I wanted to share a perspective that might resonate with people who have been using n8n for a while and have started to run into its limits. I have not abandoned it completely, but I have changed how I use it and this shift has made a massive difference in how stable and scalable my automations are.
I first started using n8n in early 2023. Like many, I was blown away by how quickly I could connect apps, build workflows visually, and get an MVP running in hours instead of days. It was my go-to for client projects, internal tools, and quick experiments.
However, over time my workflows got bigger, my client demands got more complex, and I started to notice patterns: large workflows slowed down and became harder to debug, AI integrations were cool in demos but unpredictable in production, handling files, especially big ones, was painful and full of workarounds, and changing business logic in a giant visual workflow without breaking something was stressful.
One day I decided to treat n8n as my orchestrator, not as my entire system. I kept 20% of the work, the high-level coordination, triggers, and light integrations, in n8n, and moved the other 80%, the heavy data processing, file handling, AI logic, and complex conditionals, into small, focused Python scripts or microservices. n8n triggers the scripts, gets the result, and moves on. No more trying to cram every step into a giant spaghetti workflow.
The results were amazing: my systems became 5x more reliable almost overnight, debugging was simple, scaling up was easier, and clients noticed faster performance and fewer “mystery” errors.
I know a lot of people love n8n (so do I) and for many projects it really is all you need. But if you are hitting walls with performance or complexity, you do not have to choose between all n8n or no n8n. A hybrid approach can give you the best of both worlds.
I am curious how many of you already mix n8n with custom code? And for those who do, what is your split between visual workflows and traditional programming?