Wavu: Turning WhatsApp Conversations Into Online Sales

Most e-commerce advice assumes a customer who browses a catalogue, adds to cart, enters card details and checks out. That customer exists in Africa, but they are the minority. The majority sees a product on Instagram or TikTok, opens WhatsApp, types hi, how much? and expects a conversation.

That is not a broken funnel. It is a different funnel - and it converts extremely well when it is answered. The trouble is that it is enormously labour-intensive, wildly inconsistent between staff, and it collapses the moment volume rises or the person holding the phone goes to sleep.

Wavu is upeo.ai's conversational online selling platform, built for exactly this reality: businesses whose customers arrive in a chat window rather than a checkout page.

Why conversational selling wins in Africa

Trust is established in the conversation

Buying from a business you have never met requires reassurance. Customers ask questions not because the listing lacked information, but because a human-feeling exchange is how they decide you are real.

Payment and delivery are negotiated, not fixed

Pay on delivery or pay now? Which rider? Can it reach Kisii by Friday? Is there a discount for two? These are conversational variables that a rigid checkout page cannot express.

The catalogue is fluid

Small businesses sell what they have today. Stock changes faster than any product page gets updated, so the customer asks rather than trusts the site.

Data cost and habit

WhatsApp is where people already are. Asking them to install an app or load a heavy site to buy a pair of shoes adds friction to a decision that was already made.

Where conversational selling breaks down

• Response speed. Reply in five minutes and you often have a sale. Reply in five hours and you have a read receipt.
• After hours. A large share of shopping messages arrive between 8pm and midnight, when nobody is on the phone.
• Inconsistency. Three staff, three prices, three tones, three delivery promises.
• Nothing is recorded. Orders in chat threads cannot be counted, fulfilled reliably, or reported on.
• Abandoned conversations. Customer asks the price, goes quiet, and nobody ever follows up. This is the single biggest loss in the channel.
• Volume ceiling. One person can hold maybe thirty real sales conversations a day. Growth beyond that means hiring, which means more inconsistency.

What Wavu does

Answers instantly, from your real catalogue

Price, availability, sizes, colours, delivery options and timelines - answered from live business data, at any hour, in the language and register the customer used. No let me check followed by silence.

Sells rather than just responds

A good salesperson asks what the customer is actually looking for, suggests the alternative when the exact item is finished, and mentions the item that pairs naturally with what is in the basket. That is a repeatable pattern, and AI executes it consistently instead of only when the staff member is having a good day.

Turns the conversation into a real order

This is the core of it. A rambling chat becomes structured data your operation can act on:

{
  "customer": "+254 7XX XXX XXX",
  "items": [
    { "sku": "SNK-402-BLK", "name": "Canvas sneakers", "size": "41", "qty": 1 }
  ],
  "delivery": { "town": "Nakuru", "method": "courier", "eta": "2 days" },
  "payment": { "method": "mobile money", "status": "awaiting" },
  "source": "instagram_dm",
  "confidence": "high",
  "needs_human": false
}

From there the workflow runs on its own: order created, payment instruction sent, confirmation issued, stock decremented, fulfilment notified. That is the AI Workflow Automation layer doing what it is built for - actions triggered without manual intervention.

Follows up on quiet conversations

The customer who asked the price and never replied is not lost. A single well-timed, non-annoying follow-up recovers a meaningful share of these. Nobody does this manually at volume. Wavu does it every time.

Escalates when it should

Unusual requests, complaints, bulk negotiations, anything ambiguous or emotionally charged goes to a human with the whole conversation attached - not restarted from scratch, which is the thing customers hate most about chatbots.

Gives you a real picture of the channel

Conversations started, conversion rate, average order value, most-asked questions, most-requested out-of-stock items, best-performing traffic sources, busiest hours. The questions customers ask most often are the fastest product roadmap you will ever get.

What stays human

• Bulk and wholesale negotiation. Volume pricing is a margin decision.
• Complaints, returns and refunds. Always a person, immediately.
• Anything with a payment dispute attached.
• High-value or bespoke orders. Where the buyer expects, and deserves, a relationship.
• Judgement calls on goodwill. Replacing a damaged item is a business decision, not a rule.

The goal is not to replace your salespeople. It is to stop them spending their day typing the price of the same item forty times, so they can handle the conversations where their judgement actually changes the outcome.

Where conversational selling sits on the NGAZI ladder

NGAZI - Swahili for ladder - treats AI adoption as five rungs, each earning the next.

Stage	What it looks like in online selling
Stage 0 - paper and memory	Orders in chat threads, stock in your head, no record you can count.
Stage 1 - a fast assistant	You use AI to write captions and product descriptions.
Stage 2 - it reads your files	AI answers from your real catalogue, prices and delivery policy, verifiably.
Stage 3 - it gets things done	AI holds the sales conversation, creates orders, confirms payment and triggers fulfilment.
Stage 4 - it sees the whole picture	AI reads accumulated sales history for repeat patterns, product pairings and demand signals.

Wavu is a Stage 3 product, and that is exactly why the preparation matters. If your catalogue and prices are not digital and current, no conversational AI can sell them accurately. Fix the catalogue first. It is a week of work that unlocks everything after it.

How to start

1. Get your catalogue right. Item, SKU, price, available quantity, sizes or variants, photos. Current, not aspirational.
2. Write your delivery and payment policy down clearly. Areas, costs, timelines, methods. Ambiguity here becomes a bad automated answer.
3. Collect your twenty most common questions from existing chats and write the correct answer to each.
4. Start with answering and quoting. Let humans still close.
5. Add order creation and payment confirmation once accuracy is proven.
6. Turn on follow-up for quiet conversations. Usually the fastest measurable gain.
7. Read the transcripts weekly. Every awkward reply is a fix.

Your customers, your data

The customer list and buying history built in these conversations is one of the most valuable assets a small commerce business owns. It should stay in your environment, under access control with full audit trails, and never move without your authorization. Conversational data is personal data - treat it that way, and expect your vendor to as well.

Talk to us

Wavu is part of the upeo.ai suite, built in Nairobi alongside GariSuite for car dealerships, UpeoRetail for retail and ISHI for real estate. It connects to WhatsApp, email and web, works against your live business data, automates replies, order creation and confirmations, and keeps your team in control of the decisions that need a person. It is trained on your specific business, not a generic sales bot.

• Email: hello@upeo.ai
• Phone: +254 116 888 777
• Web: https://www.upeo.ai

Open your business WhatsApp and scroll back one week. Count the conversations that ended with a customer question and no answer. That is the revenue sitting in your inbox right now.