What AI Would Actually Change for a Retailer in Ogun State

I use a retailer in Ogun State as a thought exercise when I am training commercial teams. Ask most people what AI would do for that business and you get the popup. Hello, how can I help you today. That is the front door, and it is the least interesting room in the building. The version that would change their month looks nothing like a chatbot. It is a system that knows which products are about to go out of stock based on the weather, salary week and last year's pattern. Pricing that adjusts across thousands of SKUs because a competitor moved overnight. A recommendation engine that learns each customer well enough to send the right message on the right day in the right language. Fraud detection catching a stolen card before the order ships. The warehouse rerouting itself because a delivery van broke down in Lekki traffic. A finance team closing the month in two days instead of ten because reconciliation runs itself. None of that is futuristic. All of it is running somewhere already. Very little of it is running in African FMCG and retail, and the reason is rarely the technology. ## Salary week is a forecasting signal Spend time in this sector and you learn quickly that the patterns are real, specific and local. Demand moves with salary week. It moves with school fees season, with Ramadan and Christmas, with rainfall, with fuel price changes that alter how far a consumer will travel to shop. A forecasting model built on last year's national averages will miss all of it. A model built on your own sell out data, with the local signals included, will not. This is the clearest opportunity in the sector and it is unglamorous. Better forecasting means less working capital tied up in the wrong stock, fewer out of stocks on the lines that actually move, and less waste on short dated product. Those are the three numbers that decide whether a distribution business makes money. ## Where the money is, in order **Demand forecasting and replenishment.** Start here almost always. The data exists in the sales system, the commercial value is measurable, and the result shows inside a quarter. **Distribution and route economics.** Most distributors in the region cannot tell you profitability by outlet. They can tell you volume. Those are different, and the gap is usually where margin disappears. Working out which outlets cost more to serve than they return changes the route plan. **Trade spend.** Enormous sums move through promotions and trade support with limited visibility on what worked. Attributing uplift properly is a modelling problem, and it is worth more than most digital marketing budgets in the same business. **Customer level personalisation.** Real, but later. It needs the first three working to be worth much. ## The thing sitting underneath When I train commercial teams at companies in this sector, the same pattern surfaces. Somebody in trade marketing is pasting a distributor's sales file into a free tool to build a summary for a review. Somebody in category management is dropping competitor pricing and volume data into a chatbot to get a chart. I call it BROAI. Bring Your Own AI, after the Bring Your Own Device problem. It is not laziness. It is a capable person with a deadline, no approved tool, and no guidance. The fix is an approved option, not a warning. ## What good looks like One strategy rather than a tool in each department. Approved tools. A named owner for each use case who carries a number. Company data staying inside systems the company controls. SWITCH, the six phase framework I use with clients, starts with Scan and Score because most businesses in this sector either overestimate their data quality or pick the use case that sounded impressive rather than the one that pays. Here is the question I would put to a commercial director this week. If your forecast was ten per cent more accurate, what would that be worth in working capital? If you can answer that, you already know which project to fund first. ## Frequently asked questions **Where should an FMCG company in Africa start with AI?** Demand forecasting. The data already exists in the sales system, the business case is easy to state, and the result appears within a quarter. **Does AI work when distribution data is messy?** Messy data is normal in this sector and is not a reason to wait. It is a reason to start with one region or one category rather than the whole business. **What about small retailers who do not have a data team?** The first version is usually a spreadsheet exercise done properly rather than a system purchase. Pattern before platform. **How does AI help with trade spend?** By separating promotions that drove incremental volume from those that discounted sales that would have happened anyway. Most businesses cannot currently tell the difference. **Is personalisation worth it in African retail?** Eventually. Forecasting, distribution economics and trade spend pay back faster and build the data foundation personalisation needs. *Stanislaus Martins has trained marketing, sales and ecommerce teams at companies including Nestlé, Nigerian Breweries, Jumia and Konga, and advises enterprise teams across Sub Saharan Africa on practical AI adoption. Formats are on the [speaking and training page](/speaking), or start a conversation on the [work with me page](/work-with-me).*