Walk into a hardware shop in Nairobi on a Tuesday afternoon and ask the owner which of their items made the most money last month. Not sold the most units - made the most money, after what they paid the supplier.
Most owners cannot tell you. Not because they are careless, but because the question is genuinely hard. The sales are in one place, the purchase prices are in another, and the answer needs both. So the owner does what everyone does: they go on feel. And feel is usually right about the top three items and wrong about everything below that.
That gap - between the data a shop already has and the questions the owner actually wants answered - is what Rai exists to close.
What Rai actually is#
Rai is Swahili for advice or counsel, which is exactly the job. It is an AI agent that lives inside UpeoRetail, and you talk to it the way you would talk to a very well-briefed accountant who never sleeps and has read every line of your books.
You ask: "which suppliers raised their prices on me this quarter?" You get a table, a chart, and a sentence explaining what it means. You ask: "what should I be worried about this week?" You get three specific things, each with the number behind it.
The important word there is behind. Rai does not tell you that your margins are slipping. It tells you that your margin on power tools has gone from 22% to 17% since April, that it happened after your main supplier's price moved, and that your retail price has not moved with it. Then it tells you what a corrected price would look like.
The four things it does every day#
Reports you did not have to build. Ask in plain language, get a table and a chart you can export to PDF or Excel. If it is a question you will ask again, save it and rerun it whenever you want. No one has to configure anything.
Advice with its receipts attached. Reorder decisions, pricing moves, dead stock to clear, suppliers to consolidate. Always with the figures. "Clear these 12 items - KES 340,000 tied up, zero sales in 90 days." That is a sentence you can act on this afternoon.
A morning brief. A short digest that arrives on schedule: what moved, what is at risk, and three things worth doing. This is the part owners tell us they would miss most. Not because any single brief is dramatic, but because reading it becomes a habit, and the habit is what changes how a shop is run.
Alerts when something breaks. The brief is time-based; alerts are not. A fast-moving item about to run out, a margin sliding on a category, a supplier's price jumping, a till that does not reconcile - those fire the moment they happen, because a loss you catch on Friday is cheaper than one you read about on Monday.
It comes to you, not the other way round#
An analyst you have to log in to visit is an analyst you consult twice a month.
So Rai works over WhatsApp. You ask a question the way you would message a colleague, and the answer comes back in the same thread. The daily brief arrives there too. If the network is bad, there is SMS fallback through the gateway we already run.
And it answers in the language you actually think in. English, Swahili, Sheng - whichever you typed in. This sounds like a small feature. It is not. The difference between an owner who checks their numbers because it feels natural and one who does not because it feels like homework is often just the language the numbers arrive in.
The part we are strictest about#
Rai reads your business freely. It does not write to it freely.
Everything that would change your data - drafting a purchase order for a low-stock item, turning a photo of a supplier invoice into a purchase entry - stops and waits for you. Rai prepares it, shows you exactly what it will do, and does nothing until you approve. Every one of those actions is logged, scoped to what that user is allowed to touch, and reversible.
An AI that can quietly change your stock records is not a helpful assistant. It is an unattended liability.
We are unusually firm about this because the failure mode is so ugly. An AI that reads and gets something wrong wastes five minutes of your time. An AI that writes and gets something wrong corrupts your books, and you find out weeks later when nothing reconciles.
Where it goes next#
Rai launches inside UpeoRetail because that is where the shops already are. But it was designed from the first day to outgrow that. The part of Rai that does the thinking is deliberately kept separate from the part that talks to your specific system - which means adding a different kind of business later is new plumbing, not a rewrite.
The destination is a dependable AI business analyst that any business can point at whatever they already run. The route there starts with Kenyan retail, because that is a market we know, serving owners whose questions we have been listening to for years.
If you run an electronics, hardware or spare-parts shop and you have ever wanted to just ask your business a question, that is the whole idea.
