Skip to content
AI

What AI actually looks like in an African SME

Less GPT hype, more anomaly detection. What we learned deploying Isolation-Forest models on shop-floor data.

By Ayoolumi Melehon · July 08, 2026 · 1 min read
What AI actually looks like in an African SME

Most AI conversations in Nigerian business right now are about LLMs. ChatGPT this, Gemini that. There's a place for that. But the AI that actually earns its keep in a small-to-mid-sized Nigerian business is much less glamorous.

It's anomaly detection.

We built the AandA Smart BI platform to run our own operations. Twenty-one modules, hash-chained audit log, real-time analytics — all of it. But the module that pays for itself month after month is the one nobody talks about at conferences: an Isolation Forest model that watches every cash reconciliation event and flags the ones that don't look like the historical pattern.

Here's how it works in practice. The cashier closes their shift. They count the cash drawer, log the number, and expect it to reconcile with what the till system says. Usually the two agree. Sometimes there's a ₦500 discrepancy nobody can explain but everyone accepts. And every once in a while, there's a ₦12,000 gap.

The old way — spreadsheets, gut feeling — would let that ₦12,000 hide inside a stack of small ₦500 discrepancies until someone got suspicious enough to audit. The Isolation Forest doesn't wait. It reads the amount, the shift, the day of the week, the seasonality, and produces a “how surprising is this?” score. Below a threshold, it goes to Telegram immediately. The CEO knows within 90 seconds.

We've been running this in production for 14 months. Two hard lessons:

  1. The threshold isn't fixed. Seasonality is real. December is not August. The model self-tunes against a rolling baseline so it doesn't spam alerts during known-busy periods.
  2. Alerts must include the context. “Anomaly detected” is useless. “Anomaly detected: variance ₦8,400, staff member Aisha, shift 12–8, expected variance ₦0–₦2,000” — that's actionable in one glance.

Deployed correctly, an anomaly model like this saves more money than it costs in the first quarter. Not because it catches fraud (though it does, occasionally). Because it makes the team more careful — they know something is watching.

That's the AI story that matters in a Nigerian SME right now. Not chatbots. Watchdogs.

Let’s work together

Have a project
in mind?

Tell us what you need. We come back with a written scope and quote within 48 hours.