in the community. This makes collaboration and reuse straightforward.

n8n's downsides

Technical friction points that aren’t always clear

Yes, you can test each node individually, but the error messages aren’t always explicit. When something breaks (often around data formats or API requests), it can be hard to understand the root cause, even with ChatGPT’s help. For example, I once spent hours fixing a formatting error between two nodes, without ever fully understanding why the suggested fixes weren’t working.

Little visibility on AI-related costs

Integrating AI models like GPT is simple, but tracking token usage isn’t very transparent. (A token is essentially a word or piece of a word processed by the model and is the main billing unit for models like ChatGPT.) In terms of tracking, you often don’t know exactly what was sent, how much it cost, or why a task is taking longer to execute. This can become a real issue if you’re using a paid model (which is required for certain requests) and need to anticipate or control costs.

Unnecessary complexity at times

n8n is very powerful, but it doesn’t always guide you toward the simplest solution. You might end up building a convoluted sequence when a more direct option exists (for example, via another native node or a logic change). Without much technical experience, it’s easy to overcomplicate things.

Slower when handling large volumes

For my AI agent, I had to process a large amount of data (over 5,000 items to analyze). That’s where n8n showed its limits. Each testing phase took a lot of time because I had to relaunch the entire process from the beginning.

There are likely solutions for handling larger volumes more efficiently, but they generally require more advanced technical skills.

My tips for getting started

Use AI assistants

LLMs like ChatGPT or Gemini are invaluable allies. There are even assistants already set up specifically to help create or troubleshoot n8n workflows. You can ask them to generate JavaScript for “Function” nodes, help configure certain nodes, or even advise on which steps to add to a workflow.

As with most LLM use cases, the answers aren’t always perfect, especially if your prompt isn’t very precise. But don’t hesitate to ask for multiple possible solutions and test different ones until you find what works.

Start small

Don’t try to do everything at once. My advice is to build a simple end-to-end flow first, even with fake or reduced data, to grasp the logic and test the basic nodes. Then gradually add complexity.

For projects that involve large volumes, it’s important to anticipate this upfront and set up mechanisms for segmentation or batch processing to optimize performance and maintainability.

Lean on the community

The n8n community is very active. You’ll easily find shared workflows, tutorials, bug discussions similar to your own, and even plugins. It’s a real accelerator for learning and unblocking issues faster.

n8n is an extremely powerful tool for automating simple to moderately complex tasks. It enables non-developers to prototype, connect tools, and integrate AI into their workflows.

But once you start dealing with large data volumes, multiple use cases, or complex logic, the limits become clear:

lack of clarity in error messages,
difficulty anticipating all edge cases (errors, exceptions, data formats),
long execution times on heavy workflows,
and complexity in estimating or controlling AI-related costs.

In short, n8n is perfect for targeted, lightweight, well-defined automations—but it requires deeper technical skills when you move into more complex workflows and processes.

Learn Generative AI with Thiga Academy! Artificial intelligence is reshaping the role of the Product Manager. Our training helps you and your teams integrate it into your products in a concrete way—so you can deliver more value to your users.

What you’ll learn:

Grasp the fundamentals of AI (ML, NLP, generative AI, LLMs…) and their real-world applications
Identify, prioritize, and design high-impact AI features
Master the basics of data management: collection, cleaning, annotation, governance
Frame, test, and deploy an AI model—from MVP to large-scale rollout
Balance technical performance, business value, and user experience
Collaborate effectively with Data Engineers, Data Scientists, and Analysts
Apply the principles of responsible, ethical, and regulated AI