Kasi: The AI Agent That Never Sleeps

Executive Summary

Every time, thousands of sales die in unread DMs.

Kasi is an AI-powered Sales Agent purpose-built for Nigerian social commerce. It lives inside the messaging channels where Nigerian trade already happens — Telegram and WhatsApp — and it works the hours that humans cannot. Kasi negotiates on behalf of vendors, guards their floor prices, generates branded PDF invoices the moment a deal is struck, and routes customers to payment — all without the vendor lifting a finger.

The insight behind Kasi is deceptively simple: in social commerce, speed is trust, and trust is revenue. The moment a vendor says "Check my DM" instead of quoting a price, they have introduced friction. Friction, in a market this competitive, is churn. Kasi eliminates that friction entirely.

— The Story

A Tale of Two Vendors

To understand why Kasi exists, stop thinking in metrics. Start thinking in people. Specifically, two people — both talented, both hustle-driven, one with a tool that changes everything.

— Live Demonstration

Watch the Negotiation Happen

This is a real reconstruction of the 11:47 PM conversation. Chioma is asleep. Kasi is not.

— The Problem

"Check DM" is a technical bottleneck, not a strategy

To an outsider, "Check DM" sounds like a sales tactic. To any Nigerian who has bought or sold on social media, it is a warning sign — the digital equivalent of a shop with no price tags and a slow cashier.

Nigerian social commerce lacks a checkout layer. Every platform — Shopify, Amazon, Jumia — has invested billions in reducing steps between discovery and payment. Social vendors here are running on zero of that infrastructure, using human response time as their transaction engine.

Core Product Diagnosis — Kasi Engineering Team

Human response time is bounded by sleep, school, family obligations, and the fundamental limits of biology. An AI agent has none of these constraints. But solving latency alone is insufficient.

The second, thornier challenge: how do you build an AI that understands Nigerian bargaining culture without letting it give your products away? Nigerian commerce is relational, not transactional. Customers expect to negotiate. A vendor who refuses is inflexible; one who folds too easily signals desperation. The acceptable zone of negotiation is narrow, culturally contextual, and deeply intuitive. Teaching a machine to replicate that instinct — without a rigid, brittle rule system — was the central engineering challenge of building Kasi.

— The Solution

Engineering an Agent for the African Market

— Business Model

Pay-Per-Success, Not Pay-Per-Existence

The flat subscription model has a fatal flaw when applied to small Nigerian vendors: it charges for presence, not performance. A vendor who has a slow month still pays. The product becomes a liability the moment sales slow down.

In early testing, vendors reported that receiving a credit deduction notification — paired with a near-simultaneous bank credit alert — was actively pleasurable. The deduction confirmed Kasi had worked. It was proof of performance, not a cost reminder. A platform fee that arrives as good news is extraordinarily rare in software.

UX Observation — Early Cohort Feedback

The credit system also creates natural alignment: Kasi only earns when the vendor earns. Bulk bundles reward growing vendors with volume discounts. There is no cliff edge — no moment where growth triggers a punishing new pricing tier. The cost scales smoothly with success, exactly the way a good business partner should.

— Results & Signals

Early Numbers, Strong Conviction

The WhatsApp beta cap was a deliberate product decision — not a supply constraint, but a trust-building mechanism. Capping the first cohort at 50 spots allowed the team to monitor agent performance at controlled scale before wider rollout, while simultaneously generating the kind of social proof that no marketing budget can manufacture: vendors talking to other vendors about how to get access.

This is the correct way to launch agentic products into markets that have been burned by overpromised, underdelivered tech solutions. Build trust slowly. Earn the right to scale.

— Lessons Learned

Building Agentic Workflows for the Hustle Economy

Building Kasi surfaced insights that apply beyond this product — to anyone deploying AI agents in emerging markets, with non-technical users, in high-trust, high-stakes contexts.

Constraints are features.

The floor price is not a limitation — it is the product. Giving vendors less control over the AI (it cannot sell below floor, full stop) gives them more confidence in deploying it. Agentic systems for non-technical users must have hard rails, not soft guidelines. Trust in automation is built through predictability.

Culture is infrastructure.

Building an AI that speaks Pidgin-inflected English, understands the rhythm of Nigerian bargaining, and mirrors the warmth of human vendor interactions was not a nice-to-have — it was a load-bearing requirement. An agent that sounds like a Western customer service bot would not survive contact with a Nigerian customer for more than two messages. Localization at the cultural layer is mandatory.

Meet users where payment already lives.

Every instinct in fintech points toward building new payment rails. Kasi's instinct pointed toward existing ones. The bank transfer is not a problem to be solved — it is a trusted behavior to be integrated around. Products that fight user habits lose. Products that extend user habits win.

The side-hustle economy is a serious market.

Nigeria has tens of millions of informal vendors operating on social media. These are not hobbyists — they are builders operating under extreme resource constraints with extraordinary hustle and zero safety net. Building for them means pricing models that flex with their reality, UX that respects their time, and tools that work as hard as they do.

Agentic AI's first job is trust, not capability.

The hardest problem in deploying an autonomous agent is not technical — it is psychological. A vendor must trust that the AI will not embarrass them, undersell their products, or confuse their customers. Every design decision in Kasi — the floor price enforcement, the branded invoice, the warm negotiation tone — is in service of that trust. Capability without trust is a demo. Capability with trust is a business.