When we signed a contract to handle 3,000 active loan accounts, I don’t think either side fully understood what that would mean physically.
It means roughly 50 phones. Charging in a client’s office. Each one connected to a WhatsApp number we’ve spent weeks warming up. Each one capped at 32 messages a day to avoid triggering Meta’s abuse detection.
Our ops person sits in their office and scans them. One by one.
This is not the AI future anyone imagined.
But here’s what I’ve come to understand: the phone farm isn’t a failure of ambition. It’s a proof of product-market fit. Clients will tolerate real operational pain if the underlying outcome — getting debtors to respond — actually works. And it does. Our AI outperforms their human collectors on contactability at a fraction of the cost.
The phone farm is a transitional architecture. It’s ugly, and we’re replacing it with a proper WhatsApp Business API integration. That’ll take another 6-8 weeks.
Until then, we have 50 phones in a client’s office and a very patient ops team.
I used to be embarrassed by this. Now I think the willingness to do the dirty transitional work is part of what makes a product real.
What’s the most unsexy operational thing you’ve built that actually worked?
Leave a Reply