From MERN Stack to n8n: How I Backed Into Building AI Automation for Kenyan SMEs

In 2023 I was writing React and PHP for a learning management system, six modules, role based access, the usual stack. I called myself a MERN developer because that's what the job postings wanted. I didn't know yet that the actual skill I was building, reading a business, mapping its mess, and wiring the fix, would matter more than any framework I knew.

By 2024 I'd picked up Zapier almost by accident, connecting a form to a spreadsheet for a client who just wanted leads to stop dying in someone's inbox. It worked. It felt like a magic trick. Then late 2024 I found n8n, and that's when automation stopped being a side skill and became the thing.

The Ace: fundamentals beat fancy, every time

If there's one thing two years of client work has hammered into me, it's this. The automations that actually survive contact with a real business are boring on purpose. The flashy ones, the ones with fifteen branching nodes and three AI agents talking to each other, usually break in week two because nobody understood the business well enough to build something simpler.

To anyone starting in automation right now: learn triggers, learn error handling, learn how to read a webhook payload before you touch anything labeled "agent." Grasp the fundamentals. Don't rush to build something that looks impressive in a demo and falls over the first time a client's data doesn't match the shape you expected.

Why n8n won, and why it wasn't really about n8n

Zapier taught me that non-technical clients don't want to see complexity, they want the outcome. But it also boxed me in. The moment a workflow needed real logic, a conditional that wasn't a simple if-this-then-that, I hit a wall.

n8n removed the wall. The Code node meant I could write actual JavaScript inside a workflow instead of fighting a rigid builder. That mattered more than any single feature, because it meant the tool could bend to the business instead of the business having to bend to the tool.

Pair that with Supabase for a backend I could stand up in an afternoon and Airtable for the clients who wanted something they could look at without learning SQL, and I had a stack that could move as fast as I could think. Self-hosting on a Hostinger or Railway box meant the client owned the workflow too, not just the outcome. I'd spent a year writing React components and PHP endpoints to get the same result n8n gave me in a fraction of the time, for clients who couldn't afford to pay for that build time anyway.

What nobody tells you: automation is mostly business archaeology

Here's the part that surprised me most. Building the workflow is maybe thirty percent of the job. The other seventy percent is sitting with a business owner and figuring out how their business actually runs, which is almost never how they describe it on the first call.

In one stretch, January had me inside a law firm's client intake process, learning how conflict checks and retainer agreements actually move through a small practice. February was a dental clinic, where the real problem wasn't appointment booking, it was no-shows nobody was tracking. March split between a furniture retailer automating WhatsApp sales end to end, and a venture firm that needed structured deal flow instead of a Notion page nobody updated.

Four completely different worlds in three months. Law, healthcare, retail, venture capital. None of them run on the same logic, and pretending they do is how you end up delivering a generic chatbot that solves nothing. Every one of those projects started with me learning their business model before I opened n8n once.

That's the honest, unglamorous truth about this work. It's challenging in a way that has nothing to do with code and everything to do with listening. And it's rewarding in a way that a clean pull request never quite is, watching a dental clinic's front desk stop drowning in phone calls, or a furniture shop's WhatsApp finally keep up with demand.

African problems, African builders

Most of the automation content and tooling I learned from in 2024 assumed a Western SME. Stripe, not M-Pesa. Email, not WhatsApp. A stable fiber connection, not a business that runs its entire order book through a phone that sometimes drops to 3G.

I don't think that's a small gap. A generic automation agency can bolt a chatbot onto a Kenyan business. Someone who has actually reconciled M-Pesa payments by hand, who knows why WhatsApp Business is the real storefront for most SMEs here, who understands that staff turnover means your automation has to survive being handed to someone new, builds something different. African problems get solved better by people who've lived inside them. That's not a slogan, it's just what two years of client work across Nairobi has shown me directly.

Why I started CustomCX

Somewhere around the furniture shop and the venture firm, I noticed the same failure pattern everywhere. The moment an SME starts scaling, customer experience is the first thing that quietly breaks. Response times slip. The personal touch that got them their first hundred customers disappears under the volume of the next thousand. Owners notice revenue before they notice churn, so it goes unfixed longer than it should.

That's the gap CustomCX exists to close. Not automation for its own sake, but automation that protects the customer experience while a Kenyan SME scales, using AI where it genuinely helps and staying out of the way where it doesn't. It's the same lesson from the law firm and the dental clinic and the furniture shop, just distilled into one focus instead of scattered across four industries a month.

If you're starting out

Don't chase the most complex workflow you can build. Chase the smallest one that actually fixes something for a real business, and understand that business before you touch a single node. The tools will keep changing. Zapier led in 2024, n8n took over for me by early 2025, something else will matter more in another two years. What doesn't change is whether you actually understood the problem before you built the solution.

If you're a Kenyan SME watching your customer experience slip as you grow, or a founder who wants automation built by someone who has actually sat inside businesses like yours, that's the conversation worth having. CustomCX is where that work lives now.