From 0 to 7,500 Users on a WhatsApp Banking Platform — What Broke, What Held
title: "From 0 to 7,500 Users on a WhatsApp Banking Platform — What Broke, What Held" description: "Three months building Kira AI in production. The technical decisions mattered less than I expected. The product and operational decisions mattered more." date: "2026-04-08" series: "notes" tags: ["fintech", "whatsapp", "scaling", "kira"] linkedinUrl: "https://www.linkedin.com/posts/divine-chukwu-63bb04145_from-0-to-7500-users-on-a-whatsapp-banking-activity-7444654931862503424-MKy1"
Been a couple of months since we launched Kira AI, our WhatsApp-native banking platform, and the things that actually mattered were nothing like what I expected when we started.
Broadcasting is harder than it looks on WhatsApp
We wanted to communicate with our users and since Kira lives entirely on WhatsApp, broadcasting on WhatsApp was the obvious channel. No app, no push notifications, just WhatsApp.
But Meta categorizes promotional messages under marketing and they are not delivered to all users by default. I had to learn how Meta actually works in production, not how the documentation says it works.
What looks simple on paper becomes a real operational challenge when you have thousands of users expecting to hear from you.
We over-engineered the conversation
We spent significant time building a sophisticated conversational experience because we wanted Kira to feel human and not robotic.
Turns out users primarily wanted one thing: to send money reliably and on time. Simplicity beats cleverness when real money is involved. People care about their money arriving correctly far more than they care about how elegant the conversation feels.
We learned to build for what users actually do, not what we thought they would appreciate.
Monitoring was an afterthought and we felt that cost
We built fast and shipped fast but observability came last. We were reactive instead of proactive — finding out about problems when users reported them instead of before they did.
Everyone wants to build the next elegant feature, but without deep insight into how your system is actually performing you are flying blind.
If I could go back, monitoring and observability would be built in from day one, not added later.
Three months. 7,500 users. ₦272M+ processed. The technical decisions mattered less than I expected. The product and operational decisions mattered more than I anticipated.
What's a lesson you learned the hard way building a real product?
Originally shared on LinkedIn.