No-code AI on n8n for white collars (08.02.25 workshop notes)

tldr: n8n is a cool no-code alternative for simple rudimentary tasks for white-collar folks. Try it, guys.

First things first - thanks to Bartosz Mikulski for leading the workshop and making it noob-friendly.

While diving into agentic workflows and trying to grasp the powers of Langchain and Langgraph as a noob builder with limited Python skills, I make an effort to assess the "connector" space for the general audience out there, i.e. white collars trying to simplify & speed up their work using GenAI.

For some of us the no-code solutions are an obvious first choice. Fast prototypes, simple builds you don't need to pay for, a very simple MVP if you need one.

The no-code solutions have been there for a while now - headless drag-n-drop databases, app builders like Webflow and Bubble, process automatons like Zapier or IFTTT.

But I wanted to see the components available today to integrate LLMs into those flows, the actual possibilities and the extent of skills one needs to be fully capable in the white-collar job market of today.

Well, I was quite surprised. You don't need much knowledge to integrate those models on the most basic level which is still quite a lot to play with.

During the workshop of 8 hours on n8n, we build simple workflows combining ETL with simple LLM nodes, connecting services like Notion or Open Weathermap through API and letting ChatGPT decide what to do with that data (or explicitly instructing it on the output).

A beginner agent if you wish, the one you can build in minutes for a specific rudimentary task and forget for the time being.

While n8n isn't exactly cheap (their starter package goes for 29$/month after the trial of 14 days), you can host your own instance on your own laptop using docker.

Just grab that old Lenovo laptop, burn some Linux into it and voila - your home server is up to run those flows. It takes 3 terminal instructions to spin it up.

When ready, you are able to access that platform from your own browser and start building (drug-n-drop mostly).

During the workshop we have built simple flows to explore all the basic moving pieces of the platform like extraction, HTTP calls, filtering and math operations, ChatGPT calls with a custom prompt, text summarization and even some basic work with the vector databases (you don't need to set it up though, that would be a big hassle).

Example:

As you can see, we are taking something (a bunch of links in Notion), pull the HTML content, parse it, summarize it using our beloved ChatGPT and ask questions (actions, research ideas) based on that summary. We record our results in Notion as our flow closure (just read the flow left to right).

I took a bunch of screenshots but those basically depict the very same concept: the pipeline of small operations you would build in your service by writing some code and running somewhere. The blocks on n8n replace that part for you by allowing you to focus on the actual working prototype.

While I wouldn't consider it a production-ready thing, for solo-preneurs and white-collars this thing is a definite win.

Try it out, folks.